# Phil Morton > Home of Desk Notes, a free newsletter where I share what I've learned about product design, research, AI and more. Public Ghost content for AI and LLM tooling. This file includes a bounded export of public pages first, then recent public posts. Append `.md` to any post or page URL to get the content in Markdown (for example, `/example-post.md`). ## Pages ### About URL: https://www.philmorton.co/about/ Last updated: 2025-04-11T12:39:34.000Z ![A photo of Phil Morton](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/02/Phil-Morton-portrait.jpeg) Hi 👋 I’m Phil Morton, a product design and research leader dedicated to helping teams create better products and services by putting customers at the heart of the design process. In the last 15 years, I’ve worked with the likes of PlayStation, Adobe, HSBC, Heineken, and Virgin Money. My recent work at [Foolproof](https://foolproof.co.uk/?ref=philmorton.co) has included leading a 17-person design and research team at Shell working on an e-commerce platform with $20bn annual revenue, and coaching NatWest's research team. Outside of client work, I have experience in hiring, onboarding, training, career development and knowledge management. I'm proud of the role I’ve played in mentoring UX practitioners, helping them grow into confident and capable professionals. I started my career as a researcher and have conducted over 2,000 hours of face-to-face research in 14 countries. I write **Desk Notes**, a newsletter where I share the kind of insights, habits, and thought processes you’d pick up if you sat next to me at work, covering research, design, leadership and career development. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Welcome URL: https://www.philmorton.co/welcome/ Last updated: 2026-01-27T21:25:50.000Z _This page is for subscribers only._ ### Projects URL: https://www.philmorton.co/projects/ Last updated: 2026-06-17T13:37:34.000Z [Pegs OutLeading product design and research teams. Sign up for my free newsletter, Desk Notes.![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/Phil-Morton-square-de89cfb6-6235-4877-bf4e-71b1f0f2f9d9.jpeg)Phil MortonPhil Morton![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/IMG_1179-1-1-19e31301-d9f8-472b-8a29-519519e7f1c3.jpeg)](https://www.philmorton.co/projects-pegs-out/) ### Pegs Out URL: https://www.philmorton.co/pegs-out/ Last updated: 2026-06-17T13:41:33.000Z _No content available._ ### Pegs Out privacy policy URL: https://www.philmorton.co/pegs-out-privacy-policy/ Last updated: 2026-06-17T14:08:03.000Z **Version 1.0** | **Effective: 17 June 2026** > **In short:** Pegs Out keeps your data on your phone. There are no accounts, no tracking, no ads and no server. The only thing that leaves your phone is the location needed to fetch your weather forecast. ## Who are we? Pegs Out is made by Phil Morton, an independent developer based in the United Kingdom. If you have any questions about this policy or your privacy, email us at philip.morton@gmail.com. ## What does Pegs Out collect? Nothing about you. There is no account and no sign-in, so we never ask for your name, your email address or any other personal details. To be completely clear, Pegs Out does **not**: - track you, anywhere, ever; - show ads or work with advertisers; - use analytics or any usage-tracking tools; - run a server that holds your data; - sell or share your data with anyone for marketing. ## Where does your location go? Pegs Out needs to know where your washing line is so it can fetch the right weather forecast. iOS handles your location on your device. To turn that into a forecast, two things happen, both over a secure (HTTPS) connection and both in real time: - **The weather forecast, from Open-Meteo.** We send your coordinates to Open-Meteo, a free, non-commercial weather service run as an open-data project in Germany. We send no name, no account and no device identifier. Because the request travels over the internet, Open-Meteo can also see your device's IP address. Open-Meteo's own terms say it may keep request logs (which can include the coordinates and the IP address) for up to 90 days for technical reasons, and that these are not linked to your identity. - **Place names, from Apple.** When you search for a place by name, or ask Pegs Out to use your current location, iOS sends that search text or those coordinates to Apple's location services to look up a place name. This is handled by Apple, under Apple's own privacy policy. | Who | Why | What we send | | ---------- | --------------------- | --------------------------------------------- | | Open-Meteo | Your weather forecast | Coordinates and (unavoidably) your IP address | | Apple | Place-name lookup | Coordinates, or the place name you type | We keep none of this off your phone. If a future version of the app refreshes your forecast in the background, the same applies. Weather data is provided by [Open-Meteo.com](https://open-meteo.com/?ref=philmorton.co) under the [CC BY 4.0 licence](https://creativecommons.org/licenses/by/4.0/?ref=philmorton.co). ## What stays on your phone? Everything else stays on your device and never leaves it: - the location you choose; - your washing-line and laundry settings (sun, shelter, fabric and so on); - the cached 10-day forecast. All of it lives on your phone and is deleted when you delete the app. ## Your rights Because we hold no personal data about you, there is nothing for us to look up, change or delete. If you have any question about your privacy, email us at philip.morton@gmail.com. If you live in the UK and are unhappy with how we have handled a privacy matter, you can contact the Information Commissioner's Office (ICO) at [ico.org.uk](https://ico.org.uk/?ref=philmorton.co). ## Changes to this policy If we change this policy, we will update the version number and the date at the top and publish the new version here. Previous versions are kept in our source history. ## Contact us Questions about this policy or your privacy? Email us at philip.morton@gmail.com. ### Pegs Out support URL: https://www.philmorton.co/pegs-out-support/ Last updated: 2026-06-18T17:10:13.000Z **Version 1.0** | **Updated: 18 June 2026** Pegs Out tells you whether it is worth hanging your washing outside. Not just "will it rain", but whether, given the daylight and conditions left, this load will actually dry, or whether you are better off using the dehumidifier indoors. ## Getting help Pegs Out is made by Phil Morton, an independent developer based in the United Kingdom. If something is not working, or you have a question or a suggestion, email philip.morton@gmail.com and you will get a reply. ## Frequently asked questions ### Why does Pegs Out ask for my location? To fetch the right weather forecast for where your washing line is. Your location is handled on your device by iOS, and the only thing that leaves your phone is the coordinates needed to get the forecast. There are no accounts and no tracking. The full detail is in the [privacy policy](https://www.philmorton.co/pegs-out-privacy-policy/). ### Why is the drying estimate given in hours rather than an exact time? Because drying depends on temperature, humidity, wind and sun, and forecasts are not perfect. The part you can trust is the relative ranking: that Tuesday is a better drying day than Thursday, or that the morning beats the afternoon. The exact number of minutes is a best estimate, not a promise, so Pegs Out is honest about that rather than pretending to a precision it does not have. ### How do I change the sun, shelter or fabric settings for my line? Open the Settings tab. The sunny or shaded and sheltered or exposed choices for your line are one-time picks you can change at any time, along with the typical wash you hang out. There is also a single control to nudge the estimates if they consistently feel too fast or too slow for your line. ### Why does a day further ahead not show a precise estimate? Weather forecasts get less certain the further out they go. Past about four days, Pegs Out deliberately shows the drying potential in general terms rather than as precise minutes, so you are not given false confidence in a number the forecast cannot support. ## Privacy Pegs Out keeps your data on your phone. There are no accounts, no tracking, no ads and no server. You can read the full [privacy policy](https://www.philmorton.co/pegs-out-privacy-policy/). ## Contact Questions, problems or suggestions? Email philip.morton@gmail.com. ### Pills In privacy policy URL: https://www.philmorton.co/pills-in-privacy-policy/ Last updated: 2026-07-29T11:52:25.000Z **Version 1.0** | **Effective: 29 July 2026** > **In short:** Pills In keeps your medication data on your phone. There are no accounts, no tracking, no ads and no server. The app never sends anything to us, because there is nowhere for it to send it to. ## Who are we? Pills In is made by Phil Morton, an independent developer based in the United Kingdom. If you have any questions about this policy or your privacy, email us at philip.morton@gmail.com. ## What does Pills In collect? Nothing about you. There is no account and no sign-in, so we never ask for your name, your email address or any other personal details. To be completely clear, Pills In does **not**: - track you, anywhere, ever; - show ads or work with advertisers; - use analytics or any usage-tracking tools; - run a server that holds your data; - sell or share your data with anyone, for marketing or for anything else. We also do not read your Apple Health data. Pills In keeps its own record of your medications and does not look at anything Health holds. ## Does anything leave your phone? The app itself makes no internet connections at all. There is no server behind Pills In, no third-party code inside it, and nothing it does causes your medication data to be sent to us or to anyone else. Three things are worth explaining properly, because "it all stays on your phone" is not quite the whole story. ### 1\. Calendar sync, which you turn on yourself Pills In can add your reorder and run-out dates to a separate "Pills In" calendar, so you can see them alongside everything else in your diary. **This is off when you install the app, and only ever happens because you switch it on.** When it is on, each event carries, in its notes: the name of the medication, its reorder-by date, its run-out date and the number you have left. Two things to be aware of: - **The "hide medication names" setting hides the event title, not the event.** With it on, an event reads "Reorder a medication" instead of "Reorder Rivaroxaban". The notes still contain the medication name and your current count, because an event with nothing in it would be no use to you. - **If the calendar is one of your iCloud calendars, Apple syncs it.** Pills In adds its calendar to your iCloud account where you have one, so the dates reach your iPad, Mac and Apple Watch. That means Apple's servers and all of your devices hold those events, in the same way they hold the rest of your calendar. That is Apple handling your calendar under Apple's privacy policy. Nothing goes to us. You can turn calendar sync off at any time in Settings, and Pills In removes the calendar it created. ### 2\. Your iPhone's own backup If you have iCloud Backup switched on, your iPhone backs up Pills In along with every other app on your phone, and the backup includes the medications, schedules, counts and history the app holds. This is Apple's backup of your device, it is encrypted, and it is entirely under your control in iOS Settings. We cannot see it and we are not sent a copy. We have deliberately left it that way. If we excluded Pills In from your backup, setting up a new iPhone would leave you with an empty app and no way to get your medication history back. ### 3\. Reminders on your lock screen A Pills In reminder never names a medication in the text you can see without unlocking your phone. It says something like "Time to log your 08:00 medications" and how many are due. The names appear only when you expand the reminder, which needs your phone unlocked. ## Diagnostic logs Like most apps, Pills In writes technical notes about what it is doing to your iPhone's system log. Medication names, counts and any notes you have written are marked as private, so they are hidden there. Those logs stay on your phone. They only ever go anywhere if you choose to send a diagnostic report to Apple or to us yourself. ## What stays on your phone? Everything else stays on your device: - your medications, their strengths, forms and schedules; - how many you have left, and every restock or correction you have made; - your record of doses taken, skipped and snoozed; - your reminder times, snooze setting and other preferences. All of it lives on your phone and is deleted when you delete the app. You can also erase it at any time using **Reset app** in Settings. **One exception.** If you had calendar sync turned on and you delete Pills In without first switching sync off, the "Pills In" calendar stays in your Calendar app, because a deleted app cannot tidy up after itself. You can remove it in the Calendar app like any other calendar. ## A note on security Your Pills In data is protected by your iPhone in the same way as your other apps' data, and by your passcode. So that you can log a dose straight from a reminder on the lock screen, the app's data is readable by the phone without unlocking it first, exactly as your calendar and reminders are. Anyone with your unlocked phone can see your medications, as they could see any other app. ## Your rights Because we hold no personal data about you, there is nothing for us to look up, change or delete. If you have any question about your privacy, email us at philip.morton@gmail.com. If you live in the UK and are unhappy with how we have handled a privacy matter, you can contact the Information Commissioner's Office (ICO) at [ico.org.uk](https://ico.org.uk/?ref=philmorton.co). ## Medical disclaimer Pills In is a reminder tool and does not provide medical advice or dosing. Always follow your doctor and pharmacist's instructions. ## Changes to this policy If we change this policy, we will update the version number and the date at the top and publish the new version here. Previous versions are kept in our source history. ## Contact us Questions about this policy or your privacy? Email us at philip.morton@gmail.com. ### Pills In support URL: https://www.philmorton.co/pills-in-support/ Last updated: 2026-08-03T13:25:45.000Z Pills In reminds you to take your medication and tells you when you are going to run out of it. It keeps count of what you have left, works out how quickly you are getting through it, and gives you a run-out date and a reorder-by date for every medicine, so you know when to go to the pharmacy. ## Getting help If something is not working, or you have a question or a suggestion, email philip.morton@gmail.com and you will get a reply. ## Frequently asked questions ### Why does Pills In ask to send notifications and set alarms? Notifications are how the reminders reach you, and how you log a dose without opening the app. The separate alarm permission is for the urgent alarm that can ring through silent mode if you have not responded to a reminder, which you can turn off in Settings if you would rather it did not. Both are asked for by iOS and can be changed at any time in the iOS Settings app. ### How does Pills In work out when I will run out? From the doses you actually take. It adds up how much of a medicine each of its scheduled times uses, projects those doses forward day by day, and finds the day the last one comes out of the box. That day is the run-out date. The reorder-by date is a few days earlier, so you have time to collect a new prescription, and it counts in working days so a weekend does not eat into your notice. Because it works from your real schedule rather than an average, medicines taken monthly or every few days are forecast just as accurately as a daily tablet. You can see the working behind any medicine by opening it from the Medications tab. ### My reminders have stopped arriving Nearly always this is a permission that has been switched off, so check in this order: 1. iOS Settings, then Notifications, then Pills In, and make sure notifications are allowed. 2. Check Focus modes and Scheduled Summary, both of which can hold notifications back. 3. In Pills In, open Settings and check that Medication reminders is on, and that the times under Time slots are the ones you expect. 4. Make sure the medicine has a schedule. A medication with no times set has nothing to remind you about. If reminders are still not arriving after that, please email me. A reminder that does not fire is the worst thing this app can do, and I would rather hear about it than not. ### My stock count does not match what is in the drawer Counts drift, and there is nothing wrong with correcting them. Open the medicine from the Medications tab and use **Set current stock** to enter the number you actually have. Use **Add stock** instead when you have collected a new prescription, which adds to the running total rather than replacing it. If you would rather not track the supply of a particular medicine at all, **Stop tracking stock** turns off its forecast and low-stock warnings without deleting anything else. ### Can I take the same medicine at more than one time of day? Yes, and it still comes out of one count. A tablet taken morning and evening is one medicine with two scheduled times, drawing from the same box, exactly as it works in real life. Each time can have its own quantity and its own frequency, so a medicine taken every morning and every other evening is a single entry rather than two. ### How do I change the reminder times, the snooze length or the reorder warning? All three live in the Settings tab. - **Time slots** sets the times of day reminders fire. These are shared across all your medicines rather than set per medicine, so changing Morning changes it everywhere. - **Allow me to snooze for** sets how long the Snooze action on a reminder delays it. - **Reorder reminders** sets how many days before running out you would like to be told. ### What gets added to my calendar? Only if you switch **Calendar sync** on in Settings, which is off when you install the app. When it is on, Pills In creates a separate "Pills In" calendar holding your reorder and run-out dates, and your trip dates if you use trips. There is a setting to hide medication names from the event titles. If that calendar is one of your iCloud calendars, Apple syncs it to your other devices in the same way as the rest of your diary. The [privacy policy](https://www.philmorton.co/pills-in-privacy-policy/) explains exactly what each event contains. Turning calendar sync off removes the calendar Pills In created. ### What does Pills In deliberately not do? - **Half tablets and as-needed medicines** are not supported in this version. A medicine taken only when you need it has no regular rate, so there is no honest way to forecast when it will run out. - **Apple Health** is not read or written. Pills In keeps its own record of your medications. - **There is no account and no cloud sync.** Your medication data lives on your phone. ## A reminder tool, not medical advice Pills In is a reminder tool and does not provide medical advice or dosing. It never suggests what to take, how much to take, or whether to change anything. Always follow your doctor and pharmacist's instructions. ## Privacy Pills In keeps your medication data on your phone. There are no accounts, no tracking, no ads and no server. You can read the full [privacy policy](https://www.philmorton.co/pills-in-privacy-policy/). ## Contact Questions, problems or suggestions? Email philip.morton@gmail.com. ## Posts ### A side project is the fastest way to upskill in the age of AI URL: https://www.philmorton.co/a-side-project-is-the-fastest-way-to-upskill-in-the-age-of-ai/ Last updated: 2026-07-22T09:18:30.000Z The one piece of career advice I’m giving everyone right now is to get a side project. [AI is changing how software gets made](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/), and the best way to upskill yourself (and your team, if you have one) is to **design and build something end-to-end**. In the last nine months I’ve learned more, and at a faster pace, than in any period of my career. That’s because I’ve been building things. I made a web app that helps parents deal with school emails (more on that in a future newsletter). [I built Pegs Out](https://www.philmorton.co/how-i-built-pegs-out-my-first-ios-app/), an app that helps you work out if you should hang your laundry out to try today. And I have a second iOS app which will be out in the next few weeks. Every one of these has taught me more about how software gets made (and how AI will change that) than anything else I’ve done. You can watch videos, listen to podcasts and read articles, but **nothing compares to what you learn by making**. ## The bar is far lower than you think Most UX people don’t have a portfolio of side projects because they can’t build software on their own. Until recently, the barrier was just too high. You had to know how to code, keep up with all if the new frameworks and find the time to actually do it. So the only designers with side projects tended to be the ones who already had a technical background. That barrier is broadly gone. You still benefit from having technical fluency, but **the idea that you can’t make software unless you know how to code is simply not true anymore.** The time barrier has shrunk too. Building something small used to take weeks of evenings. Now it might take one or two. I built Pegs Out in two weeks, as a busy parent. **The cost is also relatively low:** all you really need is a Claude or ChatGPT subscription, about ÂŁ20 a month. Nearly every web service (GitHub, Vercel, Supabase, etc.) has a generous free tier, and for a side project you don’t need the paid plans. If you want to ship an iOS app, you’ll need to pay Apple’s annual developer fee, but that’s about it. For people in an industry that pays above average, that’s not an insurmountable investment given how much you get back in learning. ## Go beyond Lovable and co. Most UXers dip their toes into making software by trying something like Lovable, Replit or Figma Make. Type in a prompt and out comes a website – magic! You can get quite far with these, but **you’ll quickly reach a local maximum** of what you can learn. **Tools like Lovable abstract away almost everything that’s happening under the hood.** You don’t have to worry about things like version control, but you also don’t learn about them. You get the result without understanding how things are made. What I’m suggesting goes a step further: use tools like Claude Code, Codex or Cursor to build the real thing and do more of it yourself. When you’re working a little closer to ‘the metal’, then you learn more. Lovable is nothing like how products get built commercially, whereas using Claude Code to build an iOS app is. Tools that do everything for you aren’t going to teach you much. ## How to get started If you’ve never built a digital product, how would you know where to begin? A proper guide is a newsletter in its own right, but the general approach is... **Start with a small idea.** Not a side hussle or a business, just a little utility or helper, ideally one that uses an existing free API. Solve a small problem in your own life. This is what [Pegs Out](https://apps.apple.com/gb/app/pegs-out/id6778139891?ref=philmorton.co) is. **Then resist the magical thinking that tools like Lovable encourage.** Don’t open Claude and ask it to build the whole thing straight away. That skips a lot of steps which as a UXer, you know matter. **Do your discovery and a bit of design first.** When I started on Pegs Out, I got Claude to research the physics of how clothes dry and best practices for building an iOS app, because I had no idea on either. Some people think the ‘double diamond’ is outdated, but I think it’s still worth doing some of that first diamond so you know what you’re building before you start writing code. **When it’s time to build, don’t ask it to just make the whole thing.** Ask it how you should make it: I*’m not technical, I want to build this, what are my options? What would a developer do here? What does every project need that I don’t know about?* It’ll mention something like GitHub, so you ask what GitHub is, and why you need it. That back and forth is where you learn the most. **Build it *with* AI, rather than getting it to do everything for you.** ## Don’t let “I’m not technical” hold you back I think a lot of people get intimidated by this stuff: *I’m not a developer, this is too technical*. It is slower to get started because you have so many concepts to learn, but **every time you get stuck, you can just ask AI what to do next**. I’ve got a computer science degree and nine months of building things with AI, and I still paste screenshots into Claude and ask *“what is this, what do I do now?”* all the time. Building software with AI isn’t hard. **The difficulty comes from the discomfort of working on something where you don’t know what you’re doing.** Just remember that you can take it at your own pace and ask AI as many ‘stupid’ questions as you like – it’s a teacher with infinite patience. ## Have the idea, then have a go Not everyone has the time, energy or interest for a side project, and that’s fine. Maybe now isn’t the right time. But stay open to it and **next time an idea for some little app pops into your head, don’t assume it’s out of reach.** For years it was, for most people in UX. It isn’t any more. And unlike a work project, whatever you build is yours. It can be something genuinely useful in your own life, and **it’s fun and rewarding** to make. You’re making something you want to exist. Whether you’re an IC or a team leader, this is the fastest way to build the skills that we’re all going to need in the years to come. All you need to do is to **take that first step** and see how much you learn. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why is AI bad at design? URL: https://www.philmorton.co/why-is-ai-bad-at-design/ Last updated: 2026-07-15T09:59:58.000Z **If there’s anything that AI is particularly good at, it’s writing code.** [The whole software engineering profession has been turned upside down](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/) as more and more code is written by agents, with humans supervising and reviewing. If LLMs have product-market fit in anything, it’s software engineering. Yet if you’ve tried AI design tools like Figma Make, Claude Design and so on, you might be disappointed. No matter how much context you give it, **AI tends to produce design work that looks plausible, but is very generic** and on further inspection full of (obvious to us humans) flaws. Having used AI design tools on a few projects recently, it left me wondering, *“is it me or is AI just bad at design?”* Turns out that, no, it’s not just me. **There are some fundamental technical reasons why AI is nowhere near as good at designing** as it is at coding. ## What you mean by ‘bad’? If you haven’t tried these tools in anger, here’s one example from [my iOS app, Pegs Out](https://www.philmorton.co/how-i-built-pegs-out-my-first-ios-app/). I recently added the ability to see *why* a particular day is good or bad for drying your laundry outside, based on the underlying weather conditions. It tells the story of the day in a style similar to a data journalism article from the FT, Economist, etc. This is what Claude originally designed: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/07/Simulator-Screenshot---iPhone-17---2026-07-03-at-20.16.26.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/07/Simulator-Screenshot---iPhone-17---2026-07-03-at-20.16.37.png) I’m sure you can see some issues... - Charts don’t have titles so you don’t know what they’re showing you. - There are no y-axis labels. - What do the yellow shaded areas mean? - What is the blue dotted line for? - The headers “Air” and “Energy” are too abstract to understand. - The data visualisation is flat and has no depth to it. After a few rounds of iteration and feedback, we eventually arrived at the final design which you can try out in the app yourself: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/07/Simulator-Screenshot---iPhone-17---2026-07-04-at-11.26.03.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/07/Simulator-Screenshot---iPhone-17-Pro---2026-07-05-at-18.38.42.png) Hopefully you’ll agree that this is not just clearer, but more visually appealing and inline with the rest of the app’s design. You might argue that this is subjective, but that is exactly the point... ## The root cause: code can be verified, but design can’t [The leap in AI coding ability](https://www.philmorton.co/with-claude-opus-4-5-i-can-build-software-again/) over the last couple of years came from a specific training method: letting a model attempt a task many times and using an automatic grader to score every attempt. **For code, that grader already exists. Does it compile? Do the tests pass? The answer is an unambiguous yes or no.** It’s the same reason AI has become so good at games like chess and Go: in each case there’s a definitive right answer to train against. Design has no equivalent correct answer. There’s no automatic test you can run for whether an interface feels trustworthy, expresses a brand or is beautiful. **Design quality is subjective, dependent on context and culture, and often only knowable after real people have used the thing.** You can’t compile it and see if it passes. When you can’t grade something automatically, the fallback is to have humans rate outputs instead. For design, that’s slow, expensive and unreliable. The model [*"is not learning what is objectively correct; it is learning what people tend to prefer"*](https://toloka.ai/blog/what-is-rlhf/?ref=philmorton.co) and this varies enormously by context. Almost everything else that makes AI weak at design follows from this single problem. ## Average is fine for code but not always what you want for design LLMs work by predicting the most probable next thing. **Left to their own devices, they gravitate towards the middle of everything they’ve seen.** The training used to make them helpful and safe narrows this further, because [human annotators tend to prefer outputs that look familiar](https://arxiv.org/abs/2510.01171?ref=philmorton.co). The result is a strong pull towards the average. For code, landing in the safe, conventional middle with a boring-but-correct solution is usually fine. **Sometimes average is acceptable for design as well.** If you’re creating a payment screen in a banking app using an established design system, something that is average might be a good first draft. Here an expected pattern might be the right solution because being following conventions makes it easier to use. **The trouble comes when we want to design anything that is unique, creative or distinctive.** For things like visual design, brand expression, or anything that’s meant to set a product apart, the average is not what we want. This is where AI’s pull towards the mean reduces its value to design the most. ## There’s less design to learn from, and it’s harder to learn As you probably know, LLMs are trained by feeding them enormous amounts of data. **Code is one type of data that is both abundant and easy to learn from.** GitHub, Stack Overflow and decades of open source mean there’s a huge amount of text-based, machine-readable training data. The largest open-source dataset of code ([The Stack v2](https://huggingface.co/datasets/bigcode/the-stack-v2?ref=philmorton.co)) contains 67.5TB of raw data covering over 600 programming languages across 3 billion files. **There is no such equivalent for data on design.** The largest public dataset of real interface designs ([Rico](https://papersgraph.com/datasets/ricosca?ref=philmorton.co)) only has 72,000 Android UI screens from 9,700 Android apps. Most high-quality design work is locked inside proprietary Figma and Sketch files that are inaccessible for training. **The scarcity of training data isn’t even the main problem though.** The bigger issue is that design doesn’t translate well when being turned into training data. A good interface works because of things that are hard to capture in a screenshot: hierarchy, spacing, rhythm, the reasoning behind why one element sits above another. **The model can see what an interface looks like but doesn’t know *why* it works.** This is why AI tools feel like they are [cargo culting](https://www.britannica.com/topic/cargo-cult?ref=philmorton.co) design: they’re making stuff that looks plausible but there is no thinking behind it. ## It’s designing blind There’s another reason that AI-generated design work is often poor: when one of these tools generates a design, **it usually can’t see what it’s making**. When you ask ChatGPT or Midjourney to generate an image, those models paint in a visual space and are always working with a version of the picture itself. They can ‘see’ what they produce. **But when we ask AI for an interface, it’s writing code**, placing elements by their coordinates without ever seeing the end result. It’s effectively drawing with its eyes closed. This is why AI is particularly bad at generating SVGs (vector graphics). You’re asking it to create precise geometry, exact coordinates and curve points, with no way to see the shape it’s describing, so the output is frequently distorted or wrong. Try asking AI to convert a PNG to an SVG and you’ll see what I mean: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/07/Screenshot-2026-07-13-at-15.53.41.png) This should not be hard. Newer tools and workflows are starting to close this gap. They **render the output, take a screenshot and feed it back** so the model can look at what it produced and revise. This helps, but there are still two problems with it: - The model’s vision is not very precise: it can’t see pixel-level differences. It’s like it’s short-sighted and isn’t wearing glasses. - The thing doing the looking is the same model with the same missing sense of visual taste. So the loop catches obvious errors while still being unable to judge whether the result is any good. ## Will this change or is there a fundamental limit? **Some of these limits will disappear** as the models improve. The “it’s designing blind” problem will reduce as computer vision gets better and AI can check its work more easily. **But most of the reasons are structural.** You cannot build an automatic grader for aesthetic taste, creativity, novelty, etc. because there is no correct answer to grade against. AI will improve but it’s always going to be limited in terms of what it can design for you. It will be fast and competent, but generic. Humans will always be needed to provide the differentiation, judgement and originality that make design actually good. ## Understand its limits so you can understand how to get the best out of it Of course, *AI is bad at design* is a generalisation in the same way as saying that AI is bad at anything. Just because it has limitations doesn’t mean we can’t use it to improve our design process. The practical takeaway is this: **don’t ask AI to do the design, get it to help you do *parts* of design**. The key to using AI in any task is [working out which steps should be done by humans and which by AI](https://www.philmorton.co/when-should-a-task-be-done-by-ai-and-when-should-it-be-done-by-a-human/). If you hand over the whole design process, you’re going to be disappointed for all of the reasons outlined above. Instead, **if you understand AI’s limits, you can identify which parts it can genuinely help you with.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Five reasons design teams are struggling to adopt AI URL: https://www.philmorton.co/five-reasons-design-teams-are-struggling-to-adopt-ai/ Last updated: 2026-07-08T12:08:42.000Z [AI is turning the design process upside down](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/), yet **a lot of design teams have barely gotten started** on their journey to rethink how they work. Although [design tooling is still immature](https://www.philmorton.co/the-design-to-code-ai-workflow-youre-looking-for-doesnt-exist-yet/), it’s clear that the way that designers will be working in 5 years’ time is completely different to how they work now. The technology is here, but many teams are moving slowly to make the wholesale changes that are required. The gap between what is possible and what most teams are actually doing is huge. Why? ## 1\. No-one on the team is technical enough Before AI, you didn’t hire designers for technical skills – it was a nice to have. The problem is that **some teams don’t have *anyone* who is technical**. No-one who is tinkering with Claude Code in their evenings. No-one who understands the terminal, git or how the thing they’re designing actually gets built. No-one to lead their thinking around changes to tools, ways-of-working and methods. In the [UX Tools State of Prototyping survey](https://survey.uxtools.co/spring-2026?ref=philmorton.co), the designers doing most of their building with AI were overwhelmingly the ones who already had an engineering-adjacent background. **The people best placed to build with AI are the ones who already have a technical background.** You don’t need designers to code like engineers, but you do need enough technical grounding to make sensible calls about tools, workflow and how design joins up with development. If nobody has that grounding, where does the new process come from? ## 2\. You have to bring developers with you [New ways of working](https://www.philmorton.co/what-the-vibe-engineering-workflow-tells-us-about-the-future-of-ux-roles/) mean that designers and engineers have to work closely together. The ideal is that you’re in the same code repository as the engineers, not throwing a Figma file over the wall. For most teams that’s a long way from today. Designers and developers might sit in the same squad, but they’re still siloed, with a big handover in the middle. So when you’re trying to work out what your AI design process is going to look like, you’re not just choosing new tools for yourself. **Your engineering partners have to adopt the same approach**, because it makes no sense for you to work one way and them another. If you’re starting from a position where design and engineering are already not aligned, and are used to working in their silos, then there’s a lot more to do than just choosing a tool. ## 3\. AI’s usefulness to design is uneven In the old way of working, the output was roughly the same whatever the project: a high-fidelity Figma file. Now it varies wildly. If you’re assembling a feature on an established product with a mature design system, you barely need Figma at all. **AI is good at using existing components** and you’re not asking it to do any visual design, which it’s bad at. You can sketch the rough idea, have it assemble it and iterate. There’s little point building a pixel-perfect version by hand first. But if you’re working on a new product which needs its own visual design or brand, it’s a different job. AI can get to a rough wireframe using generic components, then it stalls. **Creative and original visual design still needs a human.** So **the impact of AI on the design process is uneven.** It transforms one type of project and barely touches another. That makes scoping harder: you can’t assume every project takes the same shape, needs the same skills or moves at the same pace any more. ## 4\. AI isn’t good at design! I could write a whole newsletter on this, but the short version is... **AI is nowhere near as good at design as it is at coding**. With code there’s a verifiable right answer - does the test pass? With how an onboarding flow should work, there isn’t a ‘correct’ answer. Design requires human creativity and judgment. If you ask AI to design something, you tend to get a result that looks right but has no thinking behind it. It copies the surface and skips the reasoning. Then you have the problem that it’s terrible at novel visual design. At least with the latest models, **AI creates functional but generic design work**. You get the average of what it’s seen. Having it use an established design system to assemble pages for you gets better results because you’re not asking it to be creative or exercise any aesthetic judgement. As with any task, you get better results from AI when you give it a more limited task with a narrower scope. If you break the work into smaller steps and you do the decision-making, then it’s ok. But this also means that adopting AI into your design process isn’t a plug-and-play task. ## 5\. Nobody has time to work it out Figuring out all of this takes time, and most teams don’t have any spare. Four years ago, picking a design tool took about five minutes: it was Figma, like everyone else. Now there are countless options and they change every month. **There’s no settled answer to copy.** Working out which tools to use, how to align with engineering and what your new process looks like is a lot of work. It requires an investment of time and resources to do, which many stretched design teams can’t afford. ## This can‘t be done off the side of the desk Most teams need help to get unstuck: new hires with technical skills, outside expertise or leadership carving out protected time. **You can’t figure this out properly while delivering everything else at the same time.** The teams adapting fastest aren’t more aware of the need to change. They just have a head start: technical people already on the team and close existing relationships with engineers. If you’ve got neither, it’ll take longer. The best way to figure this stuff out is to **build something together as a proof-of-concept**. Only through [learning by doing](https://www.philmorton.co/its-never-been-easier-to-learn-by-doing/) can you figure out what’s going to work for your team. But doing so needs dedicated time. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### When should a task be done by AI, and when should it be done by a human? URL: https://www.philmorton.co/when-should-a-task-be-done-by-ai-and-when-should-it-be-done-by-a-human/ Last updated: 2026-07-01T10:43:59.000Z I think a lot of people’s AI learning follows a similar path: 1. **Start with basic queries** that chatbots handle well: prompts that are an alternative to Google, generating images, summarising documents, etc. 2. [**Learn better prompting skills**](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/) and gain experience through trial-and-error. Treat it like a teammate rather than a tool, give the LLM proper context, edit the prompt rather than giving feedback, and so on. 3. **Get overconfident and ask AI to do too much**, then get disappointed with the results. You ask it to create a whole presentation or design an entire onboarding flow, and the output is... terrible. This is when you realise that **to get the best results from AI, you have to break down tasks and figure out which parts should be done by a human and which by the LLM**. Almost every design and research team is trying to figure this out at the moment. They’re going through their processes step by step and asking the same question: what is the best blend of human and I, for any given task? ## What to consider when deciding who (or what) does each task The trap a lot of people fall into is asking the AI to do too much, in one go. When a task feels too risky to hand over or AI just isn’t hitting the quality bar, break it down further. **A high-risk task often contains smaller pieces that are lower risk.** Split it up and you’ll usually find parts you can comfortably pass to AI, even if the whole thing feels too much for it to do. This then allows you to **evaluate a task against a number of questions**: - How high does the quality need to be? - How much input will you have to give, to get that quality? - How long will it take? - What will it cost? - How risky is it: how much does it matter if it’s wrong? Transcribing a meeting is low risk and needs almost nothing from you, so you can let AI handle it. Writing the executive summary for six months of work is high risk and has a high quality bar, so you’ll need to give a lot more input. ## Picking how to work together Given the profile of the task, you then need to figure out **how you’ll work with AI**. Say if you are writing something, there are many ways to do it: - **Fully delegate:** You give it a brief, it writes the whole thing for you. - **You outline, it builds on it:** You write a skeleton and it fills in the gaps. - **Draft for you:** AI writes a first draft and then you edit and build on it. - **You dictate, it writes:** You give it the full content, but it’s doing the writing for you like a 1950s secretary. - **You write, it critiques:** You write by hand, it reviews your work and gives you feedback. There are so many ways to approach a task with AI, so which do you choose? You can easily end up picking the wrong approach and have to backtrack and try another. This is one of the best examples of where **learning how to get the best from AI simply comes from learning by doing**. But as a crude rule of thumb, here’s how I would try and summarise what I’ve learned: - **Is this a task that the AI can verify it has got the correct answer for?** (Coding, data analysis, etc.) → Safer to delegate. - **Is this something where there is no correct answer, and quality matters a lot?** (Writing, design, etc.) → You lead at the start and end of the process, but it can help you in the middle. ## A worked example Imagine you’re writing a conference talk. This is high risk. People are coming to hear your expertise, and if it’s weak it reflects on you, so the quality bar is as high as it gets. You could ask AI to create the whole thing: research, narrative, slides, speaker notes. It’ll produce something, but **it won’t be as good as what you’d make**, because it’s doing every sub-task whether it’s good at it or not. You’ll end up on stage presenting ideas that aren’t quite yours and people will know something is off. Instead, what you could do is: - You dictate your ideas for the talk and it pulls out the key arguments. - You use it as a sounding board to expand and test your ideas. - You craft the narrative yourself, because that’s the high-value part, and let AI write an outline. - Once the slides are drafted it can review the deck against your audience and flag what’s missing. - When you rehearse, it listens to your run-through and tells you which points you skipped or misrepresented. The key is to **assign tasks to who is best to do them**, whether that’s you or AI. If you just delegate everything, it forces AI to do parts it’s not good at. ## When you delegate important decisions, you get slop I love this quote from [Anu Atluru](https://www.workingtheorys.com/p/slop?ref=philmorton.co): > *Slop is the absence of decisions and, more critically, discernment.* When you produce any work, you make hundreds of tiny decisions. **For any part of work that you handover to AI, you are asking it to make those decisions for you.** When we hand over too much decision-making to AI for work that demands a high quality bar, people notice and it comes across as ‘slop’: insincere, mediocre and average. Learning *when* to use AI is therefore an important skill to develop. **People talk about a ‘human-in-the-loop’ and *who* that human is matters a lot.** It’s tempting to think this is all about domain expertise, but it isn’t enough on its own. You can be brilliant at your craft and still get mediocre output from AI, because you haven’t developed the judgement of when to use it and when not to. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How I built Pegs Out, my first iOS app URL: https://www.philmorton.co/how-i-built-pegs-out-my-first-ios-app/ Last updated: 2026-06-24T12:18:08.000Z **I’ve wanted to build an iOS for many years**, but the time I would have needed to learn Swift, Xcode and everything else was just too much. Even for someone with a technical background, the barriers to making software were too high to overcome. Of course AI has changed that in the last year. **So I am very happy to launch my first iOS app:** [**Pegs Out**](https://apps.apple.com/gb/app/pegs-out/id6778139891?ref=philmorton.co)**.** [Check it out and let me know what you think](https://apps.apple.com/gb/app/pegs-out/id6778139891?ref=philmorton.co), it’s free! ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/69---1.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/69---2.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/69---3.png) ## Get your Pegs Out Pegs Out solves the problem of ***should I put my laundry out to dry today?*** Imagine it’s 3pm on a cloudy 12ÂșC spring day. If you hung your clothes on the line, would they dry in time or not? It turns out that what causes clothes to dry outside isn’t what you might imagine. Hotter days help, but why (because warmer air can hold more moisture than colder air)? Some factors (wind) are much more important than you might realise. Others (the dew gap) most people wouldn’t have ever heard of. Pegs Out takes all of these factors and **combines them into a single ‘drying score’ from 0 to 100**. It tells you if your laundry will dry and approximately when. You can see when the last chance to hang clothes outside is and when drying stops. ## Before we code, we learn and plan I’ve been looking for an idea to build an app around, something that is simple and not something that necessarily requires me to build a business around. Once I had the idea, I had Claude **run some initial research to explore the problem space**. What is the physics behind drying? How do you build an iOS app? What APIs could I use (for free)? After some back and forth, this came together in a brief for the app: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image.png) The original brief for Pegs Out Then it was on to picking a name, because you’ve got to call the code repository something. And I find that **having a name is the difference between an idea and a real thing**. Naming is great example of a task that AI can help you with by generating a lot of ideas, but it’s never going to pick the best one. That’s down to your own taste and preferences. Once we had a good candidate, Claude went off to do its due diligence on ‘Pegs Out’. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image-1.png) Tip: for a task like this, first get it to explain how to do it, not jump straight to ideas. Once I had created the repository on GitHub, it was time to do some proper research and planning. I had Claude go off and document two things: - **The physics:** How do clothes dry outside and how can we model it? Are there well established formulas and academic research around this? It turns out that there are some well used formulas, but the research is very thin. - **How to use Claude to build iOS apps:** I wanted to know all the best practices so I could avoid the common pitfalls. Other people have done this before - what have they learned? Claude wrote all of its findings up as markdown files in a `/docs/research` folder. This was more for it than me. The point of this is to **build context that it can use later**. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image-2.png) Claude’s research prior to writing any code. Doing this upfront research isn’t strictly necessary. You can just type a prompt into Lovable and be done, but **to get the best results, getting the right context in place makes a huge difference**. You’re going to be making a lot of decisions and best to make those as transparent as possible. ## Is Claude Design any good? Before jumping into code, I wanted to visualise what I wanted to build. What is the shape of the app? What information will I have on each screen? I thought this would be a perfect opportunity to use Claude Design. I had a full brief document which I gave it and it came up with something decent, first in the wireframe mode and then “high fidelity” (which, let’s be honest, is really a coloured-in wireframe). ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image-3.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image-4.png) Claude Design - wireframes and then ‘high fidelity' Claude Design is a good thinking tool, a rapid prototype tool. I’m not sure it’s a capital-D *Design* tool because although it stumbled on some good conventions that I kept, it was terrible at solving my data visualisation and information design problems. ## Building out the basic functionality Once I had an idea of how the app would work, I got Claude to **break the work down into separate issues** and store these in GitHub. This is my preferred way to manage the project and its long-term context (which you have to do). Screens become issues/tickets and then you get Claude Code to work through one at a time. I had also set up various skills which I adapted from a previous project. This codifies my ways of working which I **loop through for each issue** (i.e. feature): - Claude writes an issue based on my input and stores it in GitHub. - `/research` looks at the issue, identifies gaps and does research to fill them, then posts back its findings and decisions I need to make. - `/gh` reads the issue and creates a plan to implement it. - `/checkplan` interrogates the plan from different angles. - Then Claude starts coding. - `/codereview` is run immediately after, looking for issues. - `/finish` commits and pushes the code, comments on the issue and closes it. - When I’m ready to merge the code into the `main` branch then I use `/release` This may feel like a lot, especially if you are coming from a non-technical background, but these skills are essentially just simple text files that contain steps for the agent (Claude Code) to follow. Since coding follows such a repetitive process, it’s **well worth doing as the time saved overall will be massive**. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/Screenshot-2026-06-17-at-14.50.07-1.png) The Xcode Simulator allows you (and Claude) to see what the app looks like. So using this loop, Claude built out the first version of Pegs Out in iOS. I also had to learn how to use Xcode (Apple’s coding app) as there are several steps you as a human need to do. Claude told me what to do whenever I got stuck. We ended up with a decent MVP connected to real data, running on my phone, that I could test with. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/IMG_5149.PNG) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/IMG_5153.PNG) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/IMG_5154.PNG) The first version of Pegs Out. ## Getting from basic components to something that feels ‘designed’ To create the MVP, I asked Claude to stick strictly to out-of-the-box Swift UI components and not to create anything too custom. I wanted to deliberately separate the functionality and information design from the visual design and branding. The trouble is that once you have the basic functionality working, **how do you go from bare bones generic app to something that looks more polished?** As someone with limited visual design skills, this was the hardest part of the project. This is where I leaned heavily on established conventions in the weather app category. I downloaded a bunch of other apps and then borrowed UI concepts which I felt would work well with Pegs Out’s niche. Hello Weather and Carrot Weather were the main inspirations, and helped me realise how to redesign the Today screen to better convey the most important information. **This was the first point at which I used Figma.** Moving away from stock-looking Swift UI, I wanted to be able to play around with how it looked without having to go through a loop with Claude. I didn’t mock up the whole app or all of the states in Figma, just enough to give Claude clearer direction. I don’t think anyone who says “Figma is dead” understands that sometimes you just need a place to freely explore ideas in this way. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/image-5.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/06/IMG_5256.PNG) My Figma mockup on the left, and the final iOS app on the right With a Figma reference, Claude broke down my redesign into issues and implemented in a couple of days. Along the way it exposed a number of decisions we needed to make, but nothing major. As you can see, the end result is very close to the mockup. ## Down the hill to release Once I had gotten [over the hill](https://basecamp.com/hill-charts?ref=philmorton.co) and got a design that I felt represented the final thing, then it was a race ‘down the hill’ to the finish line. There are **all sorts of things you need to do** to get an app live: - Write a privacy policy and support page, and host them somewhere online - Checking and fixing accessibility issues - Create an app icon - Make your App Store screenshots - Write the App Store description - Fill in the age rating, policy compliance, review info, etc. in App Store Connect But once you get through all this (with Claude’s help) then it’s time to submit it for App Store Review. 5 days later, it was time to press the button to make it live. ## You can do it too I keep telling people that [the best way to learn about AI is to build something](https://www.philmorton.co/its-never-been-easier-to-learn-by-doing/). Developers have always had side projects, but UX people rarely do, because historically we were unable to actually make software. **Now with AI, you can.** And you learn *so much* by doing so. Everyone has an idea for an app. Go and make it. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The anatomy of an AI agent URL: https://www.philmorton.co/the-anatomy-of-an-ai-agent/ Last updated: 2026-05-20T09:50:55.000Z AI agents are clearly *an important thing you should know about* but what exactly is an agent? Almost every company or tech product seems to have one, but **not everything that’s called an ‘agent’ really qualifies as one**. **People are also starting to talk about and treat agents as if they’re real people.** With [OpenClaw](https://openclaw.ai/?ref=philmorton.co), you can have an AI agent that has a name, a personality, a ‘heartbeat’ and a ‘soul’. You message them requests but they can also act on their own. They can pop up in Slack to answer questions. You need to [onboard them](https://every.to/p/what-i-learned-onboarding-our-ai-project-manager?ref=philmorton.co) like a real colleague. People give them their own email accounts and so on. So if we’re going to [anthropomorphise](https://en.wikipedia.org/wiki/Anthropomorphism?ref=philmorton.co) agents, then maybe a good way to explain how they work is to think about their anatomy. This is my attempt to explain what they are and how they work in plain English. ## What is an agent? A model with tools in a loop Imagine a simple chatbot powered by an LLM: you ask it a question and it gives you an answer. You could give this chatbot tools (e.g. the ability to search the web or post updates to Slack) and it could be useful, but it still wouldn’t be an agent. The key difference is that **an agent runs in a loop**. Give it a goal and it goes round and round, performing tasks, evaluating the results and deciding what to do next, until it gets there. ## The core parts: brain, body, home and hands If you wanted to make your own agent, you’d need the following four things. If you took any of them away, what you’ve got isn’t really an agent anymore. - **Brain (the LLM):** does the thinking. You send a message and it sends one back. If you’re making your own agent, then you would usually be using an API from Anthropic, OpenAI, etc. and paying for usage. - **Body (the harness):** the program that runs the loop in code. Claude Code is a harness. Codex is a harness. A Python script that calls an LLM API in a loop is a harness, and the simplest version is only about 50 lines of code. Without one, the brain is just a chatbot answering one question at a time. - **Home (the runtime):** where the harness process actually runs. Your laptop, a server in the cloud, a Mac Mini, etc. - **Hands (tools):** the functions the model can choose to call. A tool can be as simple as a calculator or a web search. Or it might be something more involved, like a Gmail connector. ## The supporting parts: personality, memory, keys and voice You can technically call something an agent with just the four core parts, but it would be of limited use. These next four parts are what take an agent from working to genuinely useful: - **Personality (the system prompt):** a chunk of text added to the start of every conversation telling the model who it is, what it’s doing and how to behave. Without this, the agent has no identity or direction. It’s just a generic helper pointed at random tools. - **Memory:** which has two parts. Short-term memory is the conversation history that the harness sends to the model on every turn. Long-term memory can be a file or database that tracks the history of what the agent has seen or done. Without long-term memory, every session starts from scratch. - **Keys (credentials):** the API tokens, OAuth tokens and passwords that unlock the tools that reach protected systems like Gmail, GitHub or Stripe. The agent itself never sees the raw keys – the harness loads them and uses them when a tool needs to authenticate. Without keys, the agent is limited to tools that don’t need to log in to anything (calculators, web search and local file operations). - **Voice (interface):** how you talk to the agent and how it talks back. Could be a terminal (like Claude Code), a chat UI (like ChatGPT), an email address it monitors, or a webhook that fires when something happens. The interface is whatever sends input into the loop and surfaces output back out. ## Isn't an agent meant to have autonomy? One thing I struggled with is whether or not an agent needs to be autonomous to count as one. People call Claude Code an agent, but (in most cases) you have to tell it what to do for it to act. It turns out that agents can have two types of autonomy: - **Autonomy of action:** when the model picks the steps within a task. You tell Claude Code to “fix the bug” and it figures out the rest: read the file, run the test, find the bug, edit the code. You didn’t dictate the steps, it did. This is what makes something an agent. - **Autonomy of initiation:** when the agent decides *when* to act at all. It wakes up at 7am, checks your inbox and sends a summary. Or it notices a calendar conflict and resolves it without being asked. When people talk about agents as colleagues, they’re picturing something that has both. But really, autonomy of action is what makes something an agent. Initiation is what makes it feel like a colleague. ## Understanding them for what they are Agents are getting more common and capable. They’re showing up in more products and workplaces – in your Slack channels, emails and so on. Increasingly they’re being presented as colleagues rather than tools. That means the question of what actually counts as an agent matters more than it used to. If it’s model with tools in a loop, it’s an agent. If it’s essentially just a prompt you’ve saved for repeat usage, it’s not really an agent. **Products will keep coming and going, but these core ‘body’ parts won’t.** Hopefully this gives you an easier way to understand them. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Do researchers still need to take notes in interviews? URL: https://www.philmorton.co/do-researchers-still-need-to-take-notes-in-interviews/ Last updated: 2026-04-22T09:50:16.000Z These days, **every video-calling tool transcribes your interviews**. Some do it badly (Teams), some do it well (pretty much all the others), but transcription is basically a given. So why would anyone still take notes in user research? There isn’t really an established convention here. Some researchers use a dedicated notetaker so they can focus on moderating. Others take their own notes, but how they do it comes down to personal preference. Most people default to whatever they did last time without thinking about it much. Let’s break this topic down into **three questions**: 1. Given AI transcription, does anyone need to take notes at all? 2. If yes, should it be the moderator or someone else? 3. If the moderator, how should they do it? ## Does anyone need to take notes at all? Automatic transcription is a massive timesaver, but the reality is that **transcripts only capture what was said**. Depending on the type of research you’re running, that will represent different degrees of what was communicated. In formative research where there’s no prototype and nothing shown on a screen, a good transcript covers most of what matters. It doesn’t capture how something was said, body language and so on, but you might get 80% of the meaning with a literal recording of the words that were spoken. In evaluative research, transcripts only represent a fraction of the data – you might as well not bother analysing them in some cases. What someone did with the prototype, where they hesitated, which buttons they hovered over, the small sigh before they gave up: none of this is in the transcript. If your only record is what was said, your analysis will be flawed. Across all types of research, **someone has to capture the data that AI can’t (yet) record on its own**. Another good reason for both moderators and observers to take notes – even if it’s being perfectly recorded by AI – is that **writing down the key points keeps you more engaged and helps you remember what happened**. This is true for any communication, not just user research. ## Should the moderator take notes? Some researchers delegate note-taking, getting someone else to take notes so they can focus on the conversation. There’s a case for this, especially if you’re new to moderating and find the two things hard to combine. But I still think **moderators should take notes**. You have the best vantage point, [especially in-person](https://www.philmorton.co/give-in-person-research-a-chance/). You can see the screen from where they see it. You notice the small body language, the hover before the click. You’re closest to the participant and most present in the conversation. A notetaker on a gallery view or behind a one-way mirror is getting a flatter version of the session, and their notes will reflect that. This is even more important if you are going to use AI to help you analyse the findings. Anything that is not in the transcript – body language, hesitations, pass/fail, the click that didn’t happen – needs to be captured as structured data and **the moderator is in the best position to do that**. The ‘it splits your attention’ objection is valid, but it’s an argument about training, not whether the thing is possible. With practice you learn to write without looking, develop shorthand and pace the interview to leave space for writing things down. It’s a craft skill: hard at first, then automatic. ## How should they do it? There are a whole range of ways that moderators can take notes: - **Free-form on a blank page or in a file.** The simplest option. Good for very unstructured exploratory work when you don’t know what’s coming. - **In a spreadsheet.** Columns for participant, rows for topics. Structured and easy to process later. - **Annotating the discussion guide.** Pen on printed paper (old school!) or as a PDF on an iPad. Notes and questions live in the same place. - **In a Notion database.** Templates can make this easier than a spreadsheet. Useful when you want structured data points like scores or pass/fail. - **On a Miro/FigJam board.** Sticky notes per topic or per participant. Good for analysis, but trickier to do live. Each has is pros and cons. Most researchers only ever reach for one or two, and **tend to default to the same method on every project** regardless of what it actually needs. ## Structure your notes to structure your data Whichever method you pick, **the more structure you build in, the easier the analysis will be**. If you have to sit down with 12 sets of completely free-form notes and try to compare what people said about one topic, it’s going to take a long time. The points are in there, but you’re hunting across pages of unrelated stuff. When you structure your notes, they’re much faster to work with. **A rule of thumb: is there something about each session you’d want to analyse later that won’t be in the transcript?** Body language, hesitations, a pass or fail, a score out of ten, the fact that someone clicked the wrong thing three times before noticing. If the answer is yes, you need a way to capture it. What this looks like depends on the method. If you’re annotating a discussion guide, don’t just leave blank space to write in: add boxes for the specific things you want to capture at each step. If you’re using a spreadsheet, add more rows for specific data points. If you’re using a Notion database, define the properties you’ll fill in as you go, so the fields do the remembering for you. Then **when you come to do analysis, it’ll be much easier for both you and your AI assistant**. ## The data AI can’t see With the tools we now have, it’s tempting to dump every transcript into AI and ask it to do all the analysis for you. With clean human-edited transcripts and a million-token context window, an LLM can produce a decent set of themes from interviews faster than anyone could do by hand. But **you only get good results if the data going into the LLM is complete**. Transcripts plus structured notes can give you a close-to-complete dataset. Transcripts on their own, for anything more behavioural than a formative chat, is a partial one. **The further your research sits from pure speech, the more weight your notes carry.** In usability testing, the notes are the data. In formative interviews, the transcript does more of the heavy lifting, but the notes still hold the things you only understood because you were running the interview. So before your next study, **don’t just default to your usual way of taking notes**. Think about what data you’ll need at the other end, and design your notes around it. The analysis (human or AI) will be much better for it. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Evals are for everyone URL: https://www.philmorton.co/evals-are-for-everyone/ Last updated: 2026-04-01T09:50:23.000Z There’s a big difference between writing a good prompt for ChatGPT/Claude and writing a prompt that’s used as part of an AI feature or product. When you’re using an LLM, **you alone provide the input and you can see every output**. This makes it easy to spot where it’s going wrong so you can refine your prompt to get a better output. But when you’re designing an AI feature in a product, **you can’t control the input and you can’t see the output**. You therefore need to create a prompt that is robust and predictable for a wide variety of inputs. I’m building a product to help parents deal with the overwhelming number of emails they get from schools. To do that, I need to get AI to identify events and to-do items in emails and attachments. This is a lot harder than it seems: - *“Please label all uniform and ensure long hair is tied up”* \- Sounds like a task but not really something you would put on a to-do list. - *“After February half term, pupils will be doing outdoor games”* \- Is this an event? Is there a task? - *“KS1 children should wear a plain white top and dark trousers”* \- You have to parse KS1 to include years 1 and 2. Getting an AI to handle these reliably is genuinely difficult, and in my app, if this doesn’t work, the whole app doesn’t work. That’s what **prompt engineering** is all about: the art and science of making a prompt reliable enough to use in the wild. ## What evals are and why you need them When you’re creating a prompt that’s used as part of a product, **you need a way to test its performance** given different inputs. That’s what evaluations a.k.a. ‘evals’ are for. The idea is that instead of running your prompt manually and checking the results, you **set up a repeatable test**. You give it a dataset of real inputs, run the prompt against all of them, and score the outputs systematically. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/03/Screenshot-2025-12-21-at-09.54.34.png) OpenAI’s evaluation tool. On the left is your prompt, while in the table, the rows are different inputs. The red and green columns are the graders (these are prompts that score the responses). There are several ways of doing this, but for my purposes I’ve been using the evaluation tool in OpenAI’s platform. You upload test data – in my case, a set of real school emails – and the tool runs your prompt against each one. You then set up **‘graders’ – separate AI prompts that score specific aspects of the output**. One grader might check whether the response is valid JSON. Another checks whether tasks were correctly identified. Another checks whether events were extracted with the right dates. This way you can test and iterate your prompts in a structured and evidence-based way. ## Getting sucked down the never-ending prompt engineering rabbit hole The only problem is that prompt engineering is **incredibly psychologically draining** in a way that most product work isn’t. When you build the UI, you change something and immediately see the result. There is a clear cause and effect. Progress is visible and cumulative. Prompt engineering is the opposite in almost every way. Because LLMs are non-deterministic (they give you a different response every time and you don’t know why), **it’s like rolling a dice every time you make a change**. You improve one thing and break another. You switch to a model that should be better and it’s somehow worse. You want your prompt to be ’done’ but **you have to settle for gradually drifting towards good enough**. It will never produce the results you want 100% of the time. This can be really frustrating and suck hours and hours of your time – it’s always tempting to try and make ‘just one more change’. ## How to make this loop less painful In retrospect, my biggest mistake early on was trying to do too much at once. I wrote and tested the entire prompt (which does multiple things and follows multiple rules) at once and then **when it didn’t work, I couldn’t figure out why**. The approach that actually works is the same discipline that makes good software development: **change one small thing, validate it, then move on to the next one**. If you’re only changing one variable at a time, it’s much easier to make progress. If I was starting with a new prompt, here’s what I’d do: - **Start with the simplest possible version.** Don’t try to write a prompt that does everything from day one. Figure out the smallest, most contained version of the problem and get that working first. I tried to extract tasks and events in one go. I should have started with just tasks, got that working, then added events. - **Break your graders down.** Don’t try to score everything in one grader – you can’t tell what’s passing and what’s failing. One grader for task extraction, one for event extraction, one for JSON validity. Don’t get stuck wondering whether it’s the prompt or the grader that’s wrong. - **Generate synthetic test cases.** Real emails are great for developing a prompt, but when you’re testing a specific change, ask Claude to generate test data that target that change and all the edge cases. ## Evals are for everyone, not just engineers You might think that this is something for engineers to figure out, because they’re the ones building the backend processes that use prompts. But **writing a prompt and the graders that test it aren’t purely technical problems**. In my app, the main question to solve for is ‘what counts as a task in a school email?’ That’s the sort of question that engineering can’t answer on their own. **The AI feature you’re building needs to behave in the way that the people using it expect.** So to create the right prompt and test it, you need to understand people’s mental models and expectations. And the people who are best placed to do that are typically researchers and designers. As more products get built around AI, the question of how that AI behaves won’t just be a technical implementation detail. It’ll be a product decision, a research question and a design challenge. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The soft skill that matters most when AI makes teams smaller URL: https://www.philmorton.co/the-soft-skill-that-matters-most-when-ai-makes-teams-smaller/ Last updated: 2026-03-25T10:50:12.000Z With AI [upending long-established ways of working](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/), a lot of people in our industry are wondering what the future holds for their career. [Are we cooked?](https://are-designers-cooked.vercel.app/?ref=philmorton.co) How do you get up to speed or ahead of the curve? Go on LinkedIn and it’s **tools and workflows that everyone’s talking about**. *Have you seen what Google Stitch can do? Which Figma MCP are you using? How are you using Claude Code to analyse research transcripts?* This is what I wrote all about [last week](https://www.philmorton.co/the-design-to-code-ai-workflow-youre-looking-for-doesnt-exist-yet/). You do need to start experimenting with AI and integrate it into your work, **but hard skills alone aren’t going to get you hired in 3-5 years’ time.** The soft skills that are valuable are also changing and the most important one is ownership. ## In smaller teams, soft skills matter more If there is one clear trend, it’s that **AI is making teams smaller**. Agents allow people to do more of the same work (e.g. writing more code) and more types of work (e.g. designing a screen and coding the production front-end). This means you can achieve the same output with fewer people. Imagine you’re creating an app. Two years ago, you might have a squad of ten people: five engineers, two designers, a PM, a researcher and a data scientist. Now you can achieve the same outcome with 3-4 people who are skilled at using AI. One implication is that if you are assembling a team of 3-4 people, **the soft skills of each person matter more**. In a team of ten, you can tolerate someone who is not contributing in the right way, but with three it’s a disaster. Soft skills used to be something that people only deliberately started developing when they wanted to move into a leadership role. **With smaller teams where each person has much more responsibility, this has to change.** ## Ownership is the defining skill There are lots of soft skills that are important at work, but the one that’s really going to matter in these small AI-powered product teams is ownership. If you have a small team in which each person has a broad remit, **you need people to take responsibility for broader outcomes, not tasks**. Your job is not to *test each sprint’s prototype with users* or to *design the onboarding flow*, it’ll be to *research, design and built the entire front-end of the app*. By ‘ownership’, I mean **the ability to achieve outcomes end-to-end without being told exactly what to do**: - Being comfortable with ambiguity, uncertainty and incomplete information. - Negotiating with others to get what you need and work through problems. - Maintaining momentum when things are unclear. - Being proactive in changing things for the better. - Creating clarity when there is none. Ownership means you **act like less like a contributor with a narrow scope of work and more like a general manager** who feels accountable for the whole product they’re bringing to market. **The best designers and researchers already operate this way** – they go beyond their traditional roles, take ownership of the problem and work autonomously towards the overall objective. With AI and smaller teams, though, the gap between those who have this skill and those who don’t is about to get much wider. ## How to build the ownership muscle This is already a skill that’s being emphasised across the industry. For example, Intercom’s designers are moving towards [owning the entire front-end design and build](https://ideas.fin.ai/p/intercoms-3-point-framework-for-ai?ref=philmorton.co). Before this skill becomes a baseline expectation, there’s time to develop it in a couple of ways: - **In your work:** find something in your team that nobody owns and drive it end-to-end. Has anyone explored AI moderation tools for research? Has anyone thought about how to embed accessibility into the AI coding process? Every team has a million initiatives that no-one is owning. Pick one up without waiting to be asked and own it. - **Side projects:** [build your own app or website](https://www.philmorton.co/its-never-been-easier-to-learn-by-doing/). When you have your own product, you handle everything: research, design, engineering, legal and compliance, pricing and so on. This isn’t just a learning exercise – it’s one of the best ways to demonstrate ownership to a future employer, because you’re literally the owner. Imagine a designer or researcher who, alongside their day job, has shipped a small product that real people use. That person has already proven they can operate across domains, make decisions and be accountable for outcomes. That’s the profile companies are going to be looking for. The concept of general management may feel alien to many of us – it seems like the opposite end of the spectrum from deep craft. But ownership doesn’t mean abandoning depth. It means **expanding your scope from producing outputs to being accountable for what those outputs achieve.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The design-to-code AI workflow you’re looking for doesn’t exist (yet) URL: https://www.philmorton.co/the-design-to-code-ai-workflow-youre-looking-for-doesnt-exist-yet/ Last updated: 2026-03-18T10:50:08.000Z Every design team is doing the same thing at the moment: trying to figure out what their process and tooling should be, [now that AI can write production-ready code](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/). The reality is that there isn’t a satisfying answer. **No single tool provides a complete loop between a production code-based design system and a visual design canvas.** Designers are having to use an incomplete approach because the tools simply aren’t mature enough yet. ## Two approaches, neither complete Because there’s no established end-to-end process, you see designers gravitating to one end of the spectrum or the other. **One group starts in Figma.** They’re experimenting with Make and similar tools, creating prototypes that are interactive but not production code. Some of these prototypes use the design system, but many don’t. These tools are accessible but limited: you quickly reach their limits and there’s still a handover to engineering. **The other group starts in code**, using Claude Code or Codex to build production UI directly. At [Intercom](https://ideas.fin.ai/p/intercoms-3-point-framework-for-ai?ref=philmorton.co), all their designers now ship PRs to production using tools like Cursor. They started with CSS and copy fixes and are progressing towards owning the entire frontend, with engineers handling the backend. Working directly with production code is the more progressive approach, but for many non-technical people, there’s a huge learning curve. These aren’t mutually exclusive. Most people still use a canvas for exploration and code tools for production. The problem is that **nothing connects the two ends properly.** That’s the gap everyone is trying to close. ## What ’solved’ actually looks like Before getting into what tools can and can’t do today, it’s worth defining what the ideal workflow would need to cover. Here are ten requirements I came up with: 1. **Import a code-based design system.** Can I connect an existing React (or other framework) component library so the tool understands my real production components, not just visual representations of them? 2. **Render real components visually.** Does it render my actual coded components on a canvas, so I can see exactly how they look and behave - not just a static image or approximation? 3. **Open canvas exploration.** Can I lay out multiple screens, flows or compositions side by side on a freeform canvas? 4. **Assemble layouts from components.** Can I drag, drop and arrange my coded components into new screens visually? 5. **Export production-ready code.** Can I export a new layout as clean, production-ready code that uses my actual design system components (not markup that approximates them)? 6. **Edit design system components visually.** Can I modify the styling, spacing, variants or behaviour of a component using a visual interface? 7. **Push changes back to code.** If I edit a component visually, can those changes be written back to my codebase as real code changes, not just a Figma update that needs a manual dev handover? 8. **Two-way sync.** Is there a genuine bidirectional sync where code changes update the visual canvas and visual changes update the code? 9. **AI-assisted design or prototyping.** Does the tool have AI features for generating layouts, components or prototypes? 10. **Integration with AI coding tools.** Can it work alongside or feed into AI coding assistants like Claude Code, Cursor or similar? No single tool or workflow ticks all ten at the moment. ## What the tools can actually do today Tools like [UXPin Merge](https://www.uxpin.com/merge?ref=philmorton.co), [Plasmic](https://www.plasmic.app/?ref=philmorton.co) and [Builder.io](https://www.builder.io/?ref=philmorton.co) can import your React component library, render real coded components on a visual canvas and let you assemble layouts visually. **The part that works is pulling code components *in*.** The part that doesn’t is pushing visual changes *back*. UXPin has no automated push-back to source repositories – sync is strictly one-way, code to design. Plasmic can generate and overwrite code for components authored in Plasmic, but never modifies the source files of your imported code components. [Builder.io’s Fusion](https://www.builder.io/fusion?ref=philmorton.co) is the boldest attempt, but its **sync is AI-mediated** \- it interprets your changes rather than establishing a guaranteed mapping. [Pencil.dev](https://www.pencil.dev/?ref=philmorton.co) takes a different approach by **putting a design canvas inside your IDE**, but it draws vector representations of components rather than rendering real ones, and the translation between canvas and code is AI-interpreted rather than deterministic. Figma, as the incumbent with the most to lose, has been releasing pieces that start to address this. Their [MCP server](https://www.figma.com/blog/introducing-figma-mcp-server/?ref=philmorton.co) combined with Code Connect and an AI coding tool like Claude Code or Cursor creates a chain that looks promising on paper: design in Figma, map components via Code Connect, feed context to the AI coder via MCP, generate production code that uses your real components. The [’Code to Canvas’ feature](https://www.figma.com/blog/the-future-of-design-is-code-and-canvas/?ref=philmorton.co) even captures rendered UI back into Figma for review. **But in practice, the pieces don’t add up to a seamless flow.** Code to Canvas produces editable Figma layers, but it’s a visual capture rather than a real component mapping. Changes in Figma still can’t write back to code automatically. ## Why no-one has solved this yet This isn’t a case of nobody getting round to it. **It’s a genuinely hard engineering problem.** Figma and tools like it work by rendering 2D graphics on a canvas – in Figma’s case, a custom WebGL engine compiled to WebAssembly. Everything you see is shapes, paths and vectors rendered by a graphics engine, not by a browser’s layout model. A real coded component, on the other hand, is built with HTML and CSS, rendered by a web browser. **These are completely different rendering models.** You can’t just drop a real React button onto a vector canvas, because the canvas doesn’t understand how a browser lays out elements. And you can’t take a vector drawing and reliably turn it into a real component, because the shapes don’t carry the semantic structure of the code. This is why every tool in this space makes the same trade-off. Tools that render real coded components have to embed a browser, which makes it hard to offer a freeform canvas. Tools with great canvases are drawing pictures of components, not running the real thing. **Bridging the two is the core unsolved problem.** ## What to do while you’re waiting Whoever cracks this will capture an enormous market. If Figma solves it, they would cement their dominance. If a startup gets there first, the entire design tooling landscape could get disrupted overnight. Whatever happens, **we’ll probably see it play out by the end of the year.** Of course, you don’t need to wait for the tools to catch up. In the meantime, here are three things you can do: - **Get your design system into code.** This is the prerequisite for everything. Without a coded component library, none of the emerging workflows function. [AI tools only build coherently when there’s a system to constrain them](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/). - **Map out your ideal future workflow with engineering.** Run a workshop with your engineering colleagues to define your use cases (prototyping, production edits, design system changes, etc.) Then work backwards from the process to the tools, rather than starting with the tools and trying to figure out which part of the process they address. - **Experiment with the current tools anyway.** Just because the end-to-end workflow isn’t solved doesn’t mean you shouldn’t be getting your hands dirty. [AI is something you learn by using](https://www.philmorton.co/its-never-been-easier-to-learn-by-doing/). The skills you pick up now – prompting, understanding how these systems work, learning what they get right and wrong – are transferable to any tool or workflow. If you’ve been researching this space, trying to find the tool or the workflow that does it all, don’t worry – it’s not you. **There isn’t an answer that everyone else has figured out.** This is a good time to experiment and plan for when this gap does close, presumably by the end of the year. Get your ducks in a row so that when the dominant workflow and tools emerge, you’re ready to make the most of them. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Workshops are for making things URL: https://www.philmorton.co/workshops-are-for-making-things/ Last updated: 2026-03-11T10:50:07.000Z When you think of a ‘workshop’, you imagine a series of discussions and activities, lots of post-it notes, and an attempt to get alignment. If you’re lucky, some decisions get made. The problem with most workshops is that **from the participants’ point of view, they are primarily there to provide input** and perhaps have their viewpoint on something not-so-subtlety changed. They get some time away from their BAU sharing what they think about a topic, but they’re rarely involved in the creation of an output – that’s done by someone else who writes it up after the session. While that might be fine for the people organising the workshop, it can be an anticlimax for participants. **They spent all day discussing a topic, but have nothing to show for it.** Staying focused and engaged when you’re giving input but don’t see how it’s used can also be tough. Yet this is ironic given the more traditional meaning of ‘workshop’: a place where craftspeople make things. Workshops are much more engaging for everyone when they are closer to this definition. **Have a workshop where people produce something**, not endlessly discuss it. ## People love making things (even accountants) We recently ran a two-day workshop where instead of capturing input from attendees and compiling it into a strategy deck afterwards, we had participants use [Lovable](https://lovable.dev/?ref=philmorton.co) to create a website that encapsulated the discussions and decisions from the session. About 80% of the team had never used a tool like this before. They were amazed at what they could produce and their engagement was super high. At the end, I asked people to guess who had used the most tokens. No-one got it right. Despite it being an IT leadership team, the person who used to most usage was an accountant by trade. Not someone you’d expect to be the most engaged with a vibe coding tool, but that’s the point: **everyone loves making stuff.** It doesn’t matter what discipline people come from. Making brings out a side of people that most jobs don’t allow them to express. Often, the only things they get to create are spreadsheets and slide decks. **When you give most people permission to be creative, they love it**. They feel proud of what they’ve produced and want to show it to their peers. You actually need to allow more time for playback than you’d expect, because everyone wants to share what they’ve done. ## Making forces decisions Asking your workshop participants to make things keeps them more engaged, but there’s an even better reason to do this: **it forces decision-making.** Discussion encourages divergent thinking, which is useful but you can talk all day and decide nothing. **When you create something, you have to make choices.** What goes on the homepage? What’s the name of the product? What do you show first in the video? Think of the classic cereal box exercise, where participants create a cereal box representing their team or product. You can’t have a cereal box with three names. You have to pick one, and that single decision reveals what the team actually thinks is most important. ## What can participants make? Digital tools like Lovable, ChatGPT, Canva, CapCut and so on make it easy for non-technical people to create a wide range of things: websites, videos and more. Of course if you’re in-person, it’s fun to get away from the screen and create posters, cereal boxes, paper prototypes and so on. **The level of polish doesn’t matter.** In fact, it’s better when the outputs aren’t polished because the point is exploration, not production. When choosing what to ask people to make, think about the purpose that making serves in your workshop: - **Making to decide:** As with the cereal box example, the act of creating forces decision-making. The output almost doesn’t matter – it’s the decisions people have to make that are the point. - **Making to surface thinking:** This is closer to a projection technique (like you might see in [brand/market research](https://www.philmorton.co/a-practical-introduction-to-market-research-for-ux-researchers/)). When someone builds a website representing their team, the choices they make (what goes on the homepage, what they leave out, the language they use) reveal what they actually believe, not just what they’d say in a discussion. - **Making to teach a process:** Having people make something can be an effective way to link multiple activities together, like when you’re running training. If the point of the workshop is to teach people design thinking, you can have participants gather requirements, brainstorm and prioritise ideas, prototype, get feedback and iterate on a product that they make rather than running it as a series of lectures. - **Making to communicate:** Create something that carries the message beyond the workshop: a video, prototype or a one-pager. The purpose is less about what happens during the workshop and more about having something that travels afterwards and influences people who weren’t there. ## How this changes your job as a facilitator If one of the outcomes of the workshop is to make something, then **your role shifts from a facilitator to a coach**. You’re helping people get the most out of the tools, offering tips and unblocking them when they get stuck. With tools like Lovable, you need to give people a few tips upfront and then guide them through using it without having to deliver a training course upfront. How you create your agenda is also different, and I think easier than a more standard workshop. When one of the primary goals of a workshop is to produce something, you can **work backwards from the output they’re making** and base the agenda on what participants need to know or do to create it. This simplifies planning enormously because every activity has a clear purpose: it feeds into the thing they’re making. Having done a couple of workshops like this recently, I can’t imagine going back to the more traditional style of workshops. People are more engaged and they finish the day having created something they’re proud of. The next time you’re planning a workshop, **start by asking: *what could participants make?*** You might find that the agenda practically writes itself. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Give in-person research a chance URL: https://www.philmorton.co/give-in-person-research-a-chance/ Last updated: 2026-03-04T10:50:45.000Z I’ve conducted about 2,000 hours of qualitative one-on-one research over the course of my career, and about 90% of it has been in-person. Ten years ago, multi-country projects meant getting on a plane and spending days in a lab was standard practice, while remote research was a rare exception, not the norm. Then Covid happened and like everyone else, we moved to Zoom. The tools worked well enough that most teams never switched back. **Remote research has become the default**, not because anyone made a deliberate decision, but because it was easier. No lab booking, no travel, no logistics. Just press ‘join’. But last week I got the chance to run a day of in-person usability testing for the first time in years. Six depth interviews, back to back, in our office lab. I thought it would be exhausting, but I finished the day energised. It made me realise just how much we’re missing when we only do research through a screen. ## What you’re missing if you’re just doing remote Moderating research in person again **felt like watching 4K TV after years of VHS**. The fidelity of what you can see is just so much better. In a remote usability test, you see the screen but you’re largely missing... - **Hands:** You can’t tell if someone hovers their finger over an option without tapping it. - **Body language:** You miss the moment they lean forward to squint at something, or glance away because they’ve lost confidence. - **Subtle verbal cues:** People sighing and breathing in a different way that indicates they’re frustrated. These behaviours tell you as much about the experience as anything that’s captured on a transcript. **Participants are less distracted too**, since they can’t also be on their phone without you noticing. They’re much more present and the rapport is stronger when you’re sharing the same physical space. Being off-camera can also be a relief. Depending on your lab setup (i.e. if you don’t have a huge two-way mirror in the room), **you can moderate without being visible to observers**. After years of being on camera all day, you might find moderating a little less tiring because of this. ## In-person research sharpens your instincts One of the unexpected benefits of the day was how much it refreshed my intuition as a researcher. When you’ve spent hundreds of hours watching real people use websites and apps, [you develop a feel for what will and won’t work](https://www.philmorton.co/critiquing-design-is-a-researchers-job/). This expertise isn’t a replacement for research, but it makes researchers pretty good at guessing what real people will do when they encounter a design. **Remote research can dull this over time.** You’re only seeing part of the picture, so your pattern recognition has less to work with. It felt like just one day of in-person testing refilled my intuition. ## Practical tips for getting back in the room If you haven’t done in-person research in a while, it can feel like an unnecessary complication or risk to run your sessions in a lab. Here are a few tips for overcoming the inertia and get back moderating face-to-face: - **Tidy the room.** Walk into the space you’re using and look at it through the participant’s eyes. Does it feel welcoming or like a storage cupboard? Get rid of tech that you’re not using. Borrow some plants and other decor from around the office. Get some tissues and hand sanitiser. It’s like having someone over to your house: a little effort goes a long way. - **Be a good host.** Make the effort to welcome people into the room and make them feel comfortable. Take time for small talk. Let people take their coat off. Tell them it’s fine to ask for a drink, want the aircon to be adjusted or take a toilet break. - **Run a pilot session.** When you’re doing remote research, there’s usually not much tech to test. In a lab, there’s more that can go wrong. Running a pilot interview with a colleague helps you get familiar with the setup, not just test your discussion guide. - **Don’t take notes on a laptop.** The screen creates a physical barrier between you and the participant. They can also hear you typing, which signals what you find interesting and what you don’t. Instead use pen and paper, or an iPad with a stylus. Last week I exported my discussion guide to PDF and annotated it in [Goodnotes](https://www.goodnotes.com/?ref=philmorton.co) on my iPad, which worked fine. - **Accept the nerves.** It’s natural to feel a bit rusty. But remember, the participant feels even more nervous than you do. They’re walking into an unfamiliar space, wondering if they’ll say the right things. ## Make in-person research part of your toolkit When people talk about ‘mixed methods’, they usually mean qualitative and quantitative, but maybe we should also include where the research takes place. It doesn’t always make sense to do research in-person, but **blending remote and face-to-face research within a single project** can give you the best of both. It’s also easy to assume that research with niche or B2B recruits can’t possibly be done in-person. But with many companies asking people back to the office a few days a week, there’s a decent chance you can get people into a lab in a big city like London on the right day. And don’t forget that the absolute best place to do research is **going to where your participants actually use the product**. Visiting someone's office or home takes more time, money and preparation, but you can’t beat the insight you get from doing it. Running research face-to-face isn’t as easy as sitting at home, but if you’ve spent the last five years defaulting to remote research, give it another chance. You might be surprised at what you’ve been missing. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### No workshop agenda survives contact with reality URL: https://www.philmorton.co/no-workshop-agenda-survives-contact-with-reality/ Last updated: 2026-02-25T10:50:31.000Z If you’ve got an important workshop coming up, the temptation is to plan out every five-minute slot, create lots of slides and write detailed instructions for every activity. It gives you and your stakeholders the reassuring sense that everything is in hand. But there’s a paradox with workshop planning: **the more meticulously you plan, the more brittle it becomes**. I saw this in a recent three-day workshop for a stakeholder who was anxious about it going well and therefore wanted every session to be tightly defined. We spent hours creating hundreds of slides, planning everything to the nth degree. Of course, what actually happened is that the agenda completely changed during the workshop. We adapted to the topics that emerged, adjusted the timings and some of the content we’d laboriously prepared was never used at all. ## Why workshops resist rigid plans You can’t predict how a group of people (who you may never have met or worked with) will behave. Will they grasp something quickly? Will they need more time? How will they respond to a particular task? Workshops are about exploring a topic and reaching some kind of consensus. You go through cycles of divergent and convergent thinking, and **you can’t always predict what will be discovered**. Sometimes what you find will mean you have to make changes to the workshop. And sometimes the consensus won’t align with your expectations, which can render the exercises you’d planned completely irrelevant. **The best insights tend to come from unscripted moments**: challenging discussions, disagreements and conversations that don’t fit neatly into a pre-prepared canvas. When people come together, especially in person, that’s when ideas should spark. But if you’re rigidly following a plan, you risk suffocating any spontaneous moments. ## Plan like it’s agile, not waterfall When you develop software using waterfall, you define everything upfront and estimate how long each task will take. We all know that **this never works in practice, and yet that’s how we plan most workshops**. Maybe a better way to think about running a workshop is like a more modern, agile project. You’re focused on the outcomes rather than stressing about following a time plan. You have a backlog of activities, some more important than others. You can be flexible with the order, and you can swap things in and out as needed. Running a workshop in a more agile way means **you need more content and activities than you have time for**, so you have options. To do this, you’ll need to build a personal toolkit of tried and tested methods over time. Books like [*Gamestorming*](https://gamestorming.com/?ref=philmorton.co) are a good starting point, but you’ll develop your best material through experience: activities that you know work and that you can pull out when the situation calls for it. ## Facilitate like a conductor, not a playwright **Workshops are much more like improv comedy than a scripted play.** In a play, you say the same lines regardless of how the audience reacts. In improv, you respond to what’s actually happening. It’s the same principle as when you conduct research: your discussion guide is a framework, not a script, and you adapt based on what’s actually being said. Facilitation works the same way. **Your role is to guide the group towards the desired outcome, not to execute a plan.** I did a workshop recently with my longtime collaborator (and all-round legend) [Tim Loo](https://www.linkedin.com/in/timloo/?ref=philmorton.co) that was the opposite of an over-planned client workshop. There were almost no slides, just the agenda and activities prepared in advance. We had **a timeline on a Miro board** with the activities and outcomes we wanted, but as the workshop progressed we adjusted the plan: adding things, removing things, changing the order, shortening and lengthening activities. Workshop participants don’t pay too much attention to whether you’re on track or not, and **it’s better to manage people’s creativity and energy than to stick to a rigid time plan**. ## Making flexibility work in practice You still need a plan. Stakeholders need the confidence that you know what you’re doing and that there’s a structure to the workshop. You won’t get sign-off if it seems like you’re making it up as you go along. The trick is to **treat the plan as a hypothesis, not a script.** Present it as your best prediction of how the time will be used, while communicating that you expect the agenda to evolve as the workshop progresses. **Being flexible doesn’t mean being unprepared** either. If anything, it means you’re *more* prepared. I ran a training session recently where we asked people at the start what they wanted to learn. One thing that came up was trends, which we hadn’t prepared any content for. Instead of sticking to the plan, we added a talk and slotted it in later. That’s only possible if you have a selection of content and activities ready to deploy beyond what’s on the original agenda. A few tips for making this work: - **Focus on the objective, not the agenda.** The goal is to meet the outcome of the workshop, not to follow the predefined plan. Keep that in the front of your mind and let everything else flex around it. - **Co-facilitate if you can.** This is significantly easier with two people. One can present while the other watches the time plan and thinks about what and how to adjust it. - **Experience builds confidence.** The more workshops you run, the looser you can be with the agenda because you trust your toolkit and your instincts. If it’s your first workshop, plan more tightly. Over time, you’ll learn to hold plans more loosely. Upfront planning makes us feel safe and prepared, but the willingness to abandon the original agenda and be flexible is what makes workshops actually achieve their outcomes. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why the value of research is so hard to measure and what to do about it URL: https://www.philmorton.co/why-the-value-of-research-is-so-hard-to-measure-and-what-to-do-about-it/ Last updated: 2026-02-19T09:04:17.000Z Research teams are stuck in an awkward place: **they are meant to be a value creator rather than cost centre, but their impact is hard to measure**. This often puts them in an uncomfortable and vulnerable position in their organisation. If the CFO starts asking questions about *“why are we spending ÂŁ1 million a year on this UX research team?”* then you need to have a good answer. ## Problem 1: intangibility The fundamental challenge is that **the value of research is only ever realised through communication**. Researchers don’t build the product – they advise and influence the people who do. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/02/Measuring-research---2.png) There’s **often no direct line from research spend to product outcome**. You can’t point to a feature and say ‘research built that’ in the same way engineering can. **Instead, your medium for generating value is persuasion**: helping designers, product managers and stakeholders make better decisions because of what you know. The goal of a researcher isn’t to ‘do the research’. It’s to use what they’ve learned about customers to change how other people think and act. This is [why researchers need to think like marketers](https://www.philmorton.co/why-researchers-should-think-like-marketers/) and [a viral video is one of the highest forms of research impact](https://www.philmorton.co/why-every-ux-researchers-new-year-resolution-should-be-to-create-a-viral-video/). **Communication and storytelling aren’t nice-to-have soft skills** for researchers. They are the job. You can discover the most brilliant insight in the world, but if you can’t persuade anyone to act on it, all you’ve done is cost the business money. ## Problem 2: timing Thinking about research at a purely project level, it **front-loads cost and back-loads value**, so the business is effectively investing for weeks before seeing any return. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/02/Measuring-research---1.png) An unscientific chart showing how a research project’s value is mostly realised at the end. At the start of a project, while you’re recruiting, planning and conducting sessions, the business is spending money with little tangible to show for it. Some value trickles in if stakeholders watch sessions live and start seeing things differently. But **most of the value is delivered right at the end**, when you present your findings and recommendations. That final moment carries disproportionate weight and it only works if people listen, pay attention and do something with what you’ve told them. ## How to maximise and extend the value of research If value is primarily created through communication and influence, then research teams need to think carefully about how to maximise it: - **Invest in storytelling and presentation as core craft skills.** Focus on the craft of communicating insights to [make sure they land](https://www.philmorton.co/why-your-research-isnt-landing-and-how-to-fix-it/). Don’t take it for granted that everyone in the team is already 10/10 on this. These skills deserve the same (if not more) attention as research tools and methods. - **Build long-tail artefacts.** Personas, journey maps, design principles, insight repositories and other artefacts that people can refer to over time generate value over time. This is a research team’s ‘passive income’ which turns a one-off project into a long-term asset. - **Close the loop.** Follow up with stakeholders a few weeks after a release to find out what shifted. Track which artefacts get reused and by whom. Keep a simple log linking your recommendations to the product decisions they influenced. - **Develop a system for capturing stories.** Research leaders need to own the collection of stories about how research has impacted the product or service. Building a library of insights and examples of value the team has generated is one of the most important things a research leader can do. There are also ways to increase the value that a research team can create: - **Draw on multiple data sources.** If you’re primarily doing qualitative research, there’s only so much value you can generate. Combine it with data analysis, secondary research, competitor analysis and other methods and you’ll have more insights to work with and more opportunities to create value. - **Enable others through democratisation.** Supporting designers and product managers to conduct their own research allows you to generate a lot of value from relatively few hours of your team’s time. - **Broaden the role beyond research.** [AI’s impact on coding](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/) means that [the future of research is product discovery](https://www.philmorton.co/why-product-discovery-is-the-future-of-ux-research/). Get your researchers involved in concept generation and prototyping, especially now that AI tools make this more accessible than ever. The closer a researcher gets to the making of the product, the less their value depends on persuading someone else to act. By thinking about a research team in a purely commercial and transactional way, we can make sure that we’re delivering enough value to the organisation to justify its existence. ## Why researchers resist this framing This line of thinking doesn’t sit well with a lot of people. *A research team isn’t a business. We shouldn’t be so focused on ROI. If you need to explain why we’re here, you’re doing it wrong!* Many researchers came into the field because they care deeply about understanding and advocating for people. **They see their work as inherently valuable**, self-evidently so. So when they’re asked to quantify impact or ‘sell’ their findings, it can feel like the organisation is questioning the premise of their discipline. There are a few objections that you see: - **It feels reductive.** Researchers deal in nuance, context and human complexity. Being asked to boil that down to ‘we saved ÂŁx’ can feel like it strips the richness and greater meaning out of the work. - **It feels like it shouldn’t be necessary.** Researchers often look at designers or engineers and think that those roles don’t have to justify their existence in the same way. Whether or not that’s true, the perception can create resentment. - **It conflates advocacy with self-promotion.** Researchers are trained to let evidence speak for itself. Having to actively campaign for attention feels uncomfortably close to marketing, which sits awkwardly with a discipline rooted in objectivity and rigour. But the way to reframe it is that **collecting evidence of impact isn’t proving your worth to sceptics. It’s closing the loop on your own work.** If you genuinely care about customers, you should care whether or not your insights actually changed anything. Researchers already know how to gather evidence, find patterns and build a compelling case from data. The final step is to turn those same skills inward and **treat impact tracking as the last stage of the research process**, not a bureaucratic chore. ## The research leader’s role Individual contributors have a big part to play in all of this, but **it’s the research leader who owns the success or failure of the team**. Thinking about their team in a purely commercial way doesn’t come naturally to many research leaders, whose background rarely prepares them for this. Yet it’s their job to ensure that the team generates as much value as possible for the organisation and that this is documented. By **thinking about their team as a system for generating value**, they can make sure that all the parts are working as they should: - Insights are being captured about things that the organisation cares about. - What they learn is communicated effectively, so people listen and make better decisions. - Artefacts are created to extend the lifespan of insights. - The impact of the work is captured systematically through persuasive stories and hard metrics. - Researchers are finding new ways to gather insights and contribute to the building of the product. When a research leader builds this kind of system, it does more than prepare you for the CFO’s questions. It can change how the team feels about its own work. [Research is an anxious place to be right now](https://www.philmorton.co/this-is-the-scariest-and-most-exciting-time-to-be-a-ux-researcher/). But when tracking impact is part of how you operate, **people stop worrying about whether they can justify their existence** because the evidence is there for everyone to see. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### It’s never been easier to learn by doing URL: https://www.philmorton.co/its-never-been-easier-to-learn-by-doing/ Last updated: 2026-02-11T10:50:00.000Z When I was at school, I took part in the a programme run by the entrepreneurship charity [Young Enterprise](https://www.young-enterprise.org.uk/?ref=philmorton.co). They get children to set up and run their own business, with adult mentors to guide them. You sell shares to friends and family, take on a role and try to make a profit running your company. Young Enterprise is all about ***learning by doing***, because that’s the best way to learn. You can shadow people, read books, listen to podcasts and watch tutorials, but actually doing the thing is how you truly understand it. **AI allows us to do more and therefore to learn more.** Tools like [Claude Code](https://claude.com/product/claude-code?ref=philmorton.co) and [Codex](https://openai.com/codex/?ref=philmorton.co) let UX people go beyond their traditional role and actually make the thing they normally just contribute to. With [AI changing how we design and build software](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/), the best way to learn about any of it is to just try it. ## AI has removed the barrier between designing and building In our day-to-day work as designers and researchers, we typically operate at a layer of abstraction from the making of the software. We are in service of others who are writing the code or making decisions about it. **We don’t ‘get our hands dirty’ building the thing ourselves.** Imagine if we were constructing a building. Engineers are the ones laying bricks, quarrying stone, shaping it into place. Designers are drawing pictures of what the building should look like and handing those pictures to the engineers, but they’re not actually building it themselves. Researchers are talking to the people who will eventually use the building, then advising the people drawing the pictures and the people laying the bricks. Everyone is contributing, but only the engineers are actually constructing the thing. That’s how software development has always worked. **But now tools like Claude Code let you go end-to-end**: from idea to working product, without writing the code yourself. It’s also getting easier to try this stuff out. Recent releases like the OpenAI Codex Mac app and Claude Code inside the [Claude Mac app](https://claude.com/download?ref=philmorton.co) are essentially wrappers for the terminal that make the whole experience feel a bit less intimidating. You don’t have to stare at a blank command line. You can dip your toes in through a more approachable interface and just get started. ## Building forces you to learn things you’d never otherwise have to Over the last couple of months, I’ve been building a web app in my spare time using Claude Code. It takes the emails that schools send parents and extracts the tasks and events so it’s easier to keep track of what you need to do. I feel like I’ve been learning faster than any time in my career doing this, because **building a real thing forces you to take on roles that you don’t normally**. When you’re part of a multidisciplinary team, you contribute but when you build something yourself, you *are* the whole team. Because it’s just me (and Claude), I’ve had to learn all of the following: - **Modern web stack:** understanding how web apps are built and the platforms they use. - **Prompt engineering:** writing and evaluating the core prompts that make the app work. - **Technical architecture:** designing an email processing system that is resilient and scalable. - **Security:** understanding how to protect the app from abuse and keep users’ data safe. - **Brand and visual design:** defining how the app feels, choosing typefaces, colours, designing a logo and so on. - **Marketing:** understanding what types of content work on Instagram, and filming and editing reels. - **Legals:** writing the terms of service and privacy policies. - **Pricing:** working out the best price point to sell at and model the commercials over time to ensure it’s sustainable. Before AI, it would have been too intimidating (and time consuming) to attempt all of that. But now **you can have an LLM holding your hand every step of the way**. Whenever you don’t know how to do something or you get stuck, you can ask for help. For ÂŁ20/month, you get a Claude Pro subscription that gives you access to Claude Code and an always-on teacher and safety net. ## Most people are still watching from the sidelines You’d think from reading LinkedIn that everyone is doing this, but from what I can see, **only about 10% of people in most organisations are actively exploring and experimenting with AI** beyond basic usage of ChatGPT. And only 1-2% are trying out coding tools like Claude Code. So if you are using these tools to build, it’s **putting you in a really good place career-wise**. If you’re looking for a job or will be in the future, having experience building products with these tools gives you a significant advantage over other candidates. You’re not just curious about AI, you’re taking steps to understand how it works in practice. Developers have always had side projects – it’s almost expected of you. But UX people haven’t traditionally had side projects because they could do the research and design the experience, but they couldn’t build the end product themselves. Now you can. ## Just get started My advice to pretty much anyone in UX is to **get a side project**. Think of something annoying in your day-to-day life and build a little app around it. It doesn’t need to be polished or original, just real enough to force you to learn. It’ll feel uncomfortable at first. Tools like the Terminal can be intimidating if you’ve never used them. But **AI is there to explain everything and guide you through every step**, and help you whenever you get stuck. You can work at your own pace with no pressure to ship. What you’ll discover is that not only are you learning a lot, but there’s **a huge sense of achievement from building something from nothing**. It’s addictive. Developers have known this forever but designers and researchers might not have tasted it before. As a designer or researcher, **you already have the skills to think end-to-end** about a product. You understand users, you understand the experience, you understand what good looks like. The only thing you’ve been missing is the technical bit at the end, and AI can close that gap. Go and get building and learning. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Can we still trust quant surveys? URL: https://www.philmorton.co/can-we-still-trust-quant-surveys/ Last updated: 2026-02-04T10:50:46.000Z Last year, I was on a panel at the MRS B2B Market Research conference. During the Q&A, one of the audience members shared that in a recent quant survey project, **they had to throw away 80% of responses due to fraud**. There’s always been fraud risk in online surveys. People and bots filling them in just for the money is nothing new. But AI has changed the scale of the problem dramatically. [One study](https://www.frontiersin.org/journals/research-metrics-and-analytics/articles/10.3389/frma.2024.1432774/full?ref=philmorton.co) found that **usable response rates have dropped from 75% to just 10%** in recent years. [Research from Stanford](https://www.gsb.stanford.edu/insights/ai-generated-survey-responses-could-make-research-less-accurate-lot-less-interesting?ref=philmorton.co) found that **a third of online survey takers admit to using tools like ChatGPT to answer questions**. Sean Westwood, a researcher at Dartmouth, [built an AI agent to take surveys](https://www.404media.co/a-researcher-made-an-ai-that-completely-breaks-the-online-surveys-scientists-rely-on/?ref=philmorton.co) and found it could **evade detection 99.8% of the time**. The ‘trick questions’ that researchers have relied on for years no longer work. ## Three ways AI enables fraud Bots and dishonest participants have always been a problem, but LLMs give bad actors powerful new tools. There are essentially three attack vectors: - **Fully automated bots.** These are autonomous agents that complete surveys end-to-end. They can maintain a consistent persona, answer attention checks, generate realistic responses and simulate ‘human’ pacing with deliberate imperfections. Once configured, the reward from a successful completion can far outweigh the cost of running the model. - **Real humans using AI to answer.** Here the respondent is real, but the answers aren’t. Sometimes people do it deliberately to complete surveys for the money, other times they just can’t be bothered to fill in another open text field. Here, some fraud controls may not catch it because the person’s device and identity are genuine, even if their answers aren’t. - **Misrepresentation to pass screeners.** People lie about demographics, job roles or product ownership to qualify for higher-paid studies. AI can help them sound convincing and maintain consistency across a screener and main survey. For all of these, the bigger the reward, the greater the risk. If you’re running niche B2B research with a $300 incentive for CMOs or high-net-worth individuals, this is where you need to be most careful. ## What survey companies are doing Of course the market research industry recognises this [“could become an existential issue”](https://www.research-live.com/article/news/data-quality-existential-issue-for-research-new-podcast-hears/id/5131913?ref=philmorton.co), so an arms race is underway. Panel companies and survey platforms are responding with layered defences: identity verification with government IDs or selfie videos, device fingerprinting, behavioural telemetry and so on. Some B2B expert networks even insist on phone calls with potential participants before including them in studies. All of this increases the cost to recruit participants and run surveys. Now you’re not just paying more for better quality sample, you’re paying to make sure the people answering are real in the first place. ## What researchers can do to reduce fraud risk Setting up a survey has always involved writing screeners and questions in a way to reduce the chance of people guessing their way through, but now you need to do more to counter threats from AI. - **Interrogate sample quality, not just sample size.** Ask providers how respondents’ identities are verified and how they prevent fraud. Think about using your budget for higher-quality sample rather than always going for the biggest possible sample. - **Ask for evidence of criteria to be collected.** Traditional fieldwork agencies have been doing this for years, but it’s rarer in quant. If someone says they are a director of a company, check Companies House. If they say they drive a Jaguar, have them share their registration document. AI actually helps companies do this at scale. - **Design studies to be lower effort.** In the past, if someone started to tire of your overly long survey with too many open-ended questions, they would just give up. Now they can use LLMs to cheat their way to the end. Shorter surveys with clearer questions reduce the risk that real people will reach for ChatGPT to get through your badly-designed questionnaire. - **Use mixed methods.** I know this isn’t anything new, but the greater risk of fraud in surveys means it makes even more sense to use more than one method to answer any question. Triangulate what you’re seeing with other data, including qual research. - **Source from your own customers where possible.** A survey on your website or in your app gives you much more confidence than a general population panel. At least you know people had to be using your product in the first place. - **Consider having no incentive.** It’s the money that attracts the fraud, so try removing it. One benefit of the previous point is that if you have something straightforward to ask, try popping it up on your site before committing to a full paid study. ## How to make decisions about when to use online surveys If quant carries greater risks than before, then you need to be thoughtful about when and how to use it, depending on what you’re using it for: - For **exploratory work** like early concept testing or rough prioritisation, online panels with strong QA are still defensible. But treat numbers as directional rather than the perfect truth. - For **operational decisions** like roadmap prioritisation or tracking satisfaction over time, you need better sample and consistency checks across waves to keep it valid. - For **high-stakes decisions** like market sizing or major business decisions, you need verification-heavy surveys backed up with mixed-methods to reduce the risk. If you can’t afford the quality needed for the bigger decisions, **you may be better off doing less quant, more qual and more behavioural evidence** than doing cheap quant that produces questionable results. **Think: if 10% of responses were compromised, would it change the outcome of your research?** If yes, you need either higher quality, mixed methods, or both. ## Another tool in the toolbox, not the perfect truth machine Online surveys aren’t ‘dead’ or useless, but **the days of assuming that quant’s bigger numbers will always give you greater certainty are over**. You asked lots of people a set of questions... but are they real people? Did they really answer themselves or did they get a little help from ChatGPT? Are they really the type of people you thought they were? **There has never been so much uncertainty about who (or what) is behind the other side of your survey.** Yet despite this, it’s not like we’re going to go back to in-person qual to answer every question. It just doesn’t scale. Imperfect as they are, online surveys are here to stay. **We just have to be more thoughtful about how we use them** and conscious of how they can be compromised to get value from them. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### What the vibe engineering workflow tells us about the future of UX roles URL: https://www.philmorton.co/what-the-vibe-engineering-workflow-tells-us-about-the-future-of-ux-roles/ Last updated: 2026-01-28T10:50:31.000Z Last week I outlined my thesis on [how vibe engineering will turn the product design process upside down](https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/). This week we’re taking a look at the workflow used to write production-quality code with AI, because **this tells us a lot about how UX research and product design roles will change** in the future. ## Change will flow downhill from engineering Every research and design team is trying to figure out how to incorporate AI into their work. Synthesis tools, faster prototyping and so on. **Yet the impact of AI on software engineering is so massive that its gravitational pull will be the primary driver of changes to how UX people do their work.** To be simplistic, most of what we do in UX is in service of creating working software that meets the needs of an organisation and its customers. When you make writing code 10x faster and allow more people to write it, that change is going to outweigh everything else. So by understanding the AI-assisted engineering workflow, we can understand the knock-on impact on UX roles. ## The constraints that shape the workflow If you want to understand how AI will impact UX, the best thing you can do is fire up Claude Code and start building something yourself. For the last six weeks, I’ve been creating a web app for parents that extracts to-dos and dates from school emails. The first thing you learn is that **the** [**vibe engineering**](https://simonwillison.net/2025/Oct/7/vibe-engineering/?ref=philmorton.co) **workflow is dictated by the current limitations of AI coding agents:** - **They have no long-term memory.** Each conversation with the agent starts fresh – it has no context about your project unless you share it. Providing this information is your job, which is why people talk about ‘context engineering’ so much: [quality input equals quality output](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/). - **They have limited short-term memory.** When you’re having a normal conversation with something like ChatGPT, you rarely come close to filling up the context window (think of it as the agent’s working memory). But when you’re coding, it fills up super fast because it’s constantly reading files, documentation and so on. Once it gets close to full, its performance deteriorates and bad behaviour like cutting corners starts to creep in. - **Therefore, they work best on limited, tightly constrained tasks.** Ask Claude Code to “create an e-commerce site” in one prompt and you’ll get *something*, but it won’t be robust. To get quality, you have to work one small step at a time. This is why demos of ‘one prompt to build an app’ are so misleading. That’s not how you create production software. ## The workflow in practice So what does vibe engineering actually look like when you’re trying to build something real? ### 1\. Break down the problem Once you know what you want to build, you need to break it down into the smallest possible tasks (due to the limitations we just touched on). And because you’re not doing everything in a single conversation, **you need to manage the work across tasks** to make sure it all adds up to what you want. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-21-at-07.45.43.png) Sometimes my issues have sub-issues, which have sub-issues, which have sub-issues! I’ve been using GitHub issues for this in my project. For each feature or task, I use ChatGPT to write a structured description of the work and paste this into the issue. The key benefit of GitHub issues is that **Claude Code can read all of the issues** using its command-line interface, allowing it to understand the broader context of the particular task it’s working on. Working in this way can mean you end up with numerous issues for each feature. When I was adding support for email forwarding to my app, this **one feature ended up having over 40 sub-issues** on GitHub. But by breaking down the problem, you’re ensuring that the AI has the capacity to do a good job on each piece of it. ### 2\. Planning Before you ask the AI to write any code, you need to get it to conduct research and create a plan. Claude Code has this concept built in, with a **‘plan mode’** that explores the codebase before presenting its recommendation for how to implement your intent. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-27-at-20.50.33.png) Claude Code presents its plan for you to review before coding. What you quickly learn is to **never trust the initial plan**. You always ask it to review its plan, question its assumptions and check if it hasn’t missed anything. It will *always* find something. This follows a standard rule with AI: the more time you give it to do something, the better the result. For complex features, **you might spend an entire conversation (and context window) just on planning**, going round and round until you have real confidence in the approach. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-27-at-20.48.08.png) Claude Code can ask you clarification questions in plan mode Claude Code hasn’t made any big mistakes in my project so far, except when I didn’t plan throughly enough. One of the most important features I have is backend email processing. **I didn’t give Claude enough context** and it ended up choosing an architecture that would have broken once I had around 20 users. Once I gave it better requirements around scaling, it chose something more suitable. ### 3\. Coding Once you’ve broken down your feature into sub-tasks and planned them out, the actual coding happens relatively quickly. What you learn from trying out vibe engineering is that **the actual coding is not the most time consuming part** at all. ### 4\. Reviewing Similar to the ‘never trust the initial plan’ rule, you **never trust that the code it’s created is 100% complete**. Once it’s finished, always get it to review its work and check it hasn’t missed anything. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-27-at-21.05.47.png) Always get it to check its work. I've created a slash command (I type `/codereview` in the terminal) that prompts Claude to review its work across multiple dimensions like security, scalability, software engineering best practices and so on. **Having a semi-automated process allows you to catch the most obvious errors** before they show up in production. ### The compound effect Hour-by-hour, this process doesn’t feel particularly fast. **You’re constantly checking, verifying and double-checking everything.** But when you look at what you’ve accomplished in a week, that’s where it feels like you’re making progress at a 10x speed. Once you’ve figured out **how to marry established software engineering best practices to AI** so you can work in a disciplined and structured way, you get the speed benefits without too many of the downsides. ## What this means for UX roles The main insight you gain from trying this process out for yourself is that **the bottleneck is *what* to build, not *how* to build it.** **AI coding agents can only go as fast as humans can give them direction.** Well planned and defined work flies by, but the moment you need to work out how something should work, you have to slow down and think it through, taking into account the entire context of the project. This is especially true for anything that is customer-facing. **This means UX skills become more valuable, not less.** The handover process is also different. With a human developer, you can often rely on their interpretation of design files and prototypes, and answer questions as they arise. With an AI agent, **you get better results when you provide more detailed written specifications upfront** because a coding agent is more likely to make assumptions that ask for clarification (although I guess it depends on the human developers you already work with!) **The shape of teams might also change dramatically** over time. To keep the engineering team’s velocity high, you might need more UX capacity to define what to build. ## What to do about it If you’re a fellow UXer, my one piece of advice is to **explore this workflow by building something.** Get a Claude subscription, fire up Claude Code and make something real. You’ll understand more about the future of our industry than any article can explain. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How vibe engineering will turn the product design process (and tooling) upside down URL: https://www.philmorton.co/how-vibe-engineering-will-turn-the-product-design-process-and-tooling-upside-down/ Last updated: 2026-01-21T10:50:27.000Z **The primary impact of AI on product design will come from how coding agents are being used in production**, rather than improvements in prototyping and design tooling. Here’s my thesis: - AI coding agents can now write 100% of production code - This allows products to be built much faster - But to create the front-end UI in code, you need a design system in code - Getting a design system from Figma into code is currently a slow process - Therefore it makes more sense to start production design in code - AI coding agents will allow designers to work alongside devs, building the product directly in code - This will require designers to develop new skills, learn new tools and develop a technical mindset Before we get into the details, I want to be clear on one thing: I’m talking about how software gets made once you are ready to build the real production-ready thing. Product discovery isn’t disappearing, and with these AI coding tools [it’s more important than ever](https://www.philmorton.co/why-product-discovery-is-the-future-of-ux-research/). ## AI coding agents are now good enough With the release of [Opus 4.5](https://www.philmorton.co/with-claude-opus-4-5-i-can-build-software-again/) and recent updates to Claude Code, AI coding agents are now capable of producing 100% of production code. Of course using it to do this requires strict guardrails and structured processes, but **in the right hands a single developer can do the work of ten people** not using these tools. Engineers are developing techniques to go even faster. [Ralph Wiggum](https://github.com/anthropics/claude-code/blob/main/plugins/ralph-wiggum/README.md?ref=philmorton.co) puts a coding agent in a loop where it can run for hours, persisting until it succeeds. [Gas Town](https://steve-yegge.medium.com/welcome-to-gas-town-4f25ee16dd04?ref=philmorton.co) is a structure that gives agents different roles so that they can build software like workers in a factory. Whatever concerns people have about code quality and security will melt away as methods are found to reduce these risks. And as [Steve Yegge](https://steve-yegge.medium.com/the-future-of-coding-agents-e9451a84207c?ref=philmorton.co) reminds us... > *People still don’t understand that we’ve been vibe coding since the Stone Age. Programming has always been a best-effort, we’ll-fix-sh\*t-later endeavor. We always ship with bugs. The question is, how close is it? How good are your tests? How good is your verification suite? Does it meet the customer’s needs? That’s all that matters. Today is no different from how engineering has ever been. From a company’s perspective, historically, the engineer has always been the black box. You ask them for stuff; it eventually arrives, broken, and then gradually you work together to fix it. Now the AI is that black box.* ## You need a design system in code before you can go fast If you want to build software quickly with AI, you need to of course build the front-end as well as the backend. But unless you have your design system in code, the AI is going to be making all sorts of bad decisions about your UI and its implementation. The problem is that unless you already have a robust design system in code (lucky you), **it takes forever to create one** if you start from scratch in Figma. Most organisations’ processes assume that you create designs in Figma, then create a design system in Figma, then hand that over to developers to replicate it in code. This takes ages and leaves a lot of room for mistakes and misalignment. A button with all its variants might take a week to design and build. Drawing pictures of screens in Figma and then handing over to developers to build acts as a handbrake on the whole process. And **the faster developers can go, the less prepared everyone is going to be to wait around for design to do its thing**. ## Start in code, not in Figma Therefore it makes a lot more sense to start your design system in code rather than in Figma. There are several libraries out there like [shadcn](https://ui.shadcn.com/?ref=philmorton.co) that give you accessible coded components out of the box, which you can then style and extend as you please. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-19-at-17.24.40.png) Component libraries like shadcn let you start your design system in code super fast. Type `npx shadcn@latest create` and you can have a coded library of components in your project in a few seconds. Most software uses the same primitives, so **why reinvent the button if you don’t have to?** Starting in code also means there is no Figma-code gap or developer handover where things can get lost. ## The designing-in-code tools gap The only problem with starting in code is that the design tooling is very immature in this space. You can set up [Storybook](https://storybook.js.org/?ref=philmorton.co) to see what your components look like and how they behave, but if you want to design with your coded components on a canvas like you can in Figma, the options aren’t great. I’ve tried a couple of tools recently but they’re either very expensive and/or not refined enough. [Hopefully Figma will release some new features](https://obviouslyandy.substack.com/p/what-figma-needs-to-build) around this at their Config conference in the summer, continuing their trend of pushing it closer to code. In the meantime building, styling and using components isn’t something that every designer can easily do. ## Now designers get to build the product Current tool jankiness aside, the trend is clear. Designers won’t be writing code, but they will be working in and with code. Armed with tools like Claude Code, designers can build the product themselves, working alongside developers. In the same way that Sketch and Figma lowered the technical barrier to designing software after Web 2.0 took web design beyond HTML, CSS and JavaScript, **AI coding agents lower the technical barrier** to building production software today. **When we look back in five years, the current way that software gets designed is going to seem absurdly wasteful.** We essentially draw pictures of software and then hand them over to someone to build. ## To work with code (again), designers need a different skillset If your job changes from designing in Figma and then handing over to a dev, to sitting alongside them as a co-builder, you’re going to need skills such as... - Basic technical skills and knowledge (e.g. how to use version control and GitHub) - An understanding of how software gets built and the platforms used - Breaking work into bite-sized tasks that agents can do successfully - Writing precise instructions and constraints for coding agents - Being able to spot when an agent is going awry or has misunderstood something - The curiosity to learn new things and work in different ways - Lack of fear of the terminal I think a lot of **designers and other UX folk like researchers will migrate to a more generalist ‘builder’ role** where they can truly work on something end-to-end. You do the research, identify opportunities, generate and validate concepts with prototypes and then build the real thing. Equally I think there will be plenty of room for designers who want to focus on the more visual execution side of things. AI isn’t great at creating a novel and differentiated brand and visual identity for a product. The UI it creates are good but don’t standout. When there’s more and more software being built, that’ll be increasingly important. ## Collapsing the distance between intent and reality If what I’ve written turns out to be true (and it may not be), we’ll see a significant disruption in how designers and developers work together, and in the tools they rely on to do that work. While that might sound daunting, I think it’s something **everyone involved in building products should be excited about**. We get to remove points of friction in the process, particularly handovers, and work much more closely together around the real thing we’re building. **This future is an empowering one for designers.** It places greater value on their judgement, taste and decision-making. When building becomes fast and cheap, the hardest problem isn’t how to make something, it’s deciding what’s worth making at all. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### With Claude Opus 4.5, I can build software again URL: https://www.philmorton.co/with-claude-opus-4-5-i-can-build-software-again/ Last updated: 2026-01-14T10:50:06.000Z It’s been 25 years since I built my first website. I learned HTML, CSS, PHP and MySQL, and ran [a semi-successful videogames review site](https://web.archive.org/web/20131027194818/http://www.thunderboltgames.com/) for 16 years. Then while studying computer science at uni, I discovered human-computer interaction and pivoted into UX. I lost interest in coding and haven’t really done any serious development for about a decade. I’ve always been tempted to get back into it, because you get an incredible dopamine hit from making something out of nothing. But despite this, I’ve always been put off by how complex building software has become. So much has changed in the time I’ve been away and **you need to know 10 frameworks to get anything done**. There are so many things that are alien to me: Tailwind, Node.js, React... It’s overwhelming, even for someone with years of web development experience. AI promises to democratise coding, but **so far the technology has felt like more of a cool demo than anything serious**. A couple of months ago, [I tried built a simple dashboard](https://www.philmorton.co/what-i-learned-building-my-first-website-in-a-decade-with-vibe-engineering/) with ChatGPT Codex for my writing stats. It took ages and I had to wipe my work a couple of times to get it functioning. It was promising, but clearly not quite ready yet. But with [the announcement](https://www.anthropic.com/news/claude-opus-4-5?ref=philmorton.co) of Anthropic’s Opus 4.5 model and the latest updates to Claude Code, things are very different. This is [the coding model we’ve been waiting for](https://every.to/vibe-check/vibe-check-opus-4-5-is-the-coding-model-we-ve-been-waiting-for?ref=philmorton.co). ## Opus 4.5 is a watershed moment for software development It’s hard to explain how much of a leap this model is without experiencing it yourself other than ***“holy sh!t, it actually works!”*** Opus 4.5 doesn’t get stuck after 2-3 prompts. It doesn’t trip over itself. It doesn’t get bamboozled by errors. It just keeps going. When they announced this model at the end of November, they also snuck in an update to Claude Code: **plan mode**. This allows you to toggle between planning and execution. When you're starting a new feature, you go into plan mode, tell Claude what you want to achieve, and it reviews your code and documentation. Then it comes back with a plan that you iterate and approve before it starts work. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2026/01/Screenshot-2026-01-06-at-11.13.08.png) This is quite different to using something like ChatGPT on the web. There’s a structured approach built-in. When I tried Codex, it had a plan mode, but that was more limited; it just ensured the model didn’t write code. Claude does a lot more. It asks clarifying questions and really works through the problem with you. If you look at purely the benchmark scores for Opus 4.5 vs other models, it doesn’t look like a big upgrade, but **the harness the model sits in (i.e. Claude Code) makes a huge difference**. Now you can essentially have a competent junior developer at your beck and call. ## What I’ve been building For the last four weeks, I’ve been spending an hour or two a day working on a web app for parents that sorts through all the emails they receive from schools, extracting tasks and events mentioned in newsletters and updates. This is much more complicated than the dashboard I built with Codex but so far, **Claude Code and Opus 4.5 really haven’t run into any problems**. There was one path I went down that I had to ask it to redo, but that was due to a decision we made together to use one architecture over another, which turned out to be unsuitable. The stemmed from me not explaining the context of the app, rather than it making a mistake. ## Now you’re the head chef, not the line cook Based on my experience so far, it’s now entirely possible to create software without writing any of the code yourself. But doing so requires skills that you need to learn. There’s a lot more to it than just *“make me an e-commerce site for selling houseplants”*. I’ve been reading the book [*Vibe Coding*](https://www.amazon.co.uk/Vibe-Coding-Building-Production-Grade-Software-ebook/dp/B0F8C22MDN/?ref=philmorton.co) (which despite being released in late October is already out of date) and one of the main points they make is that **if you are no longer writing the code, your role is to be the technical lead**. AI abstracts away the details like many technologies before it, so the skills that professional software engineers (and the rest of us) need to have to be successful have changed. Now you’re managing a team of agents working in parallel, not just spending days writing one tiny piece of code. ## What this means for product design and research teams All of this is going to have a massive downstream impact on how UX people work. I’ve been writing for a while about how AI coding will [reshape product teams](https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/), [shift the focus of UX research to product discovery](https://www.philmorton.co/why-product-discovery-is-the-future-of-ux-research/) and [allow more people to be a solopreneur](https://www.philmorton.co/the-solopreneur-era-is-here-and-ux-people-are-well-placed-to-take-advantage/). **I’m here to tell you that this is all possible right now.** It’s no longer a case of “when the AI gets good enough”. We have reached that point. It’s happening. **This is not a drill!** The entire software development process is changing. The research and design process many people are familiar with will be upended as a result. If you haven’t started to develop your AI skills, now is the time. Stay tuned for plenty of more in-depth content over the coming weeks. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why every UX researcher’s new year resolution should be to create a viral video URL: https://www.philmorton.co/why-every-ux-researchers-new-year-resolution-should-be-to-create-a-viral-video/ Last updated: 2026-01-08T12:25:13.000Z As I’ve [written about previously](https://www.philmorton.co/why-your-research-isnt-landing-and-how-to-fix-it/), **research is only as valuable as the impact it creates**: > *The goal of a researcher isn’t to “do the research”.* > > *The goal of a researcher is to generate empathy for customers in the minds of the people making decisions about the experience.* > > *If what we learn in our research - however brilliant - is not understood by our audience and does not persuade them to act, it has no value.* How do researchers create this impact? By [**marketing the insights they discover**](https://www.philmorton.co/why-researchers-should-think-like-marketers/) so that colleagues understand what it’s like to be a customer. Video is the best medium to do this with. So if you’re looking for a new year’s resolution, here’s an idea: this year, **create a video that goes viral inside your company**. ## Why video is the highest-impact research output The problem with a lot of research is that people attend the debrief, listen politely, and then the report gets filed away in SharePoint. Everyone forgets what was learned and nothing gets done about it. It’s either not memorable, doesn’t compel them to action, or is just another data point in a very busy work life. **Video makes the customer reality visceral** in a way that reports can’t. A well-crafted video is more engaging, more persuasive and easier to consume than any deck or Miro board. Short videos can cut through the noise of work life like an effective ad. Done well, they can **make people want to do something or changes how they think** about customers. That’s what we’re trying to achieve with our research outputs. ## Three types of research video content Creating an effective video is easier said than done. First, let’s examine the different types of content we could include: ### Type 1: Factual information The most basic type of content we could have in a video are facts about what we did and what we learned. Imagine recording a research report in presentation mode with a few transitions. *We spoke to 18 customers in segment x and three themes emerged...* This type of content is easy to produce, but it’s **unlikely to be memorable or go viral**. If your video looks like a meeting recording, it probably won’t cut through. ### Type 2: Research clips Showing carefully selected moments from interviews can be powerful because it puts stakeholders face-to-face with the reality customers face. One example: when the PS4 launched, we conducted research on the PlayStation Store and discovered that most people couldn’t even find it. We created a short video showing the statistics, the correct way to access the store, and two or three clips of people failing to complete that simple task. That video went viral within PlayStation and led to a project that fixed the issue. After the fix, traffic to the store increased by 400%. When there’s a glaring issue like this one, research clips can be incredibly powerful. But when it comes to giving your audience a broader sense of what it’s like to be a customer, **it can be hard to find the right clips to tell whole story**. Maybe you didn’t ask the right question or perhaps the participant wasn’t articulate enough to do their experience justice. That’s where we might look to the next level... ### Type 3: First- and third-person reconstruction This type of video walks us through an experience and helps the us understand how people think, to give us a sense of **what it’s like being ‘in their shoes’**. Let me show you an example from Johnny Harris. This isn’t a UX research output, but it’s the best example I know of for this type of video: People who know me are probably sick of hearing this by now, but the way he walks the viewer through what it’s like to be an American living at different income levels is brilliant. By talking to us as each of the three personas and showing us screen recordings of Craigslist, Google Maps and so on, you can **see what it’s like to be that person from a first-person perspective**. His video exemplifies what a top-tier research output looks like because it generates empathy and understanding so effectively. [He’s also done another recently](https://www.youtube.com/watch?v=PpyPB3BF-hQ&ref=philmorton.co) for $100 million, $1 billion and $100 billion income levels, which I’d also recommend watching. **Another example** of this type of video was created a few years ago by my colleagues at Foolproof for a major bank, showing what their application process was like. The team applied for the accounts themselves and then created a video that showed time-lapse recordings of them navigating the various websites, plus the various pieces of paperwork they had to deal with. Using one of their houses as a set, they filmed a friend acting through the process, getting progressively frustrated at all the hassle involved in opening an account with the bank. It’s a shame I can’t share it here, because it really brings home how bad the experience is. By showing it in a first- and third-person perspective, it’s extremely relatable to the audience. **This type of video takes a lot of time and effort to produce** compared to stitching together a few research clips, but honestly I think they are worth it. Nothing cuts through like a video that puts you in the shoes of the customer and see things from their perspective. ## Making it happen Most UX researchers don’t have extensive experience of creating and editing videos, so you might feel that this is stretching your skills too far. What I’d point out is that **video editing tools have come a long way recently** and that the barrier to doing this is storytelling, not technical. You can use all the software that people use to create Reels or TikToks. You don’t need advanced editing skills to use [Edits](https://creators.instagram.com/edits?ref=philmorton.co) or [CapCut](https://www.capcut.com/?ref=philmorton.co), or even just iMovie. Production value matters less than the clarity of the insight you’re communicating. ## Tell a story that people won’t forget Imagine turning your insight into a video that is so memorable and persuasive that it **becomes part of your organisation’s culture**. Like a successful advert, it changes how people think about customers and it helps them see the world from their perspective. Decisions get made differently. Priorities shift because of it. That is an incredibly tall order, but that level of impact is the peak of what a UX researcher can achieve. And video is the best way to do it. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### An AI system for generating tailored CVs for each job that you apply for URL: https://www.philmorton.co/an-ai-system-for-generating-tailored-cvs-for-each-job-that-you-apply-for/ Last updated: 2025-12-17T10:49:59.000Z In a recent newsletter, I explained how [**AI and applicant tracking systems (ATSs) are changing CVs**](https://www.philmorton.co/how-ai-is-reshaping-cvs/). One thing candidates are doing is **using AI to create a tailored version of their CV for each application**, rather than just having a single static version that they use everywhere. > *Manually tailoring your CV for each application used to be laborious, so most people just had one version that they used for every role.* > > *Now candidates can use AI (in something like ChatGPT or a special job hunting tool) to create a CV for each application in a few minutes. They might need to do a quick edit for tone, but otherwise this is a relatively easy task for an LLM to do.* Given how competitive the job market is, even if this gives you a 5% advantage, it’s worth doing. ## Ok, so how do you do this? I said that people are doing this, but not how. That’s what today’s newsletter is all about. Of course you could use a paid product like [Teal](https://www.tealhq.com/?ref=philmorton.co), but making a DIY version using ChatGPT gives you more control over exactly how it works. A few friends and former colleagues have been looking for new roles recently, so I created and shared this system with some of them. The reviews are positive – *“it’s awesome”* – so I thought I would share so that anyone else looking for a job can benefit. Let’s get into the detail. ## How it works There are three parts to the system: 1. **A master CV** file which is structured like a normal CV, but is 2-3 times longer than one you would submit. This serves as a library of content for the AI to pick from to make your tailored CV. 2. **A prompt to create a tailored CV**. This takes a job application and then picks relevant content from the master CV to create your tailored CV. 3. **The Marked 2 app** takes the tailored CV (which is in Markdown) and outputs it as a nicely formatted PDF. ## Step 1: Create a master CV file This Markdown file is a super long-form version of your CV, which is then used by the LLM to create your tailored CV. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-12-at-14.42.10.png) I’ve [uploaded an example file to GitHub](https://gist.github.com/phil-morton/2d8b8e52d7ba01994e62e4e5460a6fb3?ref=philmorton.co), which you can download and then edit to reflect your own information. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-12-at-14.43.05.png) Click the ‘Raw’ or ‘Download ZIP’ buttons for the file. Creating the master CV takes time, but of course ChatGPT/Claude/etc can help us here. **Try giving it your existing CV and portfolio**, with a prompt like this: ``` GOAL I need your help to create a 'master CV' as outlined in this newsletter: https://www.philmorton.co/an-ai-system-for-generating-tailored-cvs-for-each-job-that-you-apply-for CONTEXT A master CV is a file which is structured like a normal CV, but is 2-3 times longer than one you would submit. This serves as a library of content for AI to pick from to make your tailored CV. TASK 1. Review the example Master CV file that I have provided you with. 2. Review my CV and portfolio files that I have provided you with. 3. Write a Master CV file for me, using only my information. 4. Output as a Markdown file. GUIDELINES - Maintain the structure and formatting used in the example Master CV file. - Ensure that the master CV you produce only contains my details and none of those in the original example file. - Remember that the master CV should be much longer than a normal CV, so do not worry about the length of the file. ``` Once you have your master CV Markdown file, you’re ready for the next step. ## Step 2: Use a prompt to generate a tailored CV Now that you have your master CV, you can use the following prompt which **creates a tailored CV as a Markdown file based on a job ad and your master CV**. I recommend adding this prompt and your master CV to a project/custom GPT in your favourite LLM, so it’s easy to use repeatedly. Remember to edit the prompt, adding your name where it says `[YOUR NAME]` and check that the file name of your master CV matches yours. **Use thinking/reasoning mode** to ensure that the LLM has enough time to do a good job. It’ll take around 5-10 mins to process, but it’ll be worth the wait. ``` ROLE You are a senior career editor specialising in leadership CVs for design, research and strategy roles. You edit for clarity, relevance and narrative impact while preserving factual accuracy and structure. CONTEXT The file "Master CV.md" contains [YOUR NAME]'s complete professional record, written in Markdown, including full role descriptions, achievements and project summaries. For each job application you will receive a job advert or job description Your task is to create a tailored two-to-three-page CV that keeps the same structure, role order and Markdown formatting as the Master CV, but condenses and aligns it with the job description. TASK 1. Analyse the job advert - Identify the top 5–7 capabilities or themes emphasised (for example leadership, research, service design, stakeholder management) - Note the domains or industries mentioned - Determine seniority level and behavioural traits 2. Map relevance from "Master CV.md" - Review all sections and identify which achievements, responsibilities and projects are most relevant to the advert - Decide which details, skills and examples best demonstrate alignment to the role - Note areas that can be condensed or summarised without losing factual accuracy 3. Write the tailored CV - Use the same Markdown heading and bullet structure as "Master CV.md" - Keep all original job titles and section headings in the same order - Follow these target lengths as guidance: * Header: keep identical contact details and layout * Summary: 150 words in two paragraphs, linking Phil’s career themes to the role * Core Skills: 10 bullet points; choose the most relevant skills and keep the format of each bullet similar to "Master CV.md" * Experience: - Current role: up to 250 words focusing on aspects relevant to the advert - Recent past roles: 150 words each - Older roles: 50 words each * Education: copy what is written in "Master CV.md" exactly. * Public speaking: copy what is written in "Master CV.md" exactly. * Volunteer work: copy what is written in "Master CV.md" exactly. GUIDELINES - Maintain the exact Markdown structure and bullet style from "Master CV.md" - Before the Experience section header, ensure there is a to aid with subsequent PDF exports from Markdown - Use British English - Maintain a professional, evidence-based tone - Use concise phrasing with strong verbs (for example “achieved X by doing Y”) - Never alter role order or job titles - Ensure factual accuracy and consistent formatting throughout OUTPUT FORMAT - Share a link to the tailored CV as a markdown file so that I can download it. - File name format: [YOUR NAME] - [company name] [job title] - [month as three letters] [year].md ``` ## Step 3: Turn the Markdown file into a nicely formatted PDF Now you have a tailored CV, but it’s in Markdown. To easily convert it into a PDF, I recommend using the [Marked 2](https://marked2app.com/?ref=philmorton.co) app. This costs $13.99 one-off, but there’s a free trial. To use it, simply open the tailored CV markdown file in Marked 2, choose your theme and then export to PDF. One essential tip: to [force a page break](https://marked2app.com/help/Special%5FSyntax.html?ref=philmorton.co#pagebreaks), add `` to your Markdown file where you need it. ## A couple of examples Using [the example master CV I shared earlier](https://gist.github.com/phil-morton/2d8b8e52d7ba01994e62e4e5460a6fb3?ref=philmorton.co), here are two CVs the prompt above has created for the following jobs: [**Head of Product Design at Cube**](https://app.welcometothejungle.com/jobs/LOaNaVJd?ref=philmorton.co)**, a leadership role** Here you can see how it’s picked up *“If you are passionate about leveraging technology to transform regulatory compliance...”* from the job ad and put in references to compliance, relations and risk into the summary. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-12-at-14.47.48.png) [**Senior Staff Product Designer at Miro**](https://app.welcometothejungle.com/jobs/iudF2x6Y?ref=philmorton.co)**, an IC role** In this example, it’s gone for *“product and systems design leader”* because it’s noticed the requirement for *“Exceptional strategic and systems thinking
”* in the job ad. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-12-at-14.47.52.png) As you can see, each time it’s used the master CV as a base, but is emphasising different skills and experience, based on the role. The differences may not seem significant between the CVs, but all of the tiny changes add up to a make big impact overall. ## Getting you past the initial screener In today’s job market, having an AI and ATS-friendly CV is essential. If you can go beyond that and tailor it to each application, you’re giving yourself an even better chance of passing that first hurdle. I hope this proves useful for readers who are on the hunt for a new job. Try it out and let me know how you get on. 🎄 This is the last newsletter of the year. Have a great Christmas and look out for the next issue in your inbox on Weds 7th Jan. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why product discovery is the future of UX research URL: https://www.philmorton.co/why-product-discovery-is-the-future-of-ux-research/ Last updated: 2025-12-10T11:28:17.000Z In a previous article, I wrote about how [this is both the scariest and most exciting time to be a UX researcher](https://www.philmorton.co/this-is-the-scariest-and-most-exciting-time-to-be-a-ux-researcher/). **Multiple forces and trends are colliding** at once to change the UX research field faster than ever before. Every day you feel the pressure to do something, everything, *anything* differently. **Leading a research team feels like captaining a ship in a violent storm**. You’re trying to keep it all together, but it’s hard to steer and you can’t see in which direction you’re headed. So what comes next? There are a few possible futures for UX research teams, but I think a big part of the answer lies in pivoting towards **product discovery**. ## The case for specialists conducting evaluative research is eroding For a variety of reasons, it makes less and less sense to have dedicated researchers conduct the basic evaluative research: - **Research democratisation** is a blessing and a curse. It allows more research to be done (a good thing) but it demonstrates that evaluative research isn’t something you necessarily need a specialist to do. - **Platforms like UserTesting** have enabled research to be democratised, but the cost of these is consuming budget that would otherwise go towards headcount. - **Continuous discovery** is a worthy concept but it reinforces the idea that research is an activity rather than a role, and that researchers should focus on more strategic work. - **Budgets are tighter** due to macroeconomic uncertainty, so leadership naturally looks at where you *really* need specialists vs activities that can be performed as part of another role. - **AI is fuel on the fire** of all of the above. It can already moderate simple formative research and it can’t be long before it’ll cover usability testing too. AI also allows people to stretch their skills further – the trend towards ‘[full stack builders](https://www.lennysnewsletter.com/p/why-linkedin-is-replacing-pms?ref=philmorton.co)’ is blurring lines between roles. You could argue that research is better when a professional does it, and I would agree. But the question organisations are asking is: *are alternative approaches good enough?* And increasingly, the answer is yes. This doesn’t mean evaluative research is going away completely. But it does mean that **the business case for large teams doing a lot of evaluative work is under immense pressure**. ## Product discovery matters more than ever AI makes it faster to create prototypes and build production-ready code, but it doesn’t necessarily help you **build the right thing**. As [Melissa Perri points out](https://www.linkedin.com/posts/melissajeanperri%5Feveryone-says-ai-will-revolutionize-product-activity-7397249906035535872-P5Wh?ref=philmorton.co): > *The build trap was never about how fast you could build. It was about building the wrong things. AI doesn’t fix that fundamental problem - it amplifies it. If your team was building features customers didn’t want before, now they can build twice as many features customers don't want, twice as fast.* Faster build cycles exaggerate the consequences of building the wrong thing. Teams can ship more features with less effort, but not necessarily more value. **The real bottleneck is deciding what’s worth building.** Understanding customer problems, identifying opportunities and validating ideas before committing too many resources – this is where the value lies. And this is exactly what UX researchers are well suited to do. ## Researchers are well-placed for product discovery UX researchers already have many of the skills needed for product discovery: - Understanding customers and their problems - Gathering evidence to inform decisions - Validating whether ideas will resonate with people - Triangulating insights from multiple sources What’s new is the context in which these skills are applied and the adjacent activities that researchers should be prepared to do or participate in. ## Going beyond formative research **Researchers need to switch their perspective** to think more broadly about the business and product/service goals, and how any research they do fits into this. Many experienced researchers already do this, but most have plenty of room for development to understand the wider context in which they operate. You are not just ‘doing research’, you are helping your organisation solve a specific problem, look for opportunities to grow revenue or cut costs, etc. **Researchers also need to go beyond just doing the research.** You can solve a lot more of the ‘building the right thing’ problem if you can also do things like: - Generating ideas - Creating low-fidelity concepts - Building mid- and high-fidelity prototypes - Using tools like the [Business Model Canvas](https://www.strategyzer.com/library/the-business-model-canvas?ref=philmorton.co) and the [Value Proposition Canvas](https://www.strategyzer.com/library/the-value-proposition-canvas?ref=philmorton.co). **Building out your skillset in this way then allows you to do all of the following** (rather than just some of them): 1. Gather insight about customers to identify opportunities 2. Generate ideas for solutions 3. Bring ideas to life in a way that feedback can be gathered on them 4. Test the desirability of these ideas with customers 5. Build the case needed to move forward into testing an idea’s feasibility and viability ## How researchers’ skillsets need to evolve Moving towards product discovery means expanding your skillset beyond traditional research methods: - **Mixed methods:** combine qual, quant, analytics, social listening and secondary research to build a more complete picture of what customers are doing and what they need. - **Lightweight concepting:** use tools like Miro, FigJam or even PowerPoint to mock up ideas quickly. You don’t need to be a designer to create a low-fidelity concept. - **Mid-fidelity design:** bringing concepts to life even further allows you to gather more detailed feedback with relatively low effort. - **AI-assisted prototyping:** tools like Bolt, Lovable and Figma Make let you create functional prototypes from a prompt or Figma file. This is especially useful for more interactive concepts that you want to gather feedback on. - **Opportunity framing:** learn to articulate customer problems and business opportunities in a way that connects to outcomes, not just insights. ## Research as a skill, not a job I always thought it was curious that despite being a company founded as a UX research agency, [Foolproof](https://www.foolproof.co.uk/?ref=philmorton.co) has never had a job title called ‘researcher’. Instead, the people doing the research (like me) are ‘consultants’. The founders were always clear that research was a tool and that our job was about finding the ‘win/win’ between what customers want and what the business wants, rather than just representing the customer’s perspective. There are a lot of people in the industry who see themselves as ‘UX researchers’ and **feel that their identity is being eroded** by multiple forces at the same time. There will always be room for specialists, but the industry is increasingly valuing generalists and people who mostly do research as their job need to adapt. **Product discovery is the most natural future home for those of us with deep research skills.** Organisations need to *build the right thing* more than ever, and understanding and interacting with customers is the only way to do that with any certainty. If UX researchers expand their horizons and skills a little, they’ll continue to be immensely valuable for years to come. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### An AI prompt for creating recruitment criteria URL: https://www.philmorton.co/an-ai-prompt-for-creating-recruitment-criteria/ Last updated: 2025-12-03T10:50:47.000Z There are plenty of reasons to work with a specialist fieldwork (a.k.a. participant recruitment) agency when you’re conducting research, rather than a big panel like UserZoom: - You want to speak to hard-to-reach people (e.g. surgeons). - You want a greater control over participant quality (e.g. if you say you own a VW, they will actually check this is true). - You’re conducting research offline (since most platforms don’t support this). When we do this, we typically write a brief which contains the recruitment criteria i.e. who exactly do we want to speak to? ## Getting the right participants is everything As I’ve written before, [participant recruitment is the biggest risk in UX research](https://www.philmorton.co/why-participant-recruitment-is-the-biggest-risk-in-ux-research/). > *The quality of participants is the foundation upon which everything else is built. Get this wrong, and everything that follows is tainted.* Because participant recruitment often takes 2-3 weeks for niche profiles, **when you start a project, the clock is ticking** to get the brief written and signed off by your stakeholders. This can sometimes lead to rushing the recruitment brief, leading to problems later in your project. If you get the criteria wrong or what you’ve written is ambiguous, then you might have... - **The wrong participants showing up in research** → flawed insights and sceptical stakeholders - **Re-recruits necessary** → extra time and cost - **Fieldwork agencies turning down your brief** → extra time to find and onboard new vendors This is where AI can help us, and **the more complex or niche the recruitment, the more it can help**. ## A prompt for creating recruitment criteria I’ve developed a prompt that takes details about your project and who you’re looking for, and turns it into recruitment criteria. You can copy and paste the prompt below or just [use it in a custom GPT I’ve created on ChatGPT](https://chatgpt.com/g/g-692d80a1a09c8191bbd494d0e642986b-recruitment-criteria-generator?ref=philmorton.co). ``` Role: Act as a UX researcher with over 20 years of experience conducting customer research. Your job is to understand the research project the user is working on and then write recruitment criteria for a recruitment brief based on this. Context: Assume that the user will be working with an external fieldwork (participant recruitment) partner. They will be completing a written recruitment brief or quote request form, which will ask for the recruitment criteria to be outlined. Task: 1. Gather the following information from the user. Proceed once you have enough information to draft the criteria. a. Project context - What is the product/service? - What stage is it at (early idea, prototype, live product)? b. Research goals - What are you hoping to learn? - Are there key decisions this research should inform? c. Type of research - Research method - Qualitative or quantitative? - Formative or evaluative? - If evaluative, what are you testing (e.g. prototype, live site, content)? d. Audience - Who do you want to recruit to take part in the research? - Are all the participants part of the same segment or are there any significant differences? e. Constraints or preferences - Time per interview/session 2. Once enough information is gathered, create a set of recruitment criteria by following these steps: a. Review the project context and objectives. - Identify which user behaviours, mindsets, or contexts are most important to understand. b. Define the core participant profile: Identify the key inclusion criteria: these are non-negotiable traits all participants must have. - Use demographic filters only if relevant to the research (e.g. age, income, job title). c. List any key segments or quotas: Define groups you want to compare or balance between. - Avoid quotas unless comparisons are planned in analysis. d. Identify exclusion criteria: Determine who should not be included to avoid skewed results. e. Check for platform/tool specific requirements. If the research involves a specific product, app or service: - List precise usage requirements (e.g. “must have connected a Garmin watch to the app in the last month”). f. Consider logistics and practical constraints, such as: - Location (if relevant, e.g. in-person sessions). - Language fluency. - Device ownership or tech familiarity if relevant (e.g. “must use iOS 16 or higher”). 3. Sanity check the feasibility - Are they recruitable? - Are any criteria likely to result in screen-out issues or misinterpretation? 4. Merge any duplicate criteria (e.g. “Can share their phone screen during a remote video call” and “Available for a 60-minute remote interview within the study window and comfortable showing app usage live”) 5. Create a set of example profiles. These bring the criteria to life in a way that is helpful for the recruitment partner to understand. See the example I have provided below. 6. Finalise the recruitment criteria and share with the user. See the example I have provided below - use the same headers and formatting. Do not use tables. Guidelines: - Make sensible assumptions based on your experience, rather than asking the user to give you every detail. - Write criteria as concise, declarative statements (e.g. “Participants must
” or “Must not
”) - Use plain, unambiguous language free from jargon or internal terminology - Be specific with quantities, timeframes and frequencies (e.g. “logged into the app at least once in the past two weeks”) - Avoid subjective terms like “regular”, “experienced”, or “tech-savvy” without defining them - Express behavioural traits using measurable actions (e.g. “has made a payment using the app in the past month”) - State inclusion and exclusion criteria separately to avoid confusion - Use consistent structure and phrasing across all criteria - Avoid double negatives or overly complex sentence constructions - Specify required combinations of traits clearly (e.g. “must be both a parent and the primary grocery shopper”) - If quotas apply, indicate them clearly next to the relevant criteria - Mark which criteria are essential versus preferred to help prioritise during recruitment - Make each criterion independently testable - don’t rely on assumptions or inference - Avoid embedding screener questions directly - focus on defining the traits to be recruited - Write numbers as numbers, not using letters (e.g. 4, not four). Example output: **All participants** - Currently employed in a finance role such as Finance Manager, Financial Controller, Head of Finance or equivalent (essential). - Working at a UK-based scaleup with 20–500 employees and still in a high-growth phase (essential). - Directly involved in day-to-day financial operations or reporting, producing or contributing to monthly accounts (essential). - Uses cloud-based accounting software (e.g. Xero, QuickBooks, Sage, NetSuite or similar) at least 3 times a week and has done so for 6+ months in their current post (essential). - Able to speak confidently about current workflows, pain points and unmet needs without needing to share their screen (essential). - Aged 18 or over, fluent in English and based in the UK (essential). - Available for a single 60-minute remote video interview during the study window and has a reliable internet connection plus webcam (essential). **Quotas** - 12 participants in total. - Aim for an even mix of company size bands: 20–99 employees and 100–500 employees (minimum 4 in each band). - Seek variety of job seniority: at least 4 Finance Managers and at least 4 Financial Controllers or Heads of Finance. - Strive for a spread of accounting software; no single package should represent more than 50 % of the sample. **Exclusion criteria** - Works for a company that develops, sells or consults on accounting or bookkeeping software. - Professional market researchers or UX practitioners. - Took part in any research for our team within the last 6 months. - Company headcount below 20 or above 500 employees. - Not directly involved in hands-on use of accounting software. **Example profiles** - Finance Manager at a Series B fintech scaleup with 65 employees, responsible for monthly close and board reporting, uses Xero daily. - Financial Controller at a health-tech startup with 120 staff, managing multi-entity consolidation in QuickBooks Online and overseeing payroll and compliance. - Head of Finance at a SaaS scaleup with 300 employees, leading a small team, producing investor KPI dashboards and relying on NetSuite for revenue recognition. - Senior Management Accountant at a green-energy scaleup with 45 employees, preparing management accounts and cash-flow forecasts using Sage Business Cloud. - Finance Operations Lead at an ecommerce scaleup with 200 employees, handling accounts payable and receivable, integrating Shopify with Xero and tracking foreign-currency transactions. ``` ## An example As with many things, [I recommend dictating instructions to LLMs](https://www.philmorton.co/voice-dictation-is-the-biggest-productivity-boost-from-ai-that-most-people-arent-using/) because it’s so much faster and you tend to share more detail. Let’s imagine we share this project context... > *I'm working on the Athlete Intelligence feature in the Strava app – it’s a new AI tool that analyses your workout data and gives you generative AI feedback on your runs and rides. It’s already live, so this is evaluative research, and we’re trying to understand whether people find it valuable, what’s working, what’s not, and what else they might want AI to do in the app.* > > *We want to speak to 12 Strava subscribers – mainly runners and cyclists – and all of them fall into the same general segment, so no need to treat them differently. Every participant should be an active Strava subscriber and have already tried the Athlete Intelligence feature at least once, so we can ground the conversation in real experience.* > > *There’s no requirement to balance runners and cyclists, though we’ll aim for a natural mix across the 12\. The sessions will be qualitative one-to-one video interviews conducted remotely, with participants asked to share their phone screen so we can view the Strava app together during the session.* > > *We’ll use the app to prompt discussion and understand how they’ve engaged with Athlete Intelligence so far. Each session will last 60 minutes. Our main goal is to get a clearer picture of the impact this feature has had, what people are finding useful or not, and what they would like AI to help with next.* > > *Participants should be English-speaking and based in the UK, using either iOS or Android. We are not limiting by OS version, but the app must be functioning properly for them. We’re not targeting any specific age range or other demographic criteria beyond ensuring they are regular users of the app. We’ll exclude anyone who has taken part in recent UX research for Strava or who works for Strava or its direct competitors.* You can [read the full response here](https://chatgpt.com/share/692eee06-f100-8000-9a94-f83c30fe1e97?ref=philmorton.co). ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-02-at-13.47.44.png) ## Adding detail that you never would As you can see, it does a thorough job of working through and specifying exactly who you’d want to talk to. With a bit of refinement and editing, you’re good to go. The other thing you may have noticed is that **it goes beyond writing the standard recruitment criteria by generating a list of example profiles**. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/12/Screenshot-2025-12-02-at-13.49.12.png) For the most complex briefs, this gives the recruiter an idea of the types of people they’re looking for. **Rather than piecing together the various criteria in their head**, you’re spelling out the types of people you want. This is something that most UX researchers would not take the time to do, but AI can do very quickly for us, and it’ll lead to a better outcome. [Try it out](https://chatgpt.com/g/g-692d80a1a09c8191bbd494d0e642986b-recruitment-criteria-generator?ref=philmorton.co) and let me know how you get on. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why you shouldn’t run stakeholder interviews like you’d run customer research URL: https://www.philmorton.co/why-you-shouldnt-run-stakeholder-interviews-like-youd-run-customer-research/ Last updated: 2025-11-26T10:50:31.000Z Stakeholder interviews are a staple of many projects, because **understanding customers is only one part of solving any design problem**. If you have a background in UX research then you might default to treating them like customer research. After all, they are similar in many ways: a 1:1 exploratory conversation with specific objectives. Yet some of the **standard practices in customer research don’t always translate that well to stakeholder interviews**. Adapting our approach can lead to better results. ## Why stakeholder interviews need a different approach When you’re doing customer research, you typically run all your interviews within as short a timeframe as possible. You use the same discussion guide for everyone and the topics you cover are fairly consistent across participants. You probe and explore tangents when relevant, but you need to ensure the research is robust. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Stakeholder-interview---1-1.png) Stakeholder interviews are different: - **They can be spread out over weeks or even months**, giving you time to reflect and adjust between conversations. - **You’re often speaking to people with very different roles and expertise**, from product managers to legal teams to customer service leads. - **The breadth of topics can be much wider** than in customer research, where you’re usually focused on a specific problem or experience. - **Each person has unique knowledge to contribute**, so asking everyone the same questions doesn’t always make sense. - **You know who you’re talking to** before you speak to them, so you have the opportunity to research them before the interview. Instead of recruiting groups of people who are similar to each other and match a screener, we’re gathering a much wider range of disparate viewpoints. ## Start by understanding who you’re speaking to Since we know who we’re going to talk to and (presumably) why we’re talking to them, it makes sense to **do some background research** on them. Before you speak to someone, take the time to understand: - What is their role and what are they responsible for? - What unique perspective or knowledge can they offer? - How might their viewpoint differ from others we’re speaking to? Have a look on LinkedIn, speak to team members about them and try to find out as much as possible in advance. This doesn’t have to take long, but it makes a huge difference to the quality of the conversation. ## Use AI to create tailored discussion guides Once you know who you’re speaking to, you can use AI to create a custom discussion guide for each stakeholder. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Stakeholder-interview---2.png) Rather than using the same set of questions for everyone like you would for customer research, you can **generate questions that are specific to their role and expertise**. This makes the conversation more focused and productive. For example, if you’re speaking to someone in customer service, you might ask about common complaints and pain points. If you’re speaking to someone in legal, you might focus on compliance and risk. AI makes this quick and easy. **I recommend writing a base discussion guide/agenda** and then asking your favourite LLM to add questions or adapt it based on the profile of the person you’re speaking to. ## Build a knowledge base as you go Since stakeholder interviews can be more spread out than customer interviews, and since you’re not asking the same questions to everyone, **it can be harder to recall what was said**. Because of this, it’s even more important than usual to structure your note-taking and analysis: - **Transcribe every interview** (with their permission of course). - **Summarise each transcript individually** using AI, pulling out the key points and themes. This aids your recall later and is a good input for other prompts. - **Store all the transcripts in a central tool** like NotebookLM, where you can query them later. Having a live knowledge base that you can refer back to throughout the project is incredibly useful when you’re trying to piece together what you’ve learned. ## Evolve your approach as you learn The real power of this approach is that you can use what you’ve learned from previous interviews to inform the next one. After each conversation, you can: - **Update your discussion guide** to reflect new themes or gaps in your knowledge. - **Ask AI to summarise what you already know** about a particular topic, so you can focus on what’s missing. - **Bring insights from previous interviews into the conversation**, to test assumptions or explore contradictions. This makes each interview more directed and productive. You’re not just asking the same questions over and over, you’re building on what you've already learned. ## Why it matters Sometimes UX researchers approach stakeholder interviews with less preparation and structure than user interviews, but this risks not learning enough from **one of the most valuable sources of insight on any project**. Understanding the internal context – the constraints, politics and priorities – is just as important as understanding customers. It shapes every decision you make about what to design and how to deliver it. **By taking a more structured (and different) approach to stakeholder interviews**, you can unlock much richer insights and make better decisions. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How AI is reshaping CVs URL: https://www.philmorton.co/how-ai-is-reshaping-cvs/ Last updated: 2025-11-19T10:50:11.000Z If you’ve reviewed CVs in the last year or so, you’ll have noticed something: they look different. Carefully crafted multi-column designs with skill ratings and company logos are slowly disappearing. In their place are plain, text-heavy documents that look more like they were written for a machine than a human. That’s because they were. ## CVs used to be designed for humans Not long ago, UXers treated their CV like a mini portfolio piece. People put a lot of effort into the visual design, using layout and typography to make it a pleasant to read and easy to scan. They kept things concise – one or two pages maximum – and included visual elements like company logos to add credibility. The CV was an artefact designed for human consumption and to get across your personality. You’d have one version that you’d send out for most applications, maybe tweaking a line or two depending on the role. ## AI doesn’t care what your CV looks like We’re in the middle of an arms race in the job market: - AI makes it much easier for people to apply for more jobs. - This is driving up the number of applicants for each role. - Recruiters are responding by leaning more and more on AI to sift through applications. Imagine that you’re a hiring manager with 200 applications for a position. If it takes you 3 minutes to review each one, it would take 10 hours (with no breaks!) to go through them all. These days almost all companies use an applicant tracking system (ATS) to make this process easier. Most use AI and ML to triage, rank and filter candidates, assessing CVs to decide who makes it through the first gate. This has created a vicious cycle where candidates (and AI tools helping them) are adapting CVs to pass automated screening, while recruiters are refining their tools to filter more effectively. ## What a CV optimised for machines looks like ### 1\. Layout: from designed artefact to predictable template These days a CV’s layout needs to be “ATS-friendly”. That means: - Single-column layouts that are easy for machines to parse. - Everything in the main column – ATSs disregard content in the header and footer. - Conventional headings like ‘Experience’, ‘Skills’ and ‘Education’. - Minimal decorative elements that might confuse text extraction. ### 2\. Stripping back and removing formatting and imagery AI-powered ATS systems only see text, so logos, icons and profile photos are largely redundant. Now it makes sense to write your CV as plain text first (e.g. in Markdown) and then add a subtle layer of visual identity on top afterwards. If you want to express your visual design skills and personality, your portfolio and LinkedIn profile are the place to do it instead. ### 3\. Length: less of a constraint than it was It used to be the case that a CV should be no more than two pages. If you could make it one page, even better for a human to read if you are confident that they’ll read it and be impressed. Now that your CV’s primary audience is AI, the two-page limit is no longer a hard rule. An ATM isn’t going to skim read – it’s going to read everything. And the more you can give it to read, the better, up to a point. Obviously you don’t want to go to five or six pages because a human hiring manager is still going to read it, but there’s almost no penalty (if not an advantage) to aiming for three pages. ### 4\. From a single static CV to multiple tailored versions Manually tailoring your CV for each application used to be laborious, so most people just had one version that they used for every role. Now candidates can use AI (in something like ChatGPT or a special job hunting tool) to create a CV for each application in a few minutes. They might need to do a quick edit for tone, but otherwise this is a relatively easy task for an LLM to do. ## Adding it all up As an example, here’s my CV from 2016 and then what mine would look like if we applied the principles above. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/CV-comparison-1.png) The 2016 one makes it easy for a human to scan, but there’s no chance it would rank highly in an AI-powered ATS today. The one on the right is much blander, but a machine is going to have no problems parsing it. ## Form follows function For UX folk who spend their careers designing for humans, it might feel counterintuitive to write your CV for a machine, but that’s the world we live in. The CV’s main purpose is now clearing the automated first gate. Creativity is something you show through your portfolio, case studies and interviews. So the next time you’re updating your CV, remember: you’re not just writing for the hiring manager. You’re writing for the machine that decides whether the hiring manager ever sees it at all. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How to get audience insights for free using the Reddit API URL: https://www.philmorton.co/how-to-get-audience-insights-for-free-using-the-reddit-api/ Last updated: 2025-11-12T10:50:25.000Z Social listening (trawling through Facebook, Instagram, TikTok and so on to find insights) is a method **more commonly associated with brand and market research than UX research**. UX researchers typically conduct desk research to inform their work and might look at social media as part of this, but social listening isn’t really a primary method compared to qual interviews, surveys and so on. **The main reason for this is access to tooling.** Most UX research teams don’t have paid accounts for platforms like [Brandwatch](https://www.brandwatch.com/?ref=philmorton.co) and [Sprinklr](https://www.sprinklr.com/?ref=philmorton.co), making social listening a laborious task that they can’t do at any level of scale or rigour. ## Free social listening data sources If getting a paid social listening tool isn’t on the horizon, then there are a few options you can use. **All social media platforms have APIs** and some are more open than others. I got [ChatGPT deep research](https://www.philmorton.co/getting-the-most-out-of-chatgpt-deep-research/) to document public data sources and their free tier limits, [which you can read here if you want](https://chatgpt.com/s/dr%5F6913046259608191b91e56d84858fa80?ref=philmorton.co). In summary, **Reddit is the most generous and useful source**, because it has millions of users discussing a wide variety of topics, and it takes a lot before you hit the limits of their free tier. ## Building a script to query the Reddit API Having identified the [Reddit API](https://www.reddit.com/dev/api/?ref=philmorton.co) as the best source of free social listening data, I got ChatGPT to help me vibe code a simple Python script that would query it and return the results in a file. With the magic of AI, we can then take a file with thousands of Reddit posts and [use ChatGPT deep research to analyse](https://www.philmorton.co/chatgpt-deep-research-is-useful-for-more-than-just-web-search/) it. For the sake of this post, I’ll skip over the process of using ChatGPT Codex and VSCode to write the Python script, but the process was similar to [my recent post about vibe engineering](https://www.philmorton.co/what-i-learned-building-my-first-website-in-a-decade-with-vibe-engineering/). ## Using my Python script which queries the Reddit API If you’d like to try out the script, [you can find it on GitHub here](https://github.com/phil-morton/social-listening-tool?ref=philmorton.co). There’s full documentation in the link above, but let’s walk through a simple use case. First, download the [reddit-pull.py](https://github.com/phil-morton/social-listening-tool/blob/main/reddit-pull.py?ref=philmorton.co) file. Open it in TextEdit or any other code/text editor and go to line 34\. Here you’ll see some variables you need to update with your own name: ``` APP_OWNER_HANDLE = "philmorton" APP_OWNER_REDDIT_USERNAME = "philipmorton" APP_VERSION = "0.1" DEFAULT_OUTPUT_DIRECTORY = "/Users/philipmorton/Downloads" DEFAULT_OUTPUT_FILENAME = "reddit_results.jsonl" ``` Once you’ve updated the file, open up the Terminal in the directory where you’ve saved it. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Screenshot-2025-11-11-at-10.10.21.png) You’ll need to make sure you have Python and ‘requests’ installed for it to work. I recommend using ChatGPT to help you with this if identify as ‘non-technical’! Once you’re set up, then we can use the script to do one of three things: 1. Search all of Reddit for a keyword 2. Pull every recent post from a subreddit 3. Search within a subreddit It will then output a JSON file with the posts and a .manifest.json file which describes what happened when you ran the script. ## An example query Let’s imagine that we’re helping Sports Interactive respond to the recent launch of *Football Manager 26*. We know that our audience uses Reddit and we can see there’s [an active subreddit](https://www.reddit.com/r/footballmanagergames/?ref=philmorton.co). We want to collect all of the posts from this subreddit in the last month. In the Terminal, we’d write the following: `python3 reddit-pull.py --subreddit footballmanagergames --time month` ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Screenshot-2025-11-11-at-10.23.57.png) By default, it will attempt to get up to 5000 posts. The script is also written in a way that it will stay under the free tier query limit. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Screenshot-2025-11-11-at-10.24.29.png) In this example, it found 991 posts in the last month and wrote them to a file for us: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Screenshot-2025-11-11-at-10.25.25.png) This is why you need AI to help with the analysis! ## Analysing the results with ChatGPT deep research Now we need to make sense of these posts. Open up ChatGPT (or your favourite LLM) and prompt something like this: ``` ### **Role** You are a data analyst specialising in qualitative research and social listening. Your task is to review and interpret Reddit data exported from a custom Python script. The data represents posts and/or comments pulled via the Reddit API in JSON format. --- ### **Context** The JSON file contains public Reddit posts and comments about Football Manager 26 from a subreddit. Each item typically includes fields such as: * `title` (string): Post title * `body` or `selftext` (string): Main text content * `subreddit` (string): Community name * `score` (integer): Upvotes * `num_comments` (integer): Number of comments * `created_utc` (timestamp): Post date/time * `author` (string): Poster username (sometimes “deleted”) The goal is to identify key **themes, sentiments, trends, and outliers** in how people discuss the topic. --- ### **Task** 1. Load and interpret the JSON file contents. 2. Identify **dominant discussion themes** (recurring ideas, keywords, or concerns). 3. Provide **illustrative quotes** (short excerpts) that exemplify each theme. 4. Analyse **sentiment** (positive, negative, neutral) and provide a rough distribution. 5. Highlight **anomalies or unexpected insights** (e.g. a post framing the topic in a surprising way). 6. Conclude with a concise **summary of the overall discussion**, including: * What people care about most * How opinions differ across groups * The general tone or mood of the conversation --- ### **Guidelines** * Focus on meaning rather than raw counts; this is a qualitative thematic analysis. * Treat each post or comment as a text sample; short comments can still contribute to patterns. * Use bullet points, short quotes, and concise section headings for readability. * Avoid listing every post—summarise patterns and back them up with examples. * If multiple subreddits are included, compare tone or focus between them. * Use natural, narrative language rather than code or JSON. * Include brief commentary when patterns contradict expectations or reveal unique perspectives. ``` It will then output [a report](https://chatgpt.com/s/dr%5F691314bd985c81918e0a75824c68230d?ref=philmorton.co) for you like this: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/11/Screenshot-2025-11-11-at-10.50.26.png) ## It’s not perfect but it’s free Reddit is only one data source and it’s not suitable for every project, but you can’t argue with the cost! If you’re a UX researcher, adding quantitative social listening data alongside other sources of evidence will **add depth and richness to your insights**. What I’ve shared here is a tiny MVP for what could be a much broader tool. Imagine having a front-end UI and being able to connect other free public data sources. That’s a project for another day though... ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### ChatGPT deep research is useful for more than just web search URL: https://www.philmorton.co/chatgpt-deep-research-is-useful-for-more-than-just-web-search/ Last updated: 2025-11-05T10:50:41.000Z Most people think of [ChatGPT deep research](https://www.philmorton.co/getting-the-most-out-of-chatgpt-deep-research/) as a tool for trawling the web and producing long reports. But it has another use that’s easy to overlook: **it’s also excellent at analysing, summarising and manipulating files**. ## Deep research = more time on task ChatGPT (and other LLMs) offer different levels of compute depending on what you need: - **Standard mode:** Instant answers for straightforward tasks. - **Thinking/reasoning mode:** Works for up to about 5-10 minutes, for more complex tasks. - **Deep research:** Can run for up to around 40-45 minutes on a much bigger task. When you look at how deep research modes in LLMs are marketed, and the use cases you see online, it’s mostly web search. You give it a task like researching trends in a topic and it comes back with a long report. But there’s a lot more you can use it for if you **think about it as simply the amount of compute time** you have for the task. ## What else you can use it for ### 1\. Data analysis and manipulation Deep research is ideal for reviewing and updating structured files like spreadsheets. Imagine that you have a spreadsheet with 1,000 of the most recent Reddit posts from a particular subreddit. Deep research can go through the file and spot the patterns in the data, if you just direct it to look at that file alone. **It can also manipulate data**. Imagine you’ve run a workshop and generated 300 ideas. You want to rate each one on the impact if would have and how difficult it would be to implement. You could spend hours doing this manually, or you could give deep research the spreadsheet and ask it to do it for you. In my experience, ChatGPT deep research is pretty good at this if you give it clear instructions and enough examples of how to undertake the task. You might hit the time limit, but you can always split the task up into multiple steps. You’ll still need to do some cleanup and checking, but it can save a lot of time. ### 2\. Pattern-spotting and summarising If you have up to 10 documents like meeting notes or interview transcripts, ChatGPT deep research can summarise them for you. Of course the king of this task is [NotebookLM](https://notebooklm.google.com/?ref=philmorton.co), which can handle up to 50 files (or 250 on the paid plan), but while it’s good at interrogating documents, **NotebookLM is not as good at writing as ChatGPT**. I recently used ChatGPT deep research to summarise feedback for someone I manage. I talked to eight people they worked with in 15-minute meetings, and I wanted to identify their strengths, development areas, and recurring themes from the meeting transcripts. I tried this with various models. Of course, it was too much for the standard GPT-5 model to do, so the output was poor. Thinking mode was better, but not verbose enough. Deep research spent 17 minutes reading through everything and produced a much more thorough, nuanced and insightful summary. It’s not as powerful as NotebookLM for larger datasets, but for 10 or fewer documents, you can do a lot with it. ## Tips for getting the best results - **Get AI to write the prompt for you**. Since deep research usage is rationed, draft and refine the prompt using regular ChatGPT first. This way you avoid giving it poor input (and therefore poor output). - **Break tasks down.** Massive datasets or open-ended briefs can hit the limit of what deep research can do. If you run up against the 40-45 minute mark, break the task into multiple prompts and then knit it all together afterwards. - **Give it plenty of examples** of input and output when you want it to manipulate data. For example, if you want deep research to go through a spreadsheet and do something to each row, do a few edits yourself manually to help guide it how to do the rest. - **Try exporting to a table in the chat** if the CSV export doesn’t work. I found this to be very unreliable – it would show data in the report but then failed to create a CSV accurately. To solve this, get it to output to the chat where you can easily copy and paste it into Excel yourself. ## For when any task needs more time Deep research is marketed as a tool to help with in-depth web research, but it can do a lot more than that. When you need help with any task which is beyond the capabilities of the thinking/reasoning models, give it a go. It’s not perfect but it can be incredible powerful and **save you hours of manual work**, even taking into consideration the time you need to prompt it and check its work. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### What I learned building my first website in a decade with vibe engineering URL: https://www.philmorton.co/what-i-learned-building-my-first-website-in-a-decade-with-vibe-engineering/ Last updated: 2025-10-22T09:50:07.000Z Despite having a Computer Science degree and years of hobby web development under my belt, I haven’t built any software in over 10 years. That was until last week, when I decided to give ‘vibe engineering’ a go and create [a dashboard to track my writing stats](https://ghost-writing-stats-dashboard.vercel.app/?ref=philmorton.co). The platform I use for this newsletter ([Ghost](https://ghost.org/?ref=philmorton.co)) has decent analytics, but they’re focused on lagging indicators like subscribers and page views which I have limited influence over. Instead, **I wanted a way to track my inputs** (how much I write and how consistently) as these are leading indicators I control. [![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/10/Screenshot-2025-10-17-at-15.48.09.png)](https://ghost-writing-stats-dashboard.vercel.app/?ref=philmorton.co) The end product is built with Next.js, TypeScript, Tailwind and shadcn/ui for the interface. ## Vibe coding vs vibe engineering I suppose I could have vibe coded this by putting in a single prompt and tweaking the output, but I wanted to try a more structured and intentional approach. Vibe coding is great for quick, throwaway prototypes, but the end result can be brittle and difficult to build on. [Vibe engineering](https://simonwillison.net/2025/Oct/7/vibe-engineering/?ref=philmorton.co) is a slightly different approach where you **treat AI as a collaborator, not a magician**. You might get the LLM to write all the code for you (as I did) but you’re following more standard software engineering practices at the same time. **The aim is to build something robust and extendable**, rather than something that just works once. Another benefit of taking a slower and more deliberate approach is that you’re more likely to learn something from the experience. ## The tools and setup ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/10/Screenshot-2025-10-11-at-16.21.30.png) ChatGPT Codex (panel on the right) inside of VS Code. Here’s what I used and why: - [**ChatGPT Codex**](https://chatgpt.com/features/codex?ref=philmorton.co): A coding agent from OpenAI that you can use on the web, in the terminal or in a code editor. I chose this rather than Claude Code since it’s already included in my ChatGPT Plus plan. - [**VS Code**](https://code.visualstudio.com/?ref=philmorton.co): This is the most popular editor in the world and several AI-focused editors like Cursor are actually based on it. I wanted to be able to see the code and work with Codex inside the app. - **ChatGPT Mac app**: This has a couple of features which are super useful for coding: - You can give it access to apps on your desktop so it can read and write content directly. - It can take quickly take and share screenshots of apps, which you’ll be doing a lot of in something like this! - [**GitHub**](https://github.com/?ref=philmorton.co): To store my code and handle version control. - **The Terminal app**: I guess this is the part that scares a lot of non-technical people off. Using the command line can be a bit intimidating but ChatGPT can hold your hand and tell you what to do. ## Making lots of progress... until you don’t Jon Hickman at Ghost gave me a head start on [understanding the API](https://www.theplan.co.uk/counting-posts-on-ghost/?ref=philmorton.co) that the platform uses. I then gave this, plus an explanation of what I wanted to do, to ChatGPT. **I did a lot of planning with ChatGPT**, going back and forth to understand the technology and tools I would need, how it would work and so on. To summarise this, I got it to create a project brief based on our discussion and output this as Markdown, so I could give it to Codex as context later. ChatGPT helped me set up my development environment, instructing me how to install and configure all the tools I needed (git, Codex, etc.) I then got it to **produce the stats I wanted without any UI** so we could check that it worked. This worked pretty much the first time, which felt like magic. Then the pain began! I wanted to use [shadcn](https://ui.shadcn.com/?ref=philmorton.co) components for the UI, but every time I tried it ended up with rendering issues and broken layouts. **I spent hours and hours** failing to get it to debug the issue, rolling back the code and trying again. In the end, I started from scratch, building the UI first with no backend logic. Once I was happy with the front-end, I got it to build the code that queried the API and did the calculations separately. Only then did I get Codex to wire the two together. Even though Codex wrote all the code, **the experience was similar to writing code by hand:** you make really good progress and then you hit an obstacle that takes hours (or more) to overcome, then it’s back to making progress again. ## What I learned from doing this ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/10/Screenshot-2025-10-17-at-17.19.57.png) You can give ChatGPT’s Mac app access to read and write content in other apps. ### 1\. You still need (to develop) a technical mindset AI means that you don’t need to be able to write code to create software, but what **you still need software engineering skills**: - Researching and planning - Building incrementally - Problem solving **Software is complex and fragile.** It only takes one small thing for the whole system to stop working. You need to be able to work in a way that minimises problems and when they do show up, be able to work through them. AI can do a lot but but it can run into dead ends. When it does, you need to help it get back on track. ### 2\. Collaborate with AI, don’t delegate everything to it I wrote about this [in a previous newsletter](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/), but it’s much more effective to treat AI as a partner rather than a tool. In this project, I was constantly asking it to explain what it was doing and why, how things work, what alternative options are available and so on. **I wanted it to feel like having a senior engineer next to me**, coaching me, rather than someone to outsource the writing of code to. This way you can learn faster, understand the ‘why’ and feel more confident in building on top of something later. ### 3\. Learning through friction builds empathy and capability If I had just fired off a prompt, vibe coding style, I wouldn’t have learned anything about how to build anything. Coming up against a problem and spending hours going around in circles was annoying, but now I’ve learned how to approach my next project differently. You **learn by doing** and pure vibe coding doesn’t really involve much ‘doing’. You also get a bit of empathy for developers. Even building a tiny one-page app like this has taught me more about modern web development than anything else would. ## Yes, you can do this too We’re at the start of a new era in which anyone can create software. But if you want to go beyond simple vibe coded prototypes (and cool demos) **you need to get your hands dirty**. That means using the terminal, learning to use GitHub and so on. It can seem a bit scary at first if you’ve never done any coding, but there’s an opportunity to learn so much. Even as AI coding tools get better, **the value of software engineering principles and best practices are still going to hold true**. Experiencing those first-hand through a few tiny side projects like this is going to be so worthwhile. Don’t just wait until AI can do everything to a perfect production-ready level. **Have a go** now and remember that your favourite LLM can hold your hand through the whole process, one step at a time. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The solopreneur era is here (and UX people are well placed to take advantage) URL: https://www.philmorton.co/the-solopreneur-era-is-here-and-ux-people-are-well-placed-to-take-advantage/ Last updated: 2025-10-15T09:50:46.000Z As a UX person (designer, researcher, etc.) your career options are broadly: 1. Full-time employee in an organisation 2. Freelancer/contractor 3. Co-founder of a startup (with someone more technical) However, developers have always had another option: being a solopreneur. **This is someone who starts their own business, but it’s just them.** They do all the product development, marketing, distribution, support, etc. Think of [Marco Arment](https://marco.org/?ref=philmorton.co) with [Overcast](https://overcast.fm/?ref=philmorton.co) or [David Smith](https://david-smith.org/?ref=philmorton.co) creating [Widgetsmith](https://apps.apple.com/gb/app/widgetsmith/id1523682319?ref=philmorton.co) (100m+ downloads). If you want an even more in-depth example, Suhail Idrees left his job at Bain to create AI video editing tool [FireCut](https://firecut.ai/?ref=philmorton.co), which makes $1m a year: **People with software engineering skills have been able to take this alternative career path for decades**, but AI means that this option is starting to open up to non-technical people too. ## The solopreneur era is here The big difference between a freelancer and a solopreneur is scale. **If you’re just selling your time, your income is limited by how much time you have.** If you’re making a piece of software, it can potentially reach (and earn money from) millions of people. There are other types of solopreneurs (e.g. YouTubers, people selling online courses, etc.) but let’s focus on software for now. For years, **the infrastructure needed to build software has been getting more and more accessible**, with complexity being abstracted away. AWS means you don’t need to manage physical storage, the App Store handles distribution, Stripe simplifies payments and so on. Many apps are essentially a layer that sits on top of APIs. It’s never been easier to create a digital product and the final barrier for most of us – being able to write code – is now being lowered by AI. ## When anyone can write code, the question is what to build Given the progress we’re seeing with AI coding tools, we’re not far from a world where anyone could (in theory) start and run a one-person software business. You could argue we’re already there. In that world, **the question goes from being “*can you build it?”* to “*what should you build?”* and “*what will people pay for?”*** With AI tools making execution much easier, the biggest barriers to success are: - Identifying real pain points worth solving - Validating ideas before committing too many resources - Designing something people actually want - Understanding the market well enough to differentiate In other words, the skills that many UX people already have. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/10/Screenshot-2025-10-14-at-14.41.44.png) Vibe coding tools like Lovable let anyone code super fast, but the results can be brittle and when you run into a problem, it can be hard to debug. ## Why UX people make great solopreneurs If you’ve spent years working in design and research, you already know how to: - Spot problems in everyday life - Interview people to learn about their pain points - Generate and validate ideas through prototyping and testing - Iterate based on feedback There are a bunch of books for wannabe solopreneurs with cringey titles like *$100m Offers* and *The Millionaire Fastlane* which talk about how to come up with great business ideas. **The techniques that they promote are essentially just good user research**: social listening, shadowing people, reading reviews, conducting jobs-to-be-done interviews. It’s the same mindset, just applied to finding business opportunities rather than improving an existing product. The other advantage UX people have is [‘taste’](https://www.philmorton.co/why-everyones-talking-about-taste-as-the-next-differentiator/). When AI can generate designs and code, the ability to judge quality becomes more valuable. Someone with thousands of hours of design or research experience will be better at spotting what will work and what won’t. Experienced UXers who go down the path of creating their own software products have a huge advantage over those that don’t, even if they may lack the technical skills. ## The two ways to build with AI For those with the curiosity to explore this further, there are a couple of approaches to building software with AI: - **Vibe coding:** tools like Lovable, Bolt and v0 let you describe what you want and the tool builds it for you. You can go from idea to working prototype in minutes or hours. But the code is a bit of a black box, so when you run into issues it can be hard to debug. - **AI-assisted coding:** tools like Windsurf, Cursor and GitHub Copilot are primarily designed to create production-ready software that you understand and control. Simon Willison calls this [‘vibe engineering’](https://simonwillison.net/2025/Oct/7/vibe-engineering/?ref=philmorton.co), which I like. You’re getting AI to accelerate your coding (or write all of it), but you’re using software engineering best practices like automated testing to build something that is less brittle than what you’d get with vibe coding. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/10/image.png) Tools like Cursor require more technical knowledge to use, but give you more control and more robust results ## It’s time to get a side project! If you’re interested in this idea, you don’t have to quit your job and go all in. You can start with a side project. **Developers have always had side projects** – it’s almost expected. But UX people rarely do, partly because we're not usually the ones building the thing. AI can change that. Even if your side project never becomes your main income, making a little piece of software will... - Stretch your skills in new directions - Give you a newfound appreciation for software engineering - Help you understand the commercial side of running a business - Make you more marketable in an increasingly competitive industry It was a side project that got me into UX in the first place. I created and ran [a videogames review website](https://web.archive.org/web/20141010201756/http://www.thunderboltgames.com/) for 16 years. This taught me how to design, code (before I forgot it all), manage people and more. **If you’re curious, here’s a practical starting point**: whenever you notice something irritating in your own life or someone mentions a problem, write it down in a ‘problem list’. Later, when you want to try a side project, you’ll have a shortlist of ideas to explore. ## A new opportunity The solopreneur career path isn’t just for developers and content creators anymore. In an uncertain job market and with AI reshaping everything, having another option is welcome. Not everyone will want to run their own business, but for those who do, the barriers are lower than they’ve ever been. Even if you never create something so successful that you can quit your job, **tinkering with side projects will make you a better practitioner**. You’ll start to understand engineering trade-offs, think more commercially and become more well-rounded – exactly the kind of person who’ll thrive as the lines between design, product and technology continue to blur. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### This is the scariest (and most exciting) time to be a UX researcher URL: https://www.philmorton.co/this-is-the-scariest-and-most-exciting-time-to-be-a-ux-researcher/ Last updated: 2025-10-08T09:50:36.000Z In the first ten years of my career, it felt like the craft of research didn’t change that quickly. You could learn a method like personas or diary studies and then move on to the next one to slowly build your craft. Sure, there were new ideas that you’d have to incorporate into your practice – the rise of mobile, remote research, hypothesis-driven testing, the start of continuous discovery – but it was manageable. These days, everything feels like it’s in flux. It means there are a lot of opportunities to reshape how everything works – which is exciting – but it’s also the most unsettling time in a generation to be working as a UX researcher. ## The forces reshaping research There are broadly three big trends colliding to make a big mess for teams in our industry: 1. **AI:** Not sure if you’ve heard of it, but kind of a big deal... Every week it feels like [*everything is changing!*](https://youtu.be/hS1YqcewH0c?ref=philmorton.co) But at the same time, adoption is mostly at the speed of large organisations (slow). 2. **Evolving product models** are changing how research fits into the wider process. Continuous discovery, democratised research, research ops, strategic research blending into market research... all of these were already in motion before AI arrived and [now they’re colliding with it](https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/). 3. **Macroeconomics** are not helping. Ukraine, Trump, the cost of living, climate change, debt, etc, etc. Companies can’t predict what will happen next so they’re cautious about investing. Yet they also need to *change everything* because of AI. ## A million unanswered questions Every research team has an endless list of things to figure out right now: - Which parts of our workflow should we augment with AI versus fully automate? - Should AI moderate user interviews, and if so, when and how? - [How do we defend against fraudulent participants and bots](https://www.philmorton.co/why-participant-recruitment-is-the-biggest-risk-in-ux-research/) using AI to fill in surveys or fool AI moderators? - How does this change our carefully assembled process for democratising research? - What’s the AI-native version of personas, journey maps, or other core deliverables? - How do we ensure quality and rigour when analysis is AI-assisted or automated? - Who is responsible for setting up and maintaining the “plumbing” of AI research workflows? - [How do we make sure that any AI tools we spend 6-12 months getting through procurement won’t just disappear overnight?](https://www.philmorton.co/ux-research-teams-should-start-with-general-purpose-ai-tools/) - What’s the role of ethics and consent when participants interact with AI moderators? - How do we integrate research when design and engineering are also AI-driven? - What’s the balance between fast, AI-driven insights and slower, human-led depth? - All these platforms are great, but what about in-person research? What should we do with our lab? - How do we use these new tools to [market our insights internally](https://www.philmorton.co/why-researchers-should-think-like-marketers/)? - [Will AI replace entry-level research roles](https://www.philmorton.co/will-ai-really-replace-junior-roles/) and if so, how do juniors build experience? - Should researchers become generalists, or double down on specialisms AI can’t reach (like physical or regulated contexts)? - How do we plan a career when the ladder is clouded in fog and the next rung isn’t visible? Good luck being a leader of a research team right now! đŸ«  The thing that feels impossible about all of this is that **the answers to most of the questions above are either constantly changing or unanswerable**. It feels like you can’t pin down what the world will be like in 6 months, let alone 2 years. If you can’t predict what will happen, how can you make any decisions about anything? ## Speculating about the future of research We don’t know what will happen, but let’s do some guessing. Here are a few possible directions that UX research roles could go in the next few years: - **Specialists:** You go where AI can’t. If AI takes over the bulk of everyday insight gathering, then researchers will end up focusing on domains where AI can’t reach or isn’t suitable: physical products, mission-critical systems, regulated environments, etc. It becomes an even more niche role. - **AI ops managers:** Rather than doing the research, your job is to manage AI agents that do. You set up workflows, manage tools and ensure quality, much like a DevOps role but for research. - **Generalists:** You’re not a *researcher*, you’re a *product builder* who uses research alongside strategy, design, market analysis and service design. Your role is to make new products and features using whatever tools and methods you need. - **AI trainers:** As an expert in research, you write evals for AI to train it how to moderate sessions, analyse research and so on. [Experts are already in very high demand](https://www.lennysnewsletter.com/p/experts-writing-ai-evals-brendan-foody?ref=philmorton.co) for this task. - **Career shifters:** You move laterally into product management, design, engineering or something else where you can use your (very transferrable) skills. What will a research team look like in five years? Will we even have a role of ‘researcher’? We don’t know. ## What to do All of this uncertainty is uncomfortable, but there are some practical things we can do: - **Accept that it’s stress-citing.** As I’ve heard [Tim](https://www.linkedin.com/in/timloo/?ref=philmorton.co) say a few times, some of the most interesting times in your career are when it’s stressful and exciting. Two things can be true at once: it’s uncomfortable *and* it’s an opportunity to make things better. - **Treat every project as an experiment.** Don’t take your methods for granted. With each piece of work, ask yourself: is there a better way to do this? What’s one new thing I can try? - **Broaden your skill set.** Don’t limit yourself to moderating research and writing reports. [Learn adjacent skills](https://www.philmorton.co/why-and-how-to-increase-the-breadth-of-your-skills/): design, strategy, market research, data analysis and so on. Having transferable skills is the best hedge against an uncertain future. - **Build AI literacy, but don’t obsess over it.** You don’t need to be the number one AI expert in your company. Spending a couple of hours a week [watching videos](https://www.youtube.com/@howiaipodcast?ref=philmorton.co), [reading articles](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/) or [listening to podcasts](https://www.lennysnewsletter.com/podcast?ref=philmorton.co) will put you in the top 20% of most companies. Then integrate what you learn into your work. - **Stop comparing yourself.** There are pros and cons to being on the cutting edge versus being a fast follower. The people ahead might gain an advantage, but they might also waste time going down the wrong path. Don’t compare yourself to what you see on LinkedIn. - **Focus on mindset over mastery.** Curiosity, openness and pragmatism matter more than chasing every new tool or technique. Everyone is trying to figure this out. Don’t worry: everyone else feels like they’re playing catch-up too. - **Delegate.** If you’re a research leader, don’t try and do it all yourself – you’ll burnout before you manage to answer all of the questions above. Lean on your team and give them opportunities to shape the future of research. This is an uncomfortable time for many people in research. But it’s also an exciting opportunity to change how everything works. We can’t see the future, but what we do know is that **the demand for understanding how customers think and act keeps growing**. Yet *how* organisations gather and use insight in the future is an open question. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### A style guide for generating concept designs with AI URL: https://www.philmorton.co/a-style-guide-for-generating-concept-designs-with-ai/ Last updated: 2025-09-30T09:50:38.000Z 💡 There are a lot of images in this week’s newsletter, so it might be easier to view on a larger screen. [Last week](https://www.philmorton.co/how-to-prompt-ai-to-generate-images-for-concept-design/), we looked at how to prompt AI to **generate images for concept designs, storyboards and research stimulus**. To reiterate, there are two core elements to any image prompt: - **The scene:** *What* is being shown in the image, e.g. a woman looks at her phone on a train - **Visual style:** *How* the image should look, e.g. a watercolour or sketch. This week, I’ll share **a selection of visual styles that you can use in your prompts**, so you can skip the time consuming trail-and-error in finding one that works. ## Scene prompts Below, I’ll share the same scenes with different styles. For reference, the first parts of each prompt are: > *A wide-angle shot of a railway station.* > *A woman in her mid-30s is standing on a railway station platform, using a smartphone.* > *A close-up of a smartphone, which fills most of the image. It is held in one hand by an adult. The smartphone screen shows train times. The UI is a blocky concept wireframe with no detail and no text. In the background, there is a railway station platform. We only see one of the user’s hands holding the phone.* **To replicate any image**, take the first part (above) and add the style (below). ## Black and white illustration > *Minimalist black-and-white grayscale illustration with clean, precise linework and smooth tonal shading. Fewer sketch lines, with a flatter, more elegant aesthetic. Rendered like a concept illustration — calm, modern, and sophisticated, with natural light and subtle depth. Not cartoonish, but polished and design-focused. No text or writing in the scene.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1720_Elegant-Station-Illustration_simple_compose_01k4qmd4d0eh6vjct08r3vc31b.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1223_Modern-Commuter-Elegance_simple_compose_01k4q3dha9fmptwj68ahjthn5y.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1722_Sleek-Train-Schedule-Display_simple_compose_01k4qmh382f1vrhsastmz8pya1.png) ## Sharpie > *Minimal sharpie pen style. Black and white with thick outlines and block shading. Not too detailed.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1726_Minimalist-Railway-Station_simple_compose_01k4qmrc7fezevtzs7b94em75t.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1727_Woman-at-Train-Station_simple_compose_01k4qmtefsf3b9qz448rasxada.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250909_1728_Smartphone-at-Station_simple_compose_01k4qmx77afjq80gz0s0e6g6x1.png) ## Sketchnote > *Sketchnote style with hand-drawn black ink lines, uneven strokes, doodle-like forms, simple icons, and plenty of white space.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1710_Sketchnote-Railway-Station_simple_compose_01k5esce1zescrbcjfh165dxax.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1707_Woman-at-Station-Sketchnote_simple_compose_01k5es853bf6fb3838psm2vz74.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1713_Train-Schedule-Sketchnote_simple_compose_01k5esgq2cfknak4y9gw6dgcke.png) ## Pencil sketch > *Pencil sketch style with light graphite shading, soft lines, and visible pencil strokes. Minimal detail, loose and informal.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1425_Railway-Station-Sketch_simple_compose_01k5efz6a2ew0s4pz4htcpzvcm.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_0932_Woman-on-Railway-Platform_simple_compose_01k5dz6d2hf6zs90skeffprdx4.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1426_Smartphone-Train-Schedule-Sketch_simple_compose_01k5eg12tne68tqd6r5harhfh6.png) ## Character design > *Simplified character design style with flat shading and clean lines. No outlines.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4sh6sx7ertbx84aqxrr4r5g-1757498606_img_1.webp) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4sh1e1ffy2sm9rzj4tjhtx2-1757498431_img_1.webp) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4shfgemf0v8y1va6j6y5js2-1757498848_img_1.webp) Note that on this one, the colours are not consistent because we haven't specified a palette. ## Isometric vector illustration > *Isometric vector illustration at a consistent grid angle, flat colours with restrained shading, clean edges and controlled depth cues. Minimal background clutter, limited palette. No text.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250910_1139_Isometric-Railway-Station_simple_compose_01k4skajwjfm19t94623cpgn99.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250910_1136_Woman-on-Platform_simple_compose_01k4sk405qemyr47amr7fc3pgx.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250910_1140_Smartphone-Train-Times_simple_compose_01k4skca9wetttk8fk64qkptcy.png) Hopefully having this selection of visual styles will give you a better starting point for creating your own images. Have you used another style that’s worked well? Another one I should add? [Let me know on LinkedIn](https://www.linkedin.com/in/philipsamorton/?ref=philmorton.co). ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How to prompt AI to generate images for concept design URL: https://www.philmorton.co/how-to-prompt-ai-to-generate-images-for-concept-design/ Last updated: 2025-10-03T10:07:15.000Z In the early stages of discovery, when you’re exploring and shaping a new proposition, it’s common to create concept designs or storyboards. **Visualising how a new feature or business will work** is useful for aligning your team, gathering feedback from customers and communicating with stakeholders. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250918_1707_Woman-at-Station-Sketchnote_simple_compose_01k5es853cevjrcwcyeqsqcy7z.png) These visuals don’t have to be perfect, yet creating them can be time consuming and in this stage of the product development process, you want to be able to iterate ideas quickly. This is where AI can help us. ## Prompting for image generation is hard If you’ve ever tried to generate images with AI, you’ll know that it’s a lot more difficult to get the output you want. For most people, **it’s hard to precisely describe the visual that you have in mind** and because it takes a minute or two to generate, the feedback loop is longer. You can easily spend 30 mins of trial-and-error to get one image just right. **This is going to be a two-part guide**, which I hope will give you some practical tips to generate your own images for this purpose, without it taking more time than it’s worth. ## Separate style from content There are two core elements to any image prompt: - **The scene:** *What* is being shown in the image, e.g. a woman looks at her phone on a train - **Visual style:** *How* the image should look, e.g. a watercolour or sketch. When you’re writing a prompt, it’s important to separate these two so that it’s easy to change it when it doesn’t yield the output you’re looking for, and it’s more straightforward to re-use styles for different projects. This week we’ll look at how to prompt the scene and **next week, I’ll share a set of visual styles you can re-use** in your own work. [Make sure you’re subscribed](https://www.philmorton.co/#/portal/signup) so you don’t miss it! For all of the examples here, I’ll be using [Sora](https://sora.chatgpt.com/?ref=philmorton.co) (the dedicated image and video generation tool from OpenAI) but these tips should work on any comparable tool. ## Start with the core subject and action First, we need to be specific about: - **Who is in the scene:** age, role, appearance (if it matters) - **What they are doing:** in a non-ambiguous way > *A woman in her mid-30s is standing on a railway station platform, using a smartphone.* > > *Minimalist black-and-white grayscale illustration with clean, precise linework and smooth tonal shading. Fewer sketch lines, with a flatter, more elegant aesthetic. Rendered like a concept illustration — calm, modern, and sophisticated, with natural light and subtle depth. Not cartoonish, but polished and design-focused. No text or writing in the scene.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4q3desrev88htd564z2r4d1-1757416993_img_1.webp) 💡 In all of the examples shown, I’ve added the same visual style to the end of the prompt. I won’t show it every time for the sake of brevity. The more specific you can be, the better control you’ll have over the scene and the less you’ll leave for it to interpret: - *A person* vs. *a woman in her mid-30s* - *Is travelling to work* vs. *is standing on a railway station platform* - *Is at a station* vs. *is standing on a railway station platform* ## Set perspective and framing Images tend to have one of three frames: - **Wide:** for establishing shots and broad perspectives of the place the main character is in. - **Medium:** for showing the subject in context, capturing both the person and their immediate surroundings or props. - **Close-up:** for focusing tightly on the subject’s face, expression, or a specific detail like hands holding a phone. For most concept and storyboard images, medium shots work best because they balance clarity of action with enough environmental detail to give context to the scene. Wide shots are useful when location matters (e.g. hospital ward vs train carriage), while close-ups are good for emotional cues or product interactions. If you don’t specify this, Sora/ChatGPT/etc will make an assumption about how to frame the image. > *A woman in her mid-30s is standing on a railway station platform, using a smartphone. *Wide angle.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4q3f19cf1rsm9qhnc1fa0tq-1757417043_img_1.webp) > *A woman in her mid-30s is standing on a railway station platform, using a smartphone. *Close-up.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4q3h87wfdxa2tjekte0hagq-1757417116_img_1-2.webp) ## Control the environment Once you have your basic shot established, then think about any changes you want to make to the environment: props, architecture, etc. > *A woman in her mid-30s is standing on a railway station platform, using a smartphone. *The platform is very busy as it’s the morning commute rush hour.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4q57b4fewjvjxrdbr99s1be-1757418948_img_1-1.webp) ## Mood and lighting Then layer in weather, lighting and mood. > *A woman in her mid-30s is standing on a railway station platform, using a smartphone. The platform is very busy as it’s the morning commute rush hour. *It is raining. The platform has no roof, so she holds an umbrella with one hand and uses her phone with the other.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4qf1yq9e8m8w3n0tvv3ceqx-1757429240_img_1-1.webp) > *A woman in her mid-30s is standing on a railway station platform, using a smartphone. The platform is very busy as it’s the morning commute rush hour. It is raining. The platform has no roof, so she holds an umbrella with one hand and uses her phone with the other.* **She looks anxious (although not sad) as she needs to make sure she gets the next train.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4qfhhgafse84dm0e5n0fxg4-1757429714_img_0-1.webp) ## Interface shots You’ll likely need to show a device being used, even if you don't want to show a detailed UI. > *A close-up of a smartphone, which fills most of the image. It is held in one hand by an adult. The smartphone screen shows train times. The UI is a blocky concept wireframe with no detail and no text.* > > *In the background, there is a railway station platform. The platform is very busy as it’s the morning commute rush hour. It is raining and the platform has no roof. We only see one of the user’s hands holding the phone.* ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/ChatGPT-Image-Sep-9--2025-at-04_22_03-PM.png) ## Continuity between shots In a storyboard, you’ll likely want to show the same character doing different steps. ChatGPT/Sora isn't perfect at doing this, but it does support this. To use the same character in the next image, you can prompt/remix with something like: > A woman in her mid-30s is standing on a railway station platform, using a smartphone. The platform is very busy as it’s the morning commute rush hour. It is raining. The platform has no roof, so she holds an umbrella with one hand and uses her phone with the other. Then: > **The same woman**, but sat on the train that she was waiting for, reading a book. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4qhh7e4fvrsfqsby4vyfk7e-1757431838_img_0.webp) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4qhmqjtf2kbyx7nwnae41pq-1757431989_img_0.webp) Here it’s done a pretty good job: her face, hair, clothes and even handbag are accurately rendered across both shots. ## Guarding against weird quirks Whatever you do, you will find that the models won't always follow your instructions or they’ll produce weird and nonsensical images. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250905_1548_Relaxed-Home-Office_simple_compose_01k4d5j85ke7h8dsj62qq3fas0-1.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/20250331_1458_Reversed-Laptop-Screen_remix_01jqp7xjhme1h96ds8dskfpdbm.png) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/assets-task_01k4sf8kckeqva34m79zmvrw38-1757496527_img_1.webp) In many cases, you have to just try again or tweak the prompt, but there are some things you can do to reduce this. My top tip on this is to **only include relevant details**.The model will do its best to interpret and include everything you write in the prompt, so only describe things that you would see if you took a photo of the same scene. For example, if you say that there is *“a diabetes patient sitting on the sofa”* it is going to find some way to show that they have diabetes. In reality, diabetes patients look like everyone else when they are sat on the sofa, so you can just prompt *“an adult sitting on the sofa”* instead. ## Next week: visual styles If you made it this far, you can see why I split this into two posts! Next week, I’ll share a variety of visual styles you can copy and paste into your prompts, for use with concepts and storyboarding. ✹ Update: here’s part 2: [****A style guide for generating concept designs with AI**](https://www.philmorton.co/a-style-guide-for-generating-concept-designs-with-ai/) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### 9 more ways to get better results with AI URL: https://www.philmorton.co/9-more-ways-to-get-better-results-with-ai/ Last updated: 2025-09-17T09:50:24.000Z Readers seemed to find my previous newsletter [11 ways to get better results with AI](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/) useful, so here is a part 2. If you haven’t already read the previous newsletter, I’d suggest starting there and then coming back to this one for more in-depth and advanced tips. Let’s get into it. ## 12\. Get it to break down the problem first Small tasks such as summarising an email conversation are easy for LLMs to do in a single query, but they can struggle to solve a larger problem in one go. When “the OG prompt engineer” [Sander Schulhoff was on Lenny’s podcast](https://youtu.be/eKuFqQKYRrA?ref=philmorton.co), one of his biggest tips was to ask the LLM not to answer your query, but to break it down into sub-problems it needs to solve first. You might ask something like: ``` I need your help to [do some kind of complex task]. Please can you start by breaking this down into a set of sub-problems and sub-tasks that we need to work through to get to the outcome? ``` You would then refine the list of sub-tasks and then work through them one at a time, providing feedback and iterating along the way. There’s a secondary benefit to this for tasks you are less familiar with. As Hilary Gridley (who you should follow) writes about in [*‌you have to understand the job*](https://hils.substack.com/p/you-have-to-understand-the-job)... > *If you have a good grasp on the underlying process, the degree to which the AI can help you is amazing. If you don’t, you are likely producing a facsimile of actual good work.* Breaking down a problem into discrete parts not only gets you better results for complex tasks, but also helps you work through problems that are outside of your comfort zone. ## 13\. Have it critique itself Once you’ve gotten an LLM to solve a problem for you, you can ask it to critique itself. Then when it shares its review of its own work, you can get it to implement its own recommendations. ``` Review and critique what you've written above. Then refine your output based on your recommendations. ``` This is particularly good for more complex analysis tasks, where you need to be sure that your facts are correct. ## 14\. Only use roles for expressive tasks Another tip from Sander Schulhoff is that role prompting (e.g. *“You are an expert in medieval poetry with over 30 years of experience”*) doesn’t yield any performance boost to many queries with today’s models. It works for expressive tasks like writing or summarising, e.g. *“Rewrite this as if you are a McKinsey consultant”*. However giving the LLM a relevant role doesn’t provide any benefit for accuracy-based tasks such as categorising data. ## 15\. The more examples, the better One of the key points in [my previous newsletter](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/) was that giving the AI more input leads to better output. One important element of this input is examples. Just like when you give a task to a human, providing AI with an example of the output you’re looking for can dramatically improve the quality of the results. And the more examples you can provide, the better the results. ## 16\. Isolate your project’s memory This is a relatively new feature in ChatGPT. You can now isolate a project so that it doesn’t use the memory of your discussions outside of the project, and vice versa. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/image-5.png) ChatGPT’s memory feature makes it more useful and personalised, but sometimes you don’t want to have it do this. When a temporary chat isn’t enough, now you have this option too. ## 17\. Priming the LLM before you get it to do a task Indragie Karunaratne has this great tip in his article [*I Shipped a macOS App Built Entirely by Claude Code*](https://www.indragie.com/blog/i-shipped-a-macos-app-built-entirely-by-claude-code?ref=philmorton.co#priming-the-agent): > *There’s a process that I call “priming” the agent, where instead of having the agent jump straight to performing a task, I have it read additional context upfront to increase the chances that it will produce good outputs.* This makes sense intuitively and works on most tasks, like just if you took the same approach with a human. For example: ``` What are some best practices for analysing qualitative UX research data? ``` Wait for it to respond and then... ``` Now take a look at these transcripts I've attached and... ``` I’ve found this quite effective in many types of tasks, especially more complex ones. ## 18\. Rename your chats This is a simple one! The more you use ChatGPT (or other LLMs), the messier they get and the harder it is to find stuff that you’ve worked on previously. The names that ChatGPT gives your conversations are so short and generic, it can be really hard to distinguish between them, especially if you have multiple similar conversations in the same project. To solve this, rename the titles of useful conversations and use emojis as well. For example, I try to start all my deep research chats with 🔭 so that I can easily find them later. ## 19\. In ChatGPT, output to a table rather than CSV When you’re getting ChatGPT to do some data manipulation or analysis, it can be beneficial to get it to output to a table in the chat rather than CSV. I find that ChatGPT can be very unreliable with its CSV exporting – it doesn’t always follow instructions, data can be truncated or contain special characters that you have to then clean up. However, for smaller tables or outputs, if you get it to put the data in a table in the chat, it is much more reliable and then you can easily copy and paste the data with no issues. The only limitation is that tables in the chat cannot be very long. ## 20\. Branch your conversation in ChatGPT ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/image-3.png) Click the three dots at the bottom of any response to branch the conversation. This one is a recently added feature. You can now create a new chat from any point in a conversation, so you can take it in more than one direction from that point onwards. For example, let’s say you prompted it to analyse a spreadsheet of data. There are probably several explorations you want to do on that data, but rather than have to repeat that first step in every conversation, you can do it once and then create a branch for each one. The reason to do this rather than just have one very long conversation is that the longer the chat gets with an LLM, the worse the results get and the more likely you are to see hallucinations. So for any data analysis you’re doing, this feature is going to be a lot of use. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Miro is for the chaos. Notion is for control. URL: https://www.philmorton.co/miro-is-for-the-chaos-notion-is-for-control/ Last updated: 2025-09-10T09:50:40.000Z Most of the software we use at work has an obvious purpose: Outlook is for email and calendar, SharePoint is for storing and sharing files, Figma is for designing apps and websites. Then you have ‘collaboration software’, the point of which is harder to pin down. Tools like [Miro](http://miro.com/?ref=philmorton.co) and [Notion](https://www.notion.com/?ref=philmorton.co)... what do they do and why should we pay for them? Yet these are some of the most important tools your team needs, because they support two of the biggest tasks in knowledge work: - **Divergent thinking:** exploring a topic and making sense of a problem space. - **Organising and structuring:** bringing clarity and order to scattered knowledge and processes. The best two tools for this are Miro and Notion. ## Miro is for the chaos ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/image.png) **The start of every project is messy and uncertain.** You’ve got a lot of questions and few answers: - Who do we need to speak to? - What work has already been done? - What does the current thing look like? - What should the future state be like? - What do we need to do and in what order? When you have a group of people (who may not have worked together before) trying to figure this out quickly, you need a space to dump everything where everyone can see it. This is where infinite canvas tools like Miro, FigJam, Mural and so on are so valuable. Miro and co let you **figure out what’s going on without having to decide on structure upfront**. When you’re in the early stages of the project and are still trying to make sense of it, a Miro board lets you explore without wasting time overthinking the structure of how you’re going to capture and organise everything. Infinite canvases are also perfect for collaborative work because **they don’t force a particular style or structure on anyone**. Unlike working in a shared Word doc where different formatting styles might clash, everyone can work independently on different parts of the canvas without stepping on each other’s toes (unless people start adding stuff at completely different scales, which is super annoying). Miro boards don’t last forever. They fill up and however hard you try, they always become a mess. But that’s almost the point – they’re not meant to be permanent. They are chaotic because that’s the part of any project that they represent. Once you start figuring things out, you move on to another tool. ## Notion is for control ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/09/image-1.png) There’s a Notion template for everything Notion is the opposite. **Once you know what the shape of something is**, you can tame it by adding structure and process around it. Notion’s appeal is that it can give you control over your information, tasks or anything else in work or life. Because it’s like the Lego of software, you can adapt it to whatever project or situation you need. There are plenty of other tools designed to support specific use cases like project management, CRM and so on, but Notion allows you to create your own in exactly the way that you want. This is why there’s a huge community around Notion templates and productivity systems. The idea people are selling with these templates is that **if you add structure and process around something, you will be more successful at it**. *If only I had the right template, I can be improve in my studies/YouTube video production/work/life/etc.* Notion has a harsher learning curve than Miro, but a higher ceiling. The killer feature is its relational databases, which takes some getting used to for many people, but once you get it, you can create pretty much anything you need. ## What Notion and Miro have in common Although both of these tools serve different jobs-to-be-done, **they share a crucial trait: malleability**. You don’t have to decide how to structure something before you start because you can always change it later. Compare this to traditional tools like Word, PowerPoint, or Excel. Once you’ve chosen to make a spreadsheet versus slides, you’re locked in. You can’t easily transform content from one format to another. Miro and Notion both let you start messy and refine as you go. This flexibility reflects how knowledge work actually happens – we rarely know the right way to structure work from the beginning. ## Why you need both These tools complement each other perfectly because they mirror the natural flow of work: from chaos to order, from divergent to convergent thinking. **Miro helps you navigate the mess** at the start of a project when you're figuring out what's going on. **Notion helps you structure what you’ve learned** once you understand the problem better. You don’t necessarily need these specific tools – FigJam or Mural could substitute for Miro, and you could get away with using Confluence instead of Notion (although no relational databases for you). But you definitely need tools that serve these two purposes. Without a space for divergent thinking, your team will struggle with exploration and sense-making. Without a tool for convergent thinking, you’ll never bring structure to what you’ve learned. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why researchers should think like marketers URL: https://www.philmorton.co/why-researchers-should-think-like-marketers/ Last updated: 2025-09-03T09:50:16.000Z The ultimate goal of any UX researcher isn’t to *do research*. It’s to **generate empathy for customers in the minds of the people making decisions about the experience**. They do this by: 1. Gathering evidence about what customers need and how they use the product or service that they’re working on. 2. Synthesising all of the evidence they’ve gathered into meaningful insight. 3. Sharing what they’ve learned with people who make decisions about the customer experience. **Too many researchers focus on the first two steps** and neglect the third one. They collect valuable insights but fail to make an impact because their distribution of what they’ve learned falls short. This can leave researchers feeling that they’re not listened to, and their stakeholders wondering what their value really is. You can’t really blame researchers for this, because there’s so much involved in gathering evidence and turning it into insight, especially with the increasing time pressure we all find ourselves under. Yet if we step back, **there are some simple ways that we can increase the impact our research has** by looking at how professionals who do this third step for a living work. These people are marketers. ## Insight is the product you’re selling Think about insight as the ‘product’ that researchers ‘make’. How do most researchers ‘sell’ their insight? When most finish a project, they present some slides to their colleagues and then put the file on SharePoint where it may never be read again. It’s no surprise that some researchers struggle to deliver the impact that they should be, given the power that insight can have. What would a marketer do differently? When marketers sell a product, they go through a structured process: understand the audience, craft compelling messages, choose the right channels, run campaigns to drive action and measure the impact. ## Step 1: Define your audience Researchers excel at understanding customers, but often place less emphasis understanding their internal stakeholders. Think about how to apply the same rigour internally. Map out the different personas in your organisation: - Who is deeply invested in your work versus who barely knows about it? - What are their different roles, incentives and interests? - How do they currently use customer insight in their decision-making? **Different people need different things from your research**, just like different customer segments have different needs. ## Step 2: Shape your message Once you understand your audience, frame your insights to align with what they care about. If you’re speaking to someone focused on commercial outcomes, connect your findings to revenue impact. If you’re talking to an operations manager, highlight how the insight affects processes and efficiency. **You’re not changing the insight** – you’re highlighting different aspects of it, just as a marketer emphasises different product features for different audiences. ## Step 3: Choose your channels and formats Think about *all* of the communication channels available to you: - Email - Internal wikis - Slack/Teams channels - Weekly updates to leadership - The office walls, desks, etc. - Monthly all-hands presentations - One-to-one meetings with key stakeholders Then think about *all* of the different formats you could use: - Video - Email - Print - Slides - Custom GPTs - Memes and humour - Infographics and posters You need to match your format to your message and audience. A busy executive might need a two-minute video summary, whilst a product manager might want detailed findings they can reference later. Think beyond reports and imagine what would go viral in your company (i.e. PPTs don’t go viral, videos do). ## Step 4: Run a campaign Marketers don’t send one email and call it done. They create multiple touchpoints, repeat key messages and build momentum over time. Researchers often share their findings once – in a report and a debrief meeting – then wonder why nothing changes. Instead, **plan a series of communications** across different channels. Share key insights in multiple formats, at different times, through various moments: - Run a debrief for the core team you’re working with. - Create a 2 min summary video and share it on Slack. - Give a 5 min update in a monthly all-hands. - Etc. This might feel like overkill, but repetition is how messages stick. ## Step 5: Measure and improve Track how your communications land. You might not have access to marketing analytics through these channels, but you can still gather feedback: - Monitor reactions on Slack or in meetings - Check page views on internal documentation - Ask stakeholders directly what formats work best for them Use this feedback to iterate your approach, just as you’d iterate research methods based on what works. ## But I don’t have time for this! In an environment that pressures researchers to do things faster, **this seems like a lot of extra work** on top of an already demanding job. But remember: there’s no point producing an incredibly robust report if no-one acts on it. It’s also the job of the research team’s leader to make this easier for everyone. **Your Head of Research needs to act like the CMO of your insight**: - Setting up templates to make it easier to do. - Maintaining a list of your top insights that you can market a.k.a. your ‘product portfolio’. - Overseeing the sharing of the biggest insights you’re generating. - Making time and space for researchers to do marketing. - Opening up opportunities to share insights in different channels and with different audiences. The researchers who have the biggest impact aren’t necessarily those who conduct the most rigorous studies. They’re the ones who excel at getting their insights into their colleagues’ minds and influencing how decisions get made. **Your job isn’t finished when you’ve done the debrief.** It’s finished when those insights have generated empathy for customers in the people who matter most: the ones making decisions about the experience. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Voice dictation is the biggest productivity boost from AI that most people aren’t using URL: https://www.philmorton.co/voice-dictation-is-the-biggest-productivity-boost-from-ai-that-most-people-arent-using/ Last updated: 2025-08-27T09:50:56.000Z There’s been [a lot of commentary](https://calnewport.com/what-if-ai-doesnt-get-much-better-than-this/?ref=philmorton.co) in the last couple of weeks about AI progress slowing down. GPT-5 might not be the leap some expected and some are wondering if we’ve hit a plateau in its development. Yet while this is important for the future, most people (including myself) aren’t even close to using today’s AI to its full potential. When I talk to people about how they use AI, **the biggest limiting factor that I see is that people aren’t using their voice for input**. Data backs this up: [surveys](https://www.applause.com/press-release/applause-2024-generative-ai-survey/?ref=philmorton.co) [show](https://sqmagazine.co.uk/openai-statistics/?ref=philmorton.co) that only around a quarter of people are using voice input with LLMs. Dictation has unlocked a huge productivity boost for me and I think it can for pretty much everyone else. ## Speaking is faster than typing Like a lot of ‘knowledge workers’, writing is a big part of my job and it’s also one of the biggest bottlenecks. **Translating your ideas into words is slow**, partly because you have to choose the right words and partly because your fingers (at least mine) can’t move fast enough. Yet most people can [speak](https://tfcs.baruch.cuny.edu/speaking-rate/?ref=philmorton.co) [4x faster](https://www.lotpublications.nl/Documents/160%5Ffulltext.pdf?ref=philmorton.co) than they can type. Until recently, this didn’t really help with writing because although talking to someone (or yourself) about your ideas would help you better form your thoughts, it couldn’t write them down any faster. ## AI has made dictation actually useful The advancements in AI in the last couple of years have completely changed this. Before LLMs, voice input technology required you to dictate what you wanted to write exactly how you wanted it written. You had to be precise and speak in a way that you wouldn’t normally talk. But with AI, we have two huge upgrades to voice dictation: - New speech-to-text models are much better at accurately transcribing what we’re saying. - LLMs excel at taking unstructured input and making sense of it. **Now you can talk to ChatGPT just like you’d talk to a real person.** It can accurately understand what you are saying and it doesn’t mind if you talk to it in a conversational way. You can ramble, backtrack, and speak without altering your voice. It processes your messy thoughts and turns them into clean, structured text. This means we can finally realise the potential of being able to speak 4x faster than we can type. ## Where this works best in daily work The main caveat to this development is that AI is not (yet) good enough at replicating your personal tone-of-voice and writing style. Even if you give the best model (I would say Claude Sonnet 4) thousands of words of your writing as input, **AI is nowhere near good enough at imitating your style** well enough for the writing to pass as your own. I wouldn’t get AI to write this newsletter on its own – even if I dictated the content – because it wouldn’t sound like me at all. **Yet in most of the writing we do at work, it doesn’t need to be in our personal tone-of-voice.** Much of what we write just needs to be in a neutral, professional and factual style, not your personal style: - Emails sharing information or giving instructions - Project updates and status reports - Proposals and statements of work - Meeting summaries and documentation - Slide content and presentations AI is good enough to write this kind of content to an 80-90% quality level, with only light editing required. ## Doing the thinking vs doing the writing Just to be super clear, **what I’m talking about isn’t getting ChatGPT to think of the content *and* do the writing**. That does not lead to good quality output in most cases. Instead, the process I’m talking about is like this: 1. You use your expertise and knowledge to think what needs to be written. 2. You dictate your thoughts to ChatGPT, sometimes for several minutes. 3. It takes your input and writes it in the way that you need. 4. You give it feedback and iterate until you’re happy with it. 5. You do a final edit before putting it into whatever you are working on. This way, **the AI is acting as your ghostwriter, not doing the work for you**. This process dramatically speeds up the process of turning your thoughts into written text, especially when it’s the kind of content that you know well and are an expert in. ## Better input leads to better output A useful side-effect of talking to AI rather than typing is that you will get better output from it, regardless of the task you’re asking it to do. With LLMs like ChatGPT, the [more input you give it, the better the output will be](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/). Because speaking is so much easier and faster than typing, you naturally give the AI longer prompts, with more context and information. This leads to better responses and helps you get the most out of LLMs by using them as a collaborator rather than a tool. ## Getting started ![The ChatGPT interface, showing Dictation and Voice mode controls.](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/08/image.png) Make sure you're using dictation (a huge productivity boost), not voice mode (a novelty but not that useful). If you want to try this, **ChatGPT’s built-in voice dictation is incredibly good**. It handles brand names and natural speech patterns much better than most dictation tools (including the built-in ones on Mac and iPhone). You don’t have to clean up the transcript before you use it, like you would from a Teams call (because its speech-to-text model is terrible). Start with low-stakes writing: routine emails, team updates or little bit of slide content. The feeling of speaking to a computer is odd at first, but the speed gains quickly outweigh any awkwardness. It also helps that a lot of us are working from home. You can sit there all day, talking to your LLM, without bothering any of your colleagues! ## Dictation is for real this time Everyone knows that AI can create images, answer questions and write code. But for whatever reason, **I don’t think most people have realised that it’s also transformed how useful dictation is in everyday work.** Voice input helps you write *way* faster than before and makes you use AI in a way that gives you better outputs. If you haven’t already made this part of your day-to-day workflow, you’re missing out on one of the most practical applications of AI. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why I don’t ask people to think aloud in evaluative research URL: https://www.philmorton.co/why-i-dont-ask-people-to-think-aloud-in-evaluative-research/ Last updated: 2025-08-20T09:50:02.000Z Asking research participants to think aloud while they complete a task has been a core method in usability testing for decades. Book like Steve Krug’s *Don’t Make Me Think* and Jacob Neilsen’s *Usability Engineering* popularised it for digital product design and [the vast majority of researchers use it in their work](https://uxpajournal.org/practices-challenges-think-aloud-protocols-survey/?ref=philmorton.co). Yet the way it’s commonly taught and often used isn’t how I would recommend people use it, because it creates a number of undesirable side effects. ## The textbook method [The standard way](https://www.nngroup.com/articles/thinking-aloud-the-1-usability-tool/?ref=philmorton.co) that think aloud is taught is that **before you start your first task**, you *“ask test participants to use the system while continuously thinking out loud — that is, simply verbalizing their thoughts as they move through the user interface”*. Jacob Neilsen even [recommends showing participants a video of what is expected](https://www.nngroup.com/articles/thinking-aloud-demo-video/?ref=philmorton.co) before they start. This ensures that participants are fully briefed on what to do, but this can have a few unintended consequences... ## Some people take it too literally Some participants start exaggerating their behaviour and provide too much detail: *“I’m scrolling down the page and oh, there’s a red button. I’m going to click it because it’s red.”* This is not realistic. Asking a research participant to think aloud can put them in a mindset where they start **overanalysing everything they do**, which is not how they would behave at home. It’s like asking someone to go for a run but pay attention to how their feet are striking the ground – forefoot, mid-foot or heel. Because you’re thinking about something that you wouldn’t normally, you’re going to influence and compromise your natural behaviour. Once someone is in this mindset, trying to self-analyse their every action, **it can be really hard to pull them back** to a more natural and realistic frame of mind. This is a much bigger problem with unmoderated research, where you can’t spot and correct this behaviour, and some serial respondents habitually narrate in this exaggerated and unrealistic way. ## It’s not natural We try to make research as realistic as possible because we want to understand how individuals would interact with a product or service in real life. Yet research isn’t conducted in a very natural setting to begin with: you’re already speaking to a stranger, either on a call or sat next to them. Asking people to think aloud adds to this. It’s not something that someone would likely ever do at home and they might not have ever done it before. ## Instead, only prompt them if they are quiet I’ve conducted around 1,000-2,000 hours of research and apart from my very earliest interviews, I’ve not asked people to think aloud before they start a task. This has worked out fine and I would argue given me better results, for two reasons: - **Because you are sat there/on a call with them, they know that you are interested in what they are doing** and most will start talking on their own. Most people do not need to be briefed to alter their behaviour for your benefit – they will do it spontaneously. - **If they are quiet, you can always ask them** *“what are you thinking?”*, *“what’s going on here?”* or a similar question. It’s much better to have someone who is quiet and have to prompt them, than have someone who is being very literal and overthinking everything in a performative way. ## Try it out Having someone narrate every tiny detail of what they’re doing feels like the right thing to do, and is especially useful now that AI can analyse transcripts for us, but I really don’t think that briefing people in advance to think aloud is the best way to moderate evaluative research. When you ask someone to adopt an unnatural behaviour from the very start of a task, you risk putting them in a mindset that they would never be in at home. Instead, let them get on with it. Use your powers of observation to notice what they are doing and how they are reacting. Trust that they will say things when something unusual or interesting happens. If they’re too quiet, there are many ways to nudge them to share more without biasing them. If you’re not convinced, try it out in your next usability test and let me know how you get on. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### UX research teams should start with general-purpose AI tools URL: https://www.philmorton.co/ux-research-teams-should-start-with-general-purpose-ai-tools/ Last updated: 2025-08-06T09:50:50.000Z Every research team is scrambling to work out how to integrate AI into their ways-of-working. There’s a generational opportunity to improve methods of practice and a lot of pressure from leadership to *do something* with AI. Research teams already [spend a fifth of their budget on tooling](https://www.userinterviews.com/research-budget-report?ref=philmorton.co) and AI is surely set to increase that. But what types of tools should you invest in? ## The temptation of specialist tools There are so many AI tools for research that it makes your head spin evaluating them all, and new ones are popping up every week. When you’re exploring how AI can improve your research process, it’s tempting to get sucked in by all the specialist tools: [AI](https://outset.ai/?ref=philmorton.co) [moderation](https://www.strella.io/?ref=philmorton.co) [tools](https://www.userology.co/?ref=philmorton.co), [synthetic users](https://www.syntheticusers.com/?ref=philmorton.co) [and personas](https://askrally.com/?ref=philmorton.co), [predictive](https://attentioninsight.com/?ref=philmorton.co) [attention](https://www.hostinger.com/ai-heatmap?ref=philmorton.co) [heatmaps](https://howuku.com/free-tool/heatmap-ai?ref=philmorton.co) and so on. These are the kind of tools that get people (especially leadership) excited about AI, but there are a few things to keep in mind: - **Risk of picking the wrong tool:** AI is moving so fast that specialist tools could become obsolete within months. You could spend months on procurement, training and integration, only to find that a better solution has emerged by the time you’re fully up and running. - **Cost:** Buying multiple separate tools for specific use cases will quickly become expensive, putting even more burden on research budgets. - **Performance and quality:** These tools [don’t always deliver on what they promise](https://www.nngroup.com/articles/synthetic-users/?ref=philmorton.co), at least not yet. **Instead, research teams should start with general-purpose tools** like ChatGPT, Claude and NotebookLM. Only when these don’t cover your use cases should you invest in specialist tools. ## Why general-purpose tools are such a good fit for UX research Most research tasks are surprisingly well-suited to general-purpose AI tools because they involve text-based inputs and outputs: **Writing things...** - [Discussion guides](https://www.philmorton.co/an-ai-prompt-for-creating-discussion-guides/) and research plans - Recruitment criteria - Survey questions - Reports **Analysing data...** - Interview transcripts - Quant survey data - Analytics and MI **Researching the web...** - Competitor analysis - Social listening - Industry trends You can do all of these with a general-purpose tool like ChatGPT for ÂŁ25/month/user and there’s a good chance that your organisation might even foot the bill (rather than it coming out of your team’s research budget). Tools like ChatGPT and Claude also have huge resources behind them and improve rapidly, whereas smaller specialist tools might struggle to keep up. ## Making general-purpose tools work for your team The main challenge with this approach is that success depends on your ability to create, share and use good prompts. **You can’t just give people ChatGPT and expect great results** – there’s a lot of work to do upfront: - Identifying your most common and repetitive tasks - Writing and refining prompts for each one - Creating custom GPTs or projects with relevant context - Building a shared library that everyone can use It’s a significant investment, but probably no more than the time you’d spend on procurement and training for specialist tools. And once you create a prompt to analyse interview transcripts or write a discussion guide, that prompt will improve over time as the underlying models get better, without you having to do anything. ## Where specialist tools are still necessary There are a few tasks where general-purpose tools hit their limits when the input or output isn’t text-based: - **Live moderation** of interviews, since ChatGPT can’t moderate a video call with someone (yet). - **Video analysis** that goes beyond the (often flawed) transcript. - **Video output** like [an avatar which talks to you](https://www.synthesia.io/?ref=philmorton.co) as if they are a particular persona. - **Advanced prototyping** where something like [Lovable or Bolt](https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/) are much more capable. - **Journey mapping** when you want to [combine data from multiple sources](https://www.theydo.com/?ref=philmorton.co). ## Start with general-purpose tools Given how many use cases you can cover with a general-purpose tools, it makes sense to start with these and only move on to specialist tools once you’ve exhausted the more broadly applicable ones. Get your team permission to use tools like ChatGPT and NotebookLM, then map out your top use cases. Build prompts, [custom GPTs and shared projects](https://www.philmorton.co/how-to-create-an-ai-master-prompt-for-any-project/) for your most common tasks and see how far you can get. **Most people underestimate how much can be done with a general-purpose tool and overestimate how useful specialist tools are.** This approach gives you flexibility, keeps costs down and reduces the risk of picking the wrong tool in a rapidly changing industry. You can always add specialist tools later, but starting with the basics will give you a solid foundation and help you understand where the real gaps are. The goal isn’t to avoid specialist tools forever, but to be strategic about when and why you adopt them. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Will AI really replace junior roles? URL: https://www.philmorton.co/will-ai-really-replace-junior-roles/ Last updated: 2025-07-30T09:50:45.000Z There’s a narrative going around that AI will eliminate junior roles. The logic is that entry-level roles do simpler tasks, and simple tasks are exactly what AI excels at automating. The data appears to support this, with graduate job postings at multi-year lows. In the UK, graduate roles represent [the smallest share of all job vacancies since 2018](https://www.hiringlab.org/uk/blog/2025/06/25/uk-mid-year-labour-market-update-2025-continuing-to-steadily-cool/?ref=philmorton.co), down 12% since last year. Yet this dip in entry-level hiring is probably due to [wider economic factors](https://www.ft.com/content/87a393a2-0b5d-4a20-bee5-7f9484dbd870?ref=philmorton.co) (a macro slowdown, tariffs, older workers delaying retirement, etc.) and masks the actual impact of AI on younger workers’ careers. Contrary to the narrative, **I think that younger workers will actually benefit from AI** rather than have it replace them. ## The AI-native generation On [a recent episode of Lenny’s Podcast](https://www.youtube.com/watch?v=crMrVozp%5Fh8&ref=philmorton.co), Dan Shipper made the point that young people using AI as a default tool can skip much of traditional entry-level learning, accelerating straight into higher-level skills. > *My take is whenever I see a kid with ChatGPT, I’m like, holy shit, they're going to grow so much faster than any other person that I've worked with... I think generally people are going to figure out that some 20-year-old with ChatGPT subscription is super powerful if you just mentor them...* While generative AI has only been mainstream for a couple of years, for a small but growing group of people, it’s been there throughout their transition from education into the workforce. For this cohort and those that follow, **they’re starting their careers with the assumption that AI is part of how you get work done.** Even though a graduate entering the workforce will lack the expertise that their colleagues will have, they’ll likely be much more used to using AI. ## History repeating itself This pattern isn’t new. Younger workers have consistently led the adoption of tech in the workplace: - **Late 1800s:** Switchboard operators were predominantly teenagers and young adults who adapted quickly to the intensive work telephones required. - **Early 1900s:** Young people fresh from commercial colleges were recruited to use typewriters, proving faster learners than established staff. - **1980s-90s:** Companies with access to recent graduates adopted PCs and spreadsheets faster, as these workers arrived with the skills already learned. - **2010s:** Under-35s led smartphone and mobile app adoption in the workplace. AI is no different. [18-29-year-olds are most likely to use generative AI tools at work](www.nngroup.com/articles/ai-adoption-pew/#:~:text=AI%20adopters%20at%20work%20are%20more%20likely%20to%20be%20younger%20than%20non%2DAI%20adopters%20), treating them as something you simply pick up and use rather than a dramatic technological shift. ## Young people are also more entrepreneurial The other big difference between the generation of people entering the workforce now and those who came before is how entrepreneurial they are. The culture of side hustles, earning money online and ease of starting a business (due to Shopify, Gumroad, etc.) means that young people are exposed to business at an increasingly early age. - Nearly [25% of Americans aged 18-24 are actively starting or running a new business](https://www.gemconsortium.org/news/younger-generations-continue-starting-businesses-at-highest-rates%2C-according-to-latest-gem-usa-report?ref=philmorton.co#:~:text=The%20latest%20GEM,44%20age%20groups), the highest rate of any age group. - [63% of UK Gen Z](https://www.axa.co.uk/newsroom/media-releases/2024/axa-uk-research-shows-60-of-young-people-want-to-be-their-own-boss-by-the-age-of-30/?ref=philmorton.co#:~:text=The%20study%20found%20that%2063%20per%20cent%20of) have already “tried their hand at a side hustle or small business”. When I was at school, [Young Enterprise](https://www.young-enterprise.org.uk/?ref=philmorton.co) was the only way you were exposed to entrepreneurship. Now it’s all over YouTube, Instagram, TikTok and so on. This should create more workers who are naturally curious, self-starting and comfortable with experimentation. ## The expertise gap Of course, there's a catch: juniors lack the expertise to apply their skills, which still must be built up over time. As I've written before, [AI amplifies expertise rather than replacing it](https://www.philmorton.co/ai-wont-make-you-an-expert-but-it-will-make-experts-better/). Experienced workers are better at judging AI output and applying context that inexperienced workers might miss. The move to remote and hybrid working means it’s harder for people at the start of their careers to get ad-hoc coaching. On the other hand, the availability of YouTube, Skillshare and other platforms makes certain skills and tools (e.g. Figma) far easier to learn than before. ## Making the most of junior talent If you’re going to be hiring juniors – which you should be – then the best approach is to formally **pair them with a more experienced colleague**. This way you enable learning in a deliberate and intentional way. If you just stick juniors on projects with limited or ad-hoc support, in today’s hybrid world this will not give you the results you’re looking for. When you combine youth and experience, you get the best of both: the speed and innovation from the younger worker, with the judgement and expertise of the senior one. ## The case for hiring juniors When people are thinking about if AI is going to replace junior roles, I think they’re comparing AI to the juniors of the past. **Instead, you should be comparing AI to juniors using AI.** When you consider how well adapted to AI this new cohort of young people are going to be, and how much more entrepreneurial they are, the equation is different. You could argue that this generation of juniors is the most valuable that there’s ever been, because they can be so much more productive and more of them are already familiar with business. Rather than leaving young people behind, AI will make them more capable and more attractive to have in your organisation, not less. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why participant recruitment is the biggest risk in UX research URL: https://www.philmorton.co/why-participant-recruitment-is-the-biggest-risk-in-ux-research/ Last updated: 2025-07-23T09:50:02.000Z Everyone’s talking about [AI in research](https://www.philmorton.co/tag/ai/) right now (including me!) Research teams are under pressure to *do something* with AI. Stakeholders are asking how it will speed up research or improve insights. AI tools are emerging seemingly every day, creating both opportunities and threats. This is *the* burning platform for research teams right now. They need to be seen using AI and finding practical applications for it. But there’s another big challenge in UX research, which is perhaps under the radar of many people: participant recruitment. ## Quality participants are everything Participant recruitment is the essential ingredient in UX research. If you don’t have the right people to talk to, it doesn’t matter what technology you’re using. The insights you gain won’t be valid or will be of lesser quality. It doesn’t matter if you’re using AI to speed up analysis if the people providing the insights aren't the right ones. **All the technology and process optimisation in the world doesn’t matter if your participants aren’t suitable.** The quality of participants is the foundation upon which everything else is built. Get this wrong, and everything that follows is tainted. Yet recruitment gets little attention because it’s seen as boring, solved and unsexy. Unlike AI, which is flashy and new, participant recruitment has been around forever and doesn’t change often. Plus, most researchers outsource it to panels or agencies, so they don’t fully understand the challenges until they’re face-to-face with a participant or watching an unmoderated research recording. ## Big panels have their limits Almost all established research teams have access to a major platform like UserTesting, giving them tools and access to a panel with millions of participants. Yet when I’ve talked to research leaders, they often express their dissatisfaction with the quality of these panels, despite paying a six-figure annual fee. Large panels are great for broad consumer products and when you need feedback quickly, but they have significant limitations: **Serial respondents can undermine quality.** Many participants on large panels are serial responders. They’re there primarily to make money online, not because they match your criteria. Their motivation is financial first, fitting your requirements second. When a traditional recruiter reaches out to someone new, their primary reason for selection is that they match the criteria. The incentive is just that: an incentive to participate. But on large panels, people sign up primarily to earn money. **Niche audiences can be poorly served.** Big panels work well for broad consumer products, but struggle with niche requirements, especially B2B. If the people you need aren’t on the panel, you’re often left to bring your own participants. Some panels offer active recruitment but this is rare. ## Fraud risk is increasing As [Dr Maria Panagiotidi notes](https://uxpsychology.substack.com/p/when-research-participants-arent), *“several overlapping factors have made qualitative research more vulnerable to fraud”*. Remote research methods, accelerated by COVID and platforms like UserTesting, have reduced friction but also made deception easier. Without face-to-face interaction and ID checks, participants can more easily misrepresent themselves. **AI is making this worse.** With tools like ChatGPT, people can generate believable screener responses or interview answers without real experience. This is especially bad for quant surveys, which AI can easily answer in a plausible way. With AI moderation rolling out via various products and platforms, there’s even the prospect of AI participants talking to AI moderators. ## Tactics to safeguard your participant recruitment Don’t take recruitment for granted or assume that it’s someone else’s problem. - **Use multiple sources.** Don't rely solely on one massive panel. Have a range of options so you can pick the best method for each use case. Sometimes you need speed and quality is less critical. Other times, especially for strategic research, you want more bespoke recruitment where you can better guarantee quality. - **Tighten your screeners.** Think carefully about how to prevent fraudulent participants from misrepresenting themselves. Avoid screener questions that can be easily guessed. Assume bad faith when you’re reviewing them. - **Check how participants are verified.** How is the platform or recruiter you’re using validating that people are who they say they are? Some do this a lot more rigorously than others. For example, if you are looking for owners of a car brand, how do you know that they actually own that car? Platforms might rely on self-reporting, while traditional recruiters will ask for proof. - **Be cautious with unmoderated and quant tools.** When you’re not directly interacting with participants, it’s harder for you to detect fraud. Tools have varying levels of fraud-prevention. Make sure you understand what defences your tools have in place against fraud (e.g. detecting when someone pastes a response into a text box). - **Consider expert networks for B2B.** While these have the same issue with serial respondents, they often verify people better than mass-market panels. However, they can be expensive and restrictive about which tools and methods you can use. - **Build your own panel.** For B2B especially, intercepting your customers and getting them to sign up for research can provide better access to the right people. Tools like [Ethnio](https://ethn.io/?ref=philmorton.co) can help, although this requires significant investment. ## The future of participant research What research teams need when it comes to participant recruitment hasn’t changed: - **Quality:** the right people, who are who they say they are. - **Speed:** so researchers can keep up with the pace of product development. - **Scale:** so they don’t have to manage 10 different vendors. - **Cost effectiveness:** sensible prices that scale with demand. - **Tools:** to run research in different ways, using different methods. You can’t get all of this from one platform yet – although there are companies like [Askable](https://www.askable.com/?ref=philmorton.co) that are heading that way – so research teams inevitably need to have a number of options at their disposal. Research teams are understandably focused on everything AI at the moment, but while this is a valid thing to explore, they also need to not take their eye off the ball with participant recruitment. Unlike AI integration – which everyone's talking about – recruitment quality is often taken for granted. Yet the people you gather insight from are the foundation of good research. If your participant recruitment process is flawed, the entire design process built on top of it is undermined. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How to create an AI master prompt for any project URL: https://www.philmorton.co/how-to-create-an-ai-master-prompt-for-any-project/ Last updated: 2025-07-16T09:50:00.000Z When you’re using AI to help you with a frequent task or a project that you’ll have many conversations about, it makes sense to set up a project for it in ChatGPT/Claude/etc. These not only organise your chats into a folder, but allow you to give the LLM context and instructions that it will refer to every time you talk to it on that topic. As we know, [the more input we give the LLM, the better the output will be](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/). But giving it enough input is easier said than done. In this guide, I’ll explain step-by-step how to set up a master prompt[1](#fn1-15207) for any project. ## What is a ‘master prompt’? This is just a fancy name for a document which summarises the context that an LLM needs to know about the topic you want it to help you with. You can either create it as a standalone document (like we’ll do here) or just copy and paste all of the text into the ‘custom instructions’ field in your ChatGPT/Claude project. ## Step 1: Identify a repetitive task or project you want help with It’s not worth putting this level of effort into a one-off task. Instead we need to be applying this to recurring tasks such as: - Writing a discussion guide for research - Preparing weekly project status reports - Choosing plants for your garden The other use case is for ongoing projects where you’ll likely need to ask the LLM multiple queries, for example: - Launching a new product - Running a hiring process - Renovating your house ## Step 2: Ask the AI to generate interview questions Rather than writing a huge document ourselves, we’ll get the AI to interview us and then create it based on what we share. We’ll start by getting it to generate a list of questions to ask us: ``` I'd like you to help me [with a problem/task - the more specific you can be here, the better]. I am going to create a project in ChatGPT and your job is to write a long document encapsulating all of the relevant context, so that ChatGPT's answers are helpful and relevant to me. To do this, I would like you to start by generating a list of 10 questions that you would like to ask me, so that you can gather enough information to write this document. ``` Then we can ask it to refine its questions. This works especially well with a reasoning model like o3. ``` 1. Review your questions and share your feedback on them. Ask yourself: if I get the answers to these, will I be able to write a document that encapsulates the most important information I will need to have? 2. Refine the question list based on your feedback. ``` ## Step 3: Let the AI interview you one question at a time Now get the LLM to ask you to the questions, interview style: ``` Now ask me these questions one at a time. Once I have answered all 10, then please write an extremely thorough and detailed summary of the answers. Put this in Canvas so I can review it. There is no upper limit to the length of your output as it will serve as input for ChatGPT, not a human reader, so you do not need to be concise with your writing. Be extremely specific. ``` **I highly recommend using dictation to answer these**, as [you can talk four times faster than you can write](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/). You can also share documents where relevant. Note that I asked it to put the document in Canvas - this makes it easier to read, refine and export in ChatGPT. For example, you can download it straight to a PDF, which we’ll use later. ## Step 4: Refine the document Once you’ve answered the 10 questions and it’s given you the summary, check that it’s not missed any important context that it would need to assist you. You can also just ask it: ``` 1. Review the document you have created. Ask yourself: if I have this context, will I be able to effectively support the user in the task that they originally briefed me on? 2. Identify any critical gaps in your knowledge of the context around this task. 3. Share a prioritised list of follow-up questions that would further enhance your knowledge. ``` You can then pick any questions you’d like it to ask you, to add to what is already in the document. Keep iterating until you are happy that the document covers everything the LLM needs to know about the situation. ## Step 5: Set up your project Once you’re happy with the document that it’s produced, download it as a PDF, then create your project in ChatGPT/Claude/etc. and add it as a reference file. You can also add any other documents that you think are relevant – the more you can give it, the better. ## Let the AI build the foundations so you don’t have to This might feel like overkill, but once you try it, you’ll see just how much of an impact it makes. As with many things, **the more you put into an LLM, the more you will get out**. But rather than spending hours writing a huge document with all the context it needs, we can use AI (and voice dictation) to do the hard work for us. Try it out and let me know how you get on. --- 1. While not strictly a ‘prompt’, I have borrowed this terminology from Tiago Forte, who uses it in [videos such as this one](https://www.youtube.com/watch?v=D9DpUDntQRc&ref=philmorton.co). [↩](#fnr1-15207) ### Why and how to increase the breadth of your skills URL: https://www.philmorton.co/why-and-how-to-increase-the-breadth-of-your-skills/ Last updated: 2025-07-09T09:50:36.000Z No-one knows how the future of work is going to play out, but what people seem to agree is that: - [Expertise is valuable](https://www.philmorton.co/ai-wont-make-you-an-expert-but-it-will-make-experts-better/), since it lets you direct AI and judge its output - [Teams will get smaller](https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/), so people who can take on multiple roles will be more valuable. Most practitioners beyond the early stages of their career therefore need to focus on expanding their skillset to be more of a generalist. This is something **worth doing right now**, not just in anticipation of the future. The more skills you have, the more ways you can be useful to the people you work with. The more of the design and development process you can cover, the more indispensable you become. ## The five components of learning a new skill The fastest way to learn something new is to incorporate all of the following: 1. **Understand the theory**: use books, training courses, conferences, articles, podcasts, YouTube videos, ask an LLM, talk to your colleagues and so on. 2. **Observe how others work**: shadow your colleagues, look at their work-in-progress, ask them to talk you through what they’re doing. 3. **Practice the skill**: get or create opportunities to try it out yourself. Use mock exercises if it’s hard to get the opportunity to do this in real project work. 4. **Analyse your performance:** reflect on what you’ve done, get feedback from other people who already have the skill. 5. **Get coaching** from people who can do it and build relationships with people who can help you. ## Start with adjacent skills, not distant ones Learning new skills takes time and effort, which is why it can feel daunting. This is especially the case if you try to learn something that is far away from your comfort zone. Think about the five ingredients in the list above: - It’s much easier to observe how others work if they work around you every day. - It’s easier to get coaching if you already work with people who have that skill. - It’s easier to practice a skill if the work you’re already doing is related. If you want to broaden your skills, this is why **it makes most sense to start with adjacent skills**, not distant ones. Build outward from your existing knowledge rather than starting from scratch in a completely new field. If you’re a researcher, UX design is a natural next step because you already have expertise in what people find usable or not. If you're a content designer, AI prompting is essentially just another form of writing. Ask yourself: *what’s the one skill, that if I had it, would allow me to do a different role?* ## Finding your gaps If you find that question hard to answer, here are a few ways to reveal the next skill to learn: - **Use AI to interview you:** upload your CV and portfolio, then get AI to ask you 10 questions to help it identify what you should learn next. You could even feed it job descriptions of adjacent roles or other people’s profiles. Remember, [the more you give it, the better the suggestions will be](https://www.philmorton.co/11-ways-to-get-better-results-with-ai/). - **Ask the person who resources your team** about near-misses – the times when you were considered for a project or role but didn’t get it. What specific skills were you missing or perceived to have missing? - **Look at people that you admire:** what skills do they have that you don’t? You could ask them directly what they think you should add to your skillset next. - **Get feedback from those you work with:** ask them what skills, if you had them, would have made your contribution to the team even more valuable. ## You’re probably overestimating the effort it takes to get started Even if you know what you should learn next, it’s another thing to make progress towards it. Too often when we think about learning a new skill, we think of the time it will take to master it, rather than how long it will take to get to a basic level of competency. We see people who are experts and think of the many years they’ve had to get there – it seems impossible that we could do the same. But **you’re not trying to master each new skill**. That’s not what being a generalist is about. Learning new skills has never been easier, with all the resources available online (plus AI). You’d be surprised how far you could get in a few hours or days. Recently I was working on a project where the need shifted from research and strategy to executing design. The only thing stopping me from contributing was my limited Figma knowledge. I spent a few hours learning auto layout by watching YouTube videos, and suddenly I could participate much more in the process. **Adjust your expectations** and new skills seem much more attainable: - You can’t learn design in a day, but you can learn the basics of Figma in a day. - You can’t learn UX research in a day, but you can learn how to write a discussion guide in a day. - You can’t learn how to code in a day, but you can learn the basic syntax of HTML in a day. Sometimes we have several days or even whole weeks between projects – imagine what you could learn then! ## Motivation is everything If you can pick up skills so quickly, why aren’t we doing it all the time? The reality is that you need motivation to learn new things. It takes curiosity and energy, which maybe we don’t have at the end of a tough work day after we’ve put the kids to put to bed. Many people go whole years of their career without really trying to learn anything new. Sure, they pick up things in the work, but they spend almost no time deliberately learning new skills unless they’re sent on training courses. I am guilty of this in many years of my career – sometimes you just don’t have the energy or inclination to step outside your comfort zone and take on something new. What you need to do is **find a clear ‘why’ for the thing you’re learning**. If you are learning a skill because someone told you you to or because you’re afraid that AI might take your job, you are not going to follow through with it. These are extrinsic motivations. What you need is an intrinsic motivation: a reason that genuinely excites you. Something that sparks your curiosity, that connects to the kind of work you want to be doing or the person you want to become. ## Learning happens through hard work, not by magic Even with a clear goal and motivation, learning new skills takes time and energy. If you are deliberate about learning, you can develop your skills much faster than if you just come to work and get the job done with as little effort as possible. Think about two athletes: one who trains by running laps all day without thinking vs. one who is tracking their times, analysing their gait, employing a coach and thinking about how they are training. Which athlete is going to progress faster? **Think of every project as an opportunity to learn or practice a new skill.** Have a plan for what you're going to work on. Keep a learning log or tracker. Schedule dedicated time each week for skill development. Use habit systems to make it stick. Most importantly, be proactive about creating opportunities rather than waiting for them to appear. ## Step out of your comfort zone and make a start Broadening your skills is a logical and calculated response to the trends in our industry, but fear of an uncertain future is not the best motivation to start learning new skills. A better reason is that the more skills you have, the more interesting and varied your work will become. Learning keeps you engaged and prevents the boredom that comes from doing the same thing repeatedly. Plus it makes sense for your career. The present and the future belong to people who can work across boundaries, not just within them. The question isn’t whether you should broaden your skills – it’s which skill you’re going to start with next. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The Situational Leadership model URL: https://www.philmorton.co/the-situational-leadership-model/ Last updated: 2025-07-02T09:50:25.000Z About a year ago, I was working with someone who seemed frustrated. I couldn’t figure out why, until I realised I was coaching them when they just wanted support. I was being helpful, but too hands-on. This kind of misalignment is easy to fall into and more common than you might imagine. Most teams (whether a one-off project or a longstanding squad) have a implicit or explicit hierarchy. One person is responsible for overseeing and directing, and one or more people are responsible for executing. Yet it’s easy to skip over *how* this relationship is going to work in practice. ## The Situational Leadership model ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/06/266f273d162b591f0dc4f4b4e2bb6de9.png) Developed by Paul Hersey and Ken Blanchard, the model describes how two people might work together when there is a hierarchy. As you can see, there are two axis: - Along the bottom, how much direction the leader provides. - Top to bottom, how much support the leader provides. Working through the four quadrants
 - **Directing** (high direction, low support) is for when someone’s new to a task and needs clear, step-by-step instructions to get started. The leader decides what needs to be done and how to do it. Some people would see this behaviour as micro-management and recoil at the lack of autonomy, but if someone is unfamiliar with the task then it may make sense to use. - **Coaching** (high direction, high support) is best for when the person doing the task is relatively inexperienced and wants additional direction and support so that they can learn. You’re guiding the work closely and providing encouragement, asking questions, and explaining the “why” behind decisions. This is probably the most common area that people accidentally end up in, with leaders giving too much direction when it’s not wanted. - **Supporting** (low direction, high support) is for when someone has the skill but might need encouragement or a sounding board. They’re given the autonomy to figure out how to achieve the goal, but if they need help, it’s there. This is the sweet spot for many people. - **Delegating** (low direction, low support) is for when you completely trust someone to get on with things and you don’t expect them to need much direction or support. You share the goal with them and then let them take the lead, knowing that they’ll reach out if they need any help. ## How to apply it I find it best to use this model at the start of a project: - Explain the model if someone isn’t already familiar with it. - Ask them which quadrant they’d like you (the leader) to be in, to best support them. - Check in mid-project to make sure that what you’ve chosen is still the best one to be in. You don’t necessarily need to use the model formally. Just thinking carefully about how you support someone and checking that it matches what they need can make all the difference. A quick conversation upfront often saves a lot of confusion later. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### 11 ways to get better results with AI URL: https://www.philmorton.co/11-ways-to-get-better-results-with-ai/ Last updated: 2025-08-31T13:46:27.000Z One of the most compelling things about AI is how much value you can get from it, for such little effort. With just a few words, even someone with no experience can get personalised answers, write code, create images, analyse data and so on. This is why ChatGPT has 800 million users and counting. ## Drawing the wrong conclusions about AI skills Because it seems so easy to do so much, the conversation around the impact of AI tends to focus on which human skills will become obsolete and whose job will be replaced. As [a few](https://www.forbes.com/sites/sergeirevzin/2025/06/18/why-duolingos-ceo-reversed-his-ai-stance-and-what-it-signals-for-work/?ref=philmorton.co) [companies](https://tech.co/news/klarna-reverses-ai-overhaul?ref=philmorton.co) have found out recently, it’s not as straightforward as that. AI [makes experts better more than it makes everyone an expert](https://www.philmorton.co/ai-wont-make-you-an-expert-but-it-will-make-experts-better/). What we should be talking about instead is **what *new* skills humans will need** in a future with AI. ## Just because it’s easy doesn’t mean there isn’t depth to master The more you use AI, the more it becomes apparent that using it effectively is a distinct skill that everyone needs to learn. This is the realisation most people have the first time they see a 1,000 word prompt – there’s a lot more to it than asking for basic things like *“analyse this PDF”*. **Getting the most out of tools like ChatGPT and Claude** (let alone workflows and agents) requires: - Creative thinking to imagine a use case in which AI can be used. - Knowledge of the tools relevant to the task at hand. - The skill to instruct it in a clear and structured way, to get the results you need. - The discipline to apply critical thinking and your judgment, so you don’t always take the easiest and fastest route to an answer. The more you use it for actual work – complex tasks with a high quality bar – the more you need to develop these skills. ## 11 essential skills for collaborating with AI Like everyone else, I’ve been using AI more and more, trying to learn as much as possible through trial and error. Here’s what I’ve learned recently – I hope you find it useful. ### 1\. Treat AI like a teammate, not a tool This is the most important idea in this list. [Research by Jeremy Utley (Stanford d.school) and Kian Gohar (GeoLab)](https://howtofixit.ai/?ref=philmorton.co) has shown that **people who treat AI as a teammate get better results** than those who see it as a tool. > *Shifting your orientation from tool to teammate changes everything about the kinds of outcomes that you can achieve working with generative AI.* When AI gives you mediocre results, if you think about it as a tool then you may presume that it’s the tool’s fault and that it is limited. But **if you treat it as a teammate, then you try to help it to get to the right outcome** by providing it context, feedback and coaching. This mindset encourages you to give the AI more input, speaking of which... ### 2\. More input = better output Just like when you give a person a task to do, the more context, direction and guidelines you provide, the closer the outcome will be to what you had in mind. [In this video](https://www.youtube.com/watch?v=pUG7%5F03G%5FAc&ref=philmorton.co), Vicky Zhao explains the difference between: - People who input 10 words to get 1,000 words of output - People who input 1,000 words to get 1,000 words of output ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/06/Screenshot-2025-06-14-at-11.21.11.png) She highlights how **people over-index on how much *faster* they can do things with AI, rather than how much *better* they can do them**. If you focus too much on saving time, then you may skip the steps required to get a quality output. ### 3\. Talk, don’t type If giving more input is what’s required, then won’t that take a lot of time and negate the benefits of AI? Not if you use your voice. **Most people can speak 4x faster than they can type**[1](#fn1-8597) and dictation software has come a long way recently. I find [ChatGPT’s dictation feature](https://openai.com/index/whisper/?ref=philmorton.co) to be very accurate, whereas the Mac/iOS system dictation requires a lot of correction. You can also get standalone apps to give you better system-wide diction. Using your voice also has a couple of other benefits to how you provide input: - **It reinforces the idea of treating AI as a collaborator** rather than a tool. - **It frees you from having to find the perfect wording** to express your ideas. Just talk to it how you would talk to a person – AI is great at taking your unstructured thoughts and making sense of them. This is an unexpected benefit of working from home: the freedom to talk to AI without being self-conscious and annoying your colleagues! ### 4\. Ask the AI to write the prompt If you’re struggling to articulate what you want the AI to do for you or need an especially large prompt, then get ChatGPT to write the prompt for you. You can ask something like: > *You are one of the world’s foremost experts in writing prompts for ChatGPT. I’d like you to draft a prompt to get it to...* Again, the more context you can give it, the better the result, so talking is better than typing in this case. I find this especially useful for deep research requests, [which work best with a detailed brief](https://www.philmorton.co/getting-the-most-out-of-chatgpt-deep-research/). ### 5\. If you don’t want it to do something, explain why When people want the AI to avoid certain behaviour, they usually write their instructions in a super-direct way: “Do not use title case for headings”. However, I recently learned that this is not the best practice. Anthropic (makers of Claude) outline a better approach [in their documentation](https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/claude-4-best-practices?ref=philmorton.co#example-formatting-preferences). **Explaining *why* the behaviour is undesirable** leads to better instruction following: > **Less effective:** > *NEVER use ellipses* > > **More effective:** > *Your response will be read aloud by a text-to-speech engine, so never use ellipses since the text-to-speech engine will not know how to pronounce them.* ### 6\. When you get a poor response, edit the prompt first, rather than provide feedback I picked up this tip from [an episode of *How I AI*](https://www.youtube.com/watch?v=5Byg-9K8JnM&ref=philmorton.co) with Luke Harries. If you don’t get the result you want from AI, rather than replying and giving it feedback to iterate, **edit the original prompt**. This is especially valuable for any prompts or tasks you might do again in the future. If you default to editing the prompt, then over time it will become more and more refined, and give you more consistent results in the future. ### 7\. Experiment with different models As more and more models have been released over the last few years months, what’s apparent is that **the differences between them are growing**. Each has their strengths and weaknesses, things they’re good for and things that they’re less capable of. Simon Willison has a brilliant demonstration of this [in his article](https://simonwillison.net/2025/Jun/6/six-months-in-llms/?ref=philmorton.co) where he shows how well each model can draw an SVG of a pelican riding a bicycle: ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/06/ai-worlds-fair-2025-25.jpeg) Even comparing OpenAI’s models, you can see a big difference: 4o is great for most tasks and decent at writing, but o3 is much better for complex tasks and web search. ### 8\. Let AI question you As Stanford’s [Jeremy Utley explains](https://www.youtube.com/watch?v=wv779vmyPVY&ref=philmorton.co), the design of AI chat interfaces emphasises you asking it questions, but there’s actually a lot of value from using it the other way round: > *The fundamental orientation a lot of people take towards AI is: I'm the question asker; AI is the answer giver. But if you think about AI like a teammate, you say, "Hey, what are ten questions I should ask about this?" or, "What do you need to know from me in order to get the best response?"* When you have a draft of something you’re working on, try giving it to the AI and prompting something like *“Ask me 10 questions about this to expand my thinking”*. ### 9\. Create roles, not just prompts The best teams are made from people with different strengths and specialisms. So if you’re thinking about AI as a teammate rather than a tool, then try thinking about each custom GPT/project you create as a distinct team member with its own role. For example, if you’re using AI to help you write slides, then your ‘team’ of custom GPTs might look like this: - Brainstorming/idea generation assistant - Presentation outliner - Writer - Editor - Graphic designer When you think about your prompts and custom GPTs having **specific and narrow roles** (rather than trying to wrap up everything into a single one), then each can be tailored with its own instructions and examples. This modular setup makes it easier to maintain quality and develop reusable workflows for repeated tasks. ### 10\. Chain your prompts together Larger and more complex tasks are best broken down into smaller ones, which can be passed from one prompt to the next like a production line. Using the example above, you can develop your idea with a GPT whose role is to expand your thinking and brainstorm new ideas. Then you give those notes to a GPT that outlines the slides and works out what should be on each. Once you’re happy with that, you can pass the outline to a GPT that writes a first draft for you. When you’ve refined it, then you pass the draft to the editor GPT for feedback. Finally, the graphic designer GPT can create relevant images for your presentation. Using this approach allows you to **modularise your workflow**, making it easier to iterate to the quality level you need, compared to a single GPT that tries to create the whole presentation for you. ### 11\. Expect human involvement It’s so tempting to think that AI can automate away processes wholesale, but the reality is that delegating entire tasks rarely results in the quality you need. When you treat **AI as a teammate rather than a tool**, then you can essentially hire a team of specialists that you can call upon to help you with any task. Start with your perspective, ideas and expertise. End with your own words or expression of those. The messy process in the middle is where AI can add the most value. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). --- 1. See [https://tfcs.baruch.cuny.edu/speaking-rate/](https://tfcs.baruch.cuny.edu/speaking-rate/?ref=philmorton.co) and [https://www.lotpublications.nl/Documents/160\_fulltext.pdf](https://www.lotpublications.nl/Documents/160%5Ffulltext.pdf?ref=philmorton.co) [↩](#fnr1-8597) ### Yes, working through bureaucracy is your job URL: https://www.philmorton.co/yes-working-through-bureaucracy-is-your-job/ Last updated: 2025-06-18T09:50:41.000Z Ask any design leader what the least favourite part of their job is, and chances are that it’ll be getting new vendors onboarded, processing invoices, raising purchase orders and so on. In many large organisations, these tasks can feel like **an endless maze of bureaucracy**. There are huge email chains, outdated systems and processes that no-one seems to have written down. It feels like these tasks take up too much time and get in the way of doing the ‘real work’. The thing is, **working with your organisation's bureaucracy is part of the job.** Rather than complaining about it, we need to develop tactics for working through it effectively. The good news is that as UX professionals, we already have many of the skills we need to get through these internal processes with as little stress as possible. ## Why these tasks feel so painful Bureaucracy is particularly frustrating for people in product, design and research roles because we see the world through the lens of user experience. We have a high bar when it comes to interacting with organisations – including our own – and **we’re able to spot every tiny thing that’s wrong with an experience.** We’re especially sensitive to bad user experiences, and most internal processes have terrible UX. We’re also used to structured work environments where work is closely managed. In the software development process we have backlogs, prioritisation and systems to organise who’s working on what. When we ask another department to do something, we can fall into the trap of assuming that they operate the same way. In reality, many departments don't work like this. Finance, legal and HR may just have a shared inbox. Tasks aren’t managed and prioritised in the same structured way as you would find in Jira. IT is usually the exception because they use ticketing systems, but most other departments don't. ## Five ways to use your UX skills to get through the bureaucracy Instead of getting frustrated, we can repurpose our professional skills to make bureaucracy more manageable. ### Use empathy and curiosity Rather than getting irritated when things don’t move as quickly as you’d like, **use your research skills to try to understand the process** and the challenges the people you're working with face. How does their team actually work? What pressures are they under? A bit of curiosity and empathy can go a long way. ### Be the project manager Don’t just send an email to finance or legal and expect them to manage the entire process for you. If you want to move something forward, you may need to be the project manager of that task. If nothing seems to be happening, **send weekly status updates to everyone involved, summarising where things stand.** This is incredibly helpful because people often can't remember what a long email chain is about and they have 50 other similar requests to deal with. You can use AI to summarise the one or multiple threads about a task, which is especially useful when new people get added. Being the person who keeps people aligned and on track can move things forward. ### Visualise the process Journey mapping isn’t just for ‘the real work’ – it’s valuable for internal processes too. Often, the process hasn't been visualised by the team themselves, so it’s actually quite a helpful artefact for them. When you draw a diagram of your understanding, it’s much easier for stakeholders to spot gaps and quickly tell you how things really work. Visualising something helps **expose your assumptions about how the process works**. I recently had to agree a new approval process with our finance team. Instead of back-and-forth emails about which documents would be required, I created a mock email in Miro with example text and little icons for each attachment, including screenshots of what each item would look like. Visualising it **helped everyone understand how the process would work** and we got agreement on it much more quickly. We often forget to apply our design skills when communicating with people internally who aren’t usually our stakeholders in the design process. ### Don't wait for permission You might be worried about stepping on people’s toes or disrupting processes, but as long as you do things in a helpful and positive way, using your skills can speed things up and make the whole experience more pleasant for everyone working on it. **You don’t need to ask permission to draw a diagram of how something works**, and you don’t need permission to be the one who summarises the email chain. Your role in whatever task you are doing can be emergent: the ‘project manager’ is whoever is *doing* the project management, just as the ‘leader’ is whoever is *doing* the leading. Obviously, if you want to change how a process works, you need buy-in from people. But if you’re just working through a process and managing a task that you care about, you generally have the freedom to push that task forward using your skills in any reasonable way. ### Document the pain for others If you have to go through a painful bureaucratic process, documenting it can help those who come after you. I once had to navigate a laborious 30+ step process to get set up on a client’s IT systems. The instructions from their IT department weren't comprehensive (to say the least), so I documented the whole process in Notion with screenshots. My colleagues were then able to go through the same process much faster than I did. **You don’t have to change the process to make it easier to work through** – just documenting it for others is enough. ## It’s part of the job As UX professionals, we love to improve the way things work, but sometimes you just have to accept a cumbersome process and get through it the best you can. **If you tried to improve every process you encountered, you’d never get your day job done.** More often than not, you can’t change the process, but you can work through it in the best way possible for yourself and the people you’re working with. The skills we use every day – making sense of things, visualisation, documentation, project management, empathy, and problem-solving – are incredibly valuable for navigating internal bureaucracy. We just need to remember to apply them beyond our immediate work and use them to make our own – and our colleagues’ – lives easier. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why everyone’s talking about ‘taste’ as the next differentiator URL: https://www.philmorton.co/why-everyones-talking-about-taste-as-the-next-differentiator/ Last updated: 2025-06-11T09:50:23.000Z There's a word that keeps cropping up in Silicon Valley commentary lately: ‘taste’. You hear it on podcasts, read it in articles and see it on LinkedIn. The idea is that **if anyone can use AI to create software, then your differentiator isn't technical skills, it's taste.** [Anu Atluru](https://www.workingtheorys.com/p/taste-is-eating-silicon-valley?ref=philmorton.co): > *"In a world of scarcity, we treasure tools. In a world of abundance, we treasure taste. Everyone's software is good enough. Software used to be the weapon, now it's just a tool... Taste is the new weapon."* [R. J. Abbott](https://www.linkedin.com/posts/rjabbott3%5Fin-the-age-of-ai-technical-founders-are-activity-7329178728847081473-sq5K?utm%5Fsource=share&utm%5Fmedium=member%5Fdesktop&rcm=ACoAAAFYpt8BfI3cpdTynsXUzSwr%5F1zpnbr5Bog): > *"In the Age of AI, Technical Founders Are Overrated. Taste Is the New Tech Stack... AI has turned the 40-hour work week into 40-minute tasks. In an era where code is no longer scarce, by definition, technical skills are no longer special."* [Krithika Shankarraman, former VP of marketing at OpenAI, on Lenny's Podcast](https://youtu.be/QaDsk4iH1aw?si=AdOxk1MpvkMQnosY&utm%5Fsource=ZTQxO): > *"Taste is going to become a distinguishing factor in the age of AI because there's going to be so much drivel that is generated by AI... truly, the companies that are going to distinguish themselves \[are the ones that show their craft\]."* ## For each trend, there is an equal and opposite trend Now that the barrier to creating software has been lowered dramatically, anyone with enough determination can start a software business in a weekend, even without coding skills. But what AI doesn’t have is ‘taste’. It doesn't have enough context or judgment to determine what quality looks like. This is a trait that only humans have. If anyone can create software, then what's valuable isn't who can build the thing anymore, but who can **build the right thing, for the right people, in the right way.** ## Taste is expertise in disguise The way taste gets talked about is on a spectrum. On one end, it's portrayed as rare genius that only a select few possess, like having an eye for design that can't be taught. On the other end, it's framed as a learnable skill you can develop through experience and practice. I think it's the latter: **taste is just expertise in disguise.** When you make decisions without data – which you do constantly – you're drawing on everything you've learned, seen and understood about people. That accumulated knowledge helps you predict what others will find engaging, useful or beautiful. Every time you write an email, make a design decision or choose how to present information, you're using your taste. You're using what you've seen work before to anticipate what will work in the future. ## Opinion vs. data It’s easy to recoil at the trend of putting subjective opinions on a pedestal. After a decade of being encouraged to make decisions based on data and evidence, now we’re meant to abandon that and use our intuition? Some of the biggest product failures have come from leaders who believed their own genius, which is why we all groan about HiPPOs (the highest paid person's opinion). It’s a fair critique: **we shouldn't abandon evidence-based decision-making** in favour of pure intuition. But we also can't dismiss the value of accumulated expertise. The best outcomes come from blending both approaches: - **Use data and research** to understand user needs and validate assumptions. - **Use your expertise** to make the thousands of micro-decisions that data can't answer. ## Who benefits from the taste trend The focus on taste is increasing the perceived value of those who understand users and how to design for them. The [$6.4 billion OpenAI is paying for Jony Ive's agency, LoveFrom](https://www.siliconrepublic.com/business/openai-io-sam-altman-jony-ive?ref=philmorton.co), which employs around 50 people, is one data point for this. Another thing I hope comes from designers' intuition being valued more is that they get more permission to be creative and take risks. The [recent Airbnb redesign](https://news.airbnb.com/product-releases/airbnb-2025-summer-release/?ref=philmorton.co) hints at this shift. Even though it’s not revolutionary, the skeuomorphic icons and motion design have garnered a lot of attention. If you look at software from 20-30 years ago – think about [WinAmp](https://www.theverge.com/tldr/21430347/winamp-skin-museum-nostalgia-90s-00s-internet-art-history-ui?ref=philmorton.co) or early skeuomorphic design in iOS – there was so much more playfulness. Today's interfaces often feel sterile and lifeless. ## Finding a balance between instinct and insight The rise of taste as a differentiator shouldn't be about abandoning rigorous product development. It should be about recognising that in a world where the technical barriers are lower, human judgement and expertise become more important. **Good taste isn't magic – it's pattern recognition developed through years of experience.** The people who will thrive are those who can combine their accumulated expertise with solid research and data. After all, taste without validation is just opinion. But data without human interpretation is just noise. The future belongs to people that can do both. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### An AI prompt for creating discussion guides URL: https://www.philmorton.co/an-ai-prompt-for-creating-discussion-guides/ Last updated: 2025-06-04T09:50:12.000Z If you’re thinking about how you can integrate AI into your research process, writing a first draft of a discussion guide is surely one of the most obvious places to start. - Writing a discussion guide usually takes several hours. - They follow a standard structure and set of guidelines. - AI is good at taking unstructured input and making sense of it. - Reasoning models can reliably follow step-by-step instructions, review their own work, and take the time they need to create quality output. I've been experimenting with this and below is a prompt that I've developed, based on the process that a human researcher would follow to create a discussion guide. ## The prompt 💡 This has been designed for and tested with the o3 model in ChatGPT. It probably won't work as well in a non-reasoning model (e.g. 4o). ``` Role: Act as a UX researcher with over 20 years of experience conducting qualitative customer research. Your job is to understand the research project the user is working on and then write a discussion guide for them based on this. Task: 1. Gather the following information from the user. Do not proceed until you have understood all of the following: 1. Project context - What is the product/service? - What stage is it at (early idea, prototype, live product)? 2. Research goals - What are you hoping to learn? - Are there key decisions this research should inform? 3. Type of research - Formative or evaluative? - If evaluative, what are you testing (e.g. prototype, live site, content)? 4. Audience - Who will be taking part in the research? - Are all the participants part of the same segment or are there any significant differences? 5. Constraints or preferences - Time per interview/session - Particular areas to avoid or focus on 2. Once enough information is gathered, create a structured draft discussion guide by following these steps: 1. First, define the business objectives and research objectives for the study. 2. Then create an interview structure to achieve these objectives in the time allocated for the session. Typically a discussion guide has the following sections: 1. Title e.g. "Discussion Guide — [Project name]" 2. Objectives - the business and research objectives 3. About this document - A short note explaining that this is a flexible guide, not a script 4. Introduction - Introduce self - Explain that today we’re looking at [the subject of the interview] - Explain format & timing of the interview - Explain session confidentiality (any personal details will not be passed on or used outside of this testing - The interview will be recorded (obtain permission) - Explain focus is on understanding individual’s thoughts, opinions and experiences, and that there are no right or wrong answers. Reassure respondents that this is not a test of them and that we are just interested in gathering different opinions - Start recording 5. Context of use - Set participants at ease with a few easy questions; personal/job background; current behaviours, needs, or workarounds; triggers and motivations. 6. Core topics, tasks and reflections - Formative: Scenarios, attitudes, pain points, expectations - Evaluative: Task-based walkthroughs, usability probes, comprehension checks - Encourage free play when relevant 7. Final reflections and questions 3. Then once you have a structure for the discussion guide, create a set of 1-3 objectives for each section. This outlines the purpose of that part of the interview. 4. Review the structure of the discussion guide so far and objectives for each section. Compare these to the overall business and research objectives to ensure that your structure will deliver what is needed. Refine your structure, headings and objectives as required. 5. Write the questions for each section, to meet the objectives of that section. 6. Review the entire discussion guide to ensure it meets the objectives of the study and flows well. Make any final refinements before sharing the output with the user. Guidelines: - A discussion guide should have approximately 0.9 questions per minute. - Start with broad, open-ended questions to encourage rich, unbiased responses - Use everyday language that participants will easily understand, avoiding jargon or technical terms - Focus on behaviours and experiences rather than opinions or hypotheticals - Ask one thing at a time to avoid confusing or leading participants - Keep questions neutral and non-judgemental to avoid influencing responses - Use follow-up probes to explore details, motivations and context - Ensure questions are relevant to the participant’s role, knowledge and usage - Phrase questions to invite stories or examples rather than yes/no answers - Avoid asking about future intentions unless you're exploring unmet needs - Tailor your questions to the stage of the session, building from easy to more reflective topics - Format each section from “Context of use” onwards like this: ## Section name Objectives: - [objectives] Questions: 1. [questions] ``` Paste the above into your favourite reasoning model and it should respond with something like this: ``` Absolutely, Phil. To put together a strong discussion guide I’ll need a bit more context from you first. Could you please tell me: 1. **Project context** * What is the product or service? * Which stage are you at (early idea, prototype, live product, or something else)? 2. **Research goals** * What are you hoping to learn from this study? * Are there specific decisions that the findings should inform? 3. **Type of research** * Is the study formative (exploring needs and concepts) or evaluative (testing something that exists)? * If evaluative, what exactly will we put in front of participants (e.g. interactive prototype, live site, content)? 4. **Audience** * Who will take part? * Are all participants drawn from the same segment or do we need to cater for different groups? 5. **Constraints or preferences** * How long do you expect each session to last? * Any areas you want to avoid or focus on in particular? Once I have this information, I can draft the discussion guide for you. ``` Then provide it with input covering the areas it's asked for. I recommend using voice for this, since we can speak four times faster than we can type, on average. If you are using AI in a secure environment, you could also share files and other materials to provide additional context. ## Example input Here's an example you can use to test the prompt: ``` I'm working on the Athlete Intelligence feature in the Strava app – it’s a new AI tool that analyses your workout data and gives you generative AI feedback on your runs and rides. It’s already live, so this is evaluative research, and we’re trying to understand whether people find it valuable, what’s working, what’s not, and what else they might want AI to do in the app. We’re speaking to 12 Strava subscribers – mainly runners and cyclists – and all of them fall into the same general segment, so no need to treat them differently. The research is pretty open, but we’ll have the Strava app in front of us and use it to prompt conversation based on how they’ve interacted with the feature. Each remote session will run for 60 minutes. Our main goal is to get a clearer picture of the impact this feature has had so far and gather ideas about what to add, change or potentially remove in future updates. Avoid discussions about subscription pricing as we don’t have influence over that. ``` ## Example output With the above input, o3 returns the following: 💡 [View the whole conversation on ChatGPT](https://chatgpt.com/share/683b29c1-40d0-8000-b62c-c306ae3f9dd2?ref=philmorton.co) for easier-to-read formatting. ``` ## Discussion Guide: Strava Athlete Intelligence – Evaluative Customer Interviews --- ### 1. Objectives **Business objectives** * Improve the live Athlete Intelligence feature so it drives subscriber retention and positive word of mouth * Identify quick-win enhancements and longer-term opportunities for AI-powered coaching in Strava * De-risk future roadmap decisions by learning what to add, change or remove **Research objectives** * Gauge the perceived value and real-world impact of Athlete Intelligence on training behaviour and motivation * Pinpoint usability or comprehension issues that hinder uptake or satisfaction * Discover unmet needs and feature ideas users would welcome * Understand how feedback generated by the AI influences trust and continued engagement --- ### 2. About this document This guide is a flexible framework rather than a rigid script. Follow the participant’s language and probe naturally. Aim for roughly 54 questions across the 60-minute session (about 0.9 questions per minute), adjusting on the fly if a topic proves especially rich or thin. --- ### 3. Session outline | Section | Time | Objectives | | ---------------------------------- | ------ | ---------------------------------------------------------------------------------------- | | A. Introduction & housekeeping | 5 min | Set context, secure consent, build rapport | | B. Warm up & context of use | 10 min | Understand athletic background, current Strava habits, discovery of Athlete Intelligence | | C. Current experience | 10 min | Explore perceived value, memorable moments, pain points | | D. Walkthrough & task-based probes | 25 min | Observe real interaction, uncover usability hurdles, capture in-the-moment reflections | | E. Future needs & ideation | 7 min | Gather ideas for improvements, new use cases, AI boundaries | | F. Final reflections & close | 3 min | Surface overall sentiment, open floor for extra comments, end on positive note | --- ### 4. Detailed guide #### A. Introduction & housekeeping *Objectives* * Explain purpose and format * Obtain recording permission and assure confidentiality * Reinforce that we are testing the feature, not the participant *Questions / talking points* 1. Quick self-intro of moderator and role 2. Explain we will talk about Strava and a specific AI feature the participant already uses 3. Confirm 60-minute timing and remote setup is still convenient 4. Explain recording – ask for verbal consent 5. Emphasise honesty, no right or wrong answers, product team not present --- #### B. Warm up & context of use *Objectives* * Learn athletic profile and Strava usage patterns * Discover how they first encountered Athlete Intelligence * Surface goals, motivations and routines *Questions* 1. Tell me a bit about your typical week of running or riding. 2. How long have you been using Strava? 3. Which Strava features do you use most often and why? 4. When did you first notice the Athlete Intelligence summaries? 5. What caught your attention about them? 6. Before Athlete Intelligence, how did you interpret your workout data? 7. Can you recall a recent workout that stood out to you? What made it memorable? 8. What goals are you currently training for, if any? --- #### C. Current experience *Objectives* * Understand perceived usefulness and trust in AI feedback * Identify aspects that delight, confuse or annoy * Capture behavioural impact *Questions* 1. On a scale of 1-10, how valuable have you found Athlete Intelligence so far? Why that number? 2. Describe the most helpful insight you received. What did you do with it? 3. Have you ever ignored or dismissed an Athlete Intelligence message? What led to that reaction? 4. How well do you feel the explanations match your real-world effort? 5. Do you share these AI insights with anyone else? 6. Have you changed any training decisions because of the feature? Please describe. 7. What, if anything, feels repetitive or unnecessary? 8. How quickly after a workout do you look at the AI feedback? 9. If you could improve one thing about how it presents insights, what would that be? 10. Are there moments when you actively avoid the AI summary? --- #### D. Walkthrough & task-based probes *Objectives* * Observe live interaction with the feature * Spot usability issues, comprehension gaps and emotional reactions * Gather verbatim comments while they navigate *Tasks & probes (encourage think-aloud)* 1. Please open Strava and go to your most recent activity. What draws your eye first? 2. Read aloud the Athlete Intelligence section. What stands out? 3. What does this phrase mean to you? (Probe on any jargon or metrics) 4. Is anything unclear or surprising? 5. Show me where you would tap to find more detail. 6. If you wanted to compare this insight with a past workout, how would you do that? 7. Scroll through a few older activities. How consistent is the feedback tone and detail? 8. Do any graphics or charts need more explanation? Which ones? 9. Imagine you want to share this with a coach or friend. Walk me through how you’d do that. 10. Thinking aloud, bookmark or save a takeaway you consider important. How easy or hard was that? 11. Are there points where the AI language feels off or too generic? 12. How comfortable are you with the accuracy of pace, power or heart-rate interpretations? 13. What questions come to mind that the feature does not answer yet? 14. Does the AI ever prompt you to set new goals? How do you feel about that? 15. Overall, what emotions do you experience while reading these insights? --- #### E. Future needs & ideation *Objectives* * Explore desired enhancements and boundaries for AI coaching * Prioritise potential additions or removals for the roadmap *Questions* 1. If Athlete Intelligence disappeared tomorrow, what would you miss most? 2. What new capability would you love to see it provide? 3. Are there scenarios where you would rather not receive AI-generated advice? 4. How would you feel about real-time in-activity AI feedback (e.g. via headphones or watch prompts)? 5. Thinking of recovery, nutrition or goal-setting, where could AI help you next? 6. Is there any data you would never want the AI to analyse? Why? 7. If you could ask the AI one question after every workout, what would it be? --- #### F. Final reflections & close *Objectives* * Capture overall sentiment and final suggestions * End on a positive, appreciative note *Questions* 1. In one sentence, how would you summarise Athlete Intelligence to a fellow athlete? 2. What is the single biggest improvement we could make? 3. Any other comments you would like to share that we have not covered? 4. Thank participant, reiterate confidentiality and next steps for incentives --- ### 5. Checks for alignment * Questions avoid subscription pricing as requested * Flow moves from easy context to deeper evaluation then future thinking * Each section’s objectives map directly to the overarching business and research goals * Total of roughly 55 question prompts fits the 60-minute allowance --- Feel free to tweak timings or wording to suit your moderation style. Let me know if you would like any revisions or a different format. ``` ## Give it a go The example response above isn't perfect, but in 5 minutes you can create a good draft on which you can then build. You can either give the model feedback to refine it, or just copy it into Word/Notion/etc to continue working on it by hand. Try it out and let me know how you get on. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Stop asking “can you give me an example of when you
” in job interviews URL: https://www.philmorton.co/stop-asking-can-you-give-me-an-example-of-when-you-in-job-interviews/ Last updated: 2025-05-28T09:50:30.000Z There’s a certain type of question that interviewers *love* to ask candidates: - *Can you give me an example of when you successfully managed multiple priorities?* - *Can you give me an example of when you had to work with a difficult stakeholder?* - *Can you give me an example of when you led a team through change?* These are classic behavioural questions, based on the idea that past behaviour is the best predictor of future performance. That idea isn’t wrong. In fact, behavioural interviewing can be useful when done well. But the way these questions are often phrased - vague, open-ended, and reliant on memory recall under pressure - is where things fall apart. ## These questions test recall, not competence These questions may seem reasonable to ask, but in practice you’re asking the candidate to: 1. Search their memory for relevant examples across years of experience 2. Quickly evaluate and compare them 3. Choose the one that best matches your intent 4. Structure it into a coherent, persuasive story 5. Deliver it fluently while under stress They need to do this in a few seconds, under pressure. These questions don’t tell you how someone will perform. They tell you how well they can narrate their career under pressure. ## Why they’re hard to answer Take a question like *“Can you give me an example of when you solved a problem creatively?”* Unless you anticipated it and rehearsed an answer in advance, it’s difficult to answer because: - **It’s broad and unspecific**, meaning that many memories could qualify. This increases the breadth of the ‘search’ that you have to undertake within your memory. - It’s easy to answer *“When did you last go on a train?”* because there is one factual answer. - It’s hard to answer *“What’s the most interesting train journey you’ve ever been on?”* because there are many that you would need to recall and evaluate. - [**Stress impairs memory retrieval**](https://pmc.ncbi.nlm.nih.gov/articles/PMC7879075/?ref=philmorton.co)**.** An interview is a stressful situation and you’re putting the candidate on the spot to recall a memory quickly. They might have plenty of relevant experience, but if they can’t bring it to mind immediately, they’ll come across as underqualified. ## We wouldn’t ask this type of question in UX research There’s a category of mistake in research moderation that you see inexperienced interviewers make: **asking a research objective as a direct question**. For example, if your objective is to learn whether people notice and understand discounts on an e-commerce site then you might ask *“Did you see the promotion on the homepage?”*, but this is flawed because it’s leading. Instead, **you need to be more subtle in your questioning** and achieve your objective in a more roundabout way. In this case you might ask them to complete a relevant task (like finding a discounted product) and then follow up by asking what helped them make their decision or what stood out along the way. So when people interview a candidate and want to know if someone solves problems creatively, asking *“Can you give me an example of when you solved a problem creatively?”* is the same type of interviewing mistake. Just as leading questions distort research data, overly open-ended behavioural prompts can distort hiring decisions, rewarding confidence, polish, or preparation rather than substance. ## What to use instead So if you want to know if someone is competent, what should you ask instead? ### 1ïžâƒŁ Ask “*how do you
”* first and then ask for an example second Instead of: *“Can you give me an example of when you dealt with a difficult team member?"* Ask: *“How do you typically approach working with colleagues you disagree with?”* This is far **easier to answer because it’s not a memory recall challenge.** It focuses on process first. If the candidate doesn’t offer an example as part of their response, then you can ask *“Can you walk me through a time when this approach worked well?”* Because you’ve got them thinking about this topic with an easier first question, **if you do ask them to recall a specific example, it’s much easier** as they have had more time to think about it. ### 2ïžâƒŁ Give them a scenario and ask how they’d approach it A variation on the above, this reveals how someone reasons, without the cognitive overhead of memory retrieval. Instead of: *“Can you give me an example of when you showed leadership?”* You would ask something like: *“Imagine you are leading a new, unfamiliar team. What steps would you take in the first month to build trust and momentum?”* If you want a specific example from their past, you can always ask, but again this will be easier for them to recall because you’ve given them more time to ponder the topic before putting them on the spot. ### 3ïžâƒŁ A task or case study Another alternative is to give someone a task to complete or a case study walkthough. These have their own flaws - tasks can be very stressful and case studies can be rehearsed - but they’re not testing someone’s memory recall in the same way and there’s less chance that they’ll completely blank on you. ## Try it out The next time you run interviews, try to do them without any *“can you give me an example of when you
”* questions and replace these with the alternatives above. You might be surprised to find that you aren’t any less able to judge candidates’ competency than before. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### A practical introduction to market research, for UX researchers URL: https://www.philmorton.co/a-practical-introduction-to-market-research-for-ux-researchers/ Last updated: 2025-05-21T09:50:42.000Z There are two trends that I’m seeing at the moment: ## Trend 1: Researchers are doing more strategic research The democratisation of UX research is changing the role of researchers in two ways: - To help designers, product managers and others conduct their own research, they’re expected to do more coaching and governance work (e.g. creating templates). - The research they are doing is more strategic in nature and often [sits outside of product squads](https://www.producttalk.org/2021/07/user-research-and-continuous-discovery/?ref=philmorton.co). Research teams are increasingly organising themselves to reflect this. [A common emerging pattern](https://cms.greatquestion.co/blog/rethinking-webflows-research-structure?ref=philmorton.co) is to divide the research team in two, with some researchers focused on supporting product squads while the others work on more strategic research briefs. ## Trend 2: Strategic briefs can’t be solved with UX research methods alone I’m seeing more and more research briefs with broad objectives that can’t be solved by UX research methods alone. These include things like: - Market sizing - Societal, cultural and industry trends analysis - Social listening UX research and market research are similar in many ways, but the two disciplines rarely intermingle. Many large companies have separate teams for UX research (reporting into the CPO/CTO) and market research (reporting into the CMO). Previously stakeholders would send their briefs to one team or the other, but **I see more and more briefs which require the skills of both**. ## It’s time to learn about market research Like many people who were trained as UX researchers, I haven’t had much exposure to market research tools and techniques beyond basic desk research and quant surveys. It was only recently when I started working with the wonderful [Jane Hovey](https://www.linkedin.com/in/jane-hovey-26b3062/?ref=philmorton.co) \- who comes from a brand and marketing strategy background - that I started to learn all about the tools and techniques used in market and brand research. **The more I learn, the more I think it’s important for UX researchers to upskill in this area.** When you can combine primary qual and quant, secondary research and data analysis, there aren’t any insights you can’t gather. I couldn’t find anything like this out there, so I thought I’d write it for myself. I hope it’s of use to other people with a UX research background in getting to grips with the tools and techniques that you can add to your skillset. ## Similar methods, but with a different purpose There are plenty of articles about the difference between UX and market research, so I’ll keep it brief: - **Market research** is focused on the market for a product or service and answers questions about things like: - *Who are our competitors and what are they doing?* - *Who are our customers (or potential customers)?* - *How big is the market for this product?* - *What are the trends in the industry?* - *How do people perceive our brand?* - **UX research** is focused on how people interact with a product or service and (as I’m sure you know) answers questions like: - *What needs and challenges do people have?* - *How do people use this product?* - *Is it meeting their needs?* There’s of course a lot of overlap - especially when it comes to research into customers’ needs, attitudes and behaviours - but fundamentally the difference is in the purpose of the research i.e. *why* you are doing it. The main adjustment you have to make as a UX researcher coming to this for the first time is to **zoom out and think about broader consumer attitudes and market trends**. ## Categories of market research UX researchers should find these categories familiar because all market research methods **ultimately boil down to either talking to people, observing people or analysing data,** just like UX research. - **Primary vs secondary research:** Primary research means you are collecting new data firsthand for a specific purpose (e.g. running a survey, conducting a focus group or interviewing customers directly). Secondary research (a.k.a. desk research) means you are analysing data that already exists, which could be from numerous sources - more on that later. - **Qualitative vs quantitative methods:** similar to UX research, there are qual methods where we are seeking to understand the ‘why’ and go into a lot of depth with a few people, and quant methods that seek broader and more representative findings about attitudes and behaviours by speaking to many more people. - **Formative vs evaluative research:** Formative is about understanding the problem space (e.g. what people’s needs are) while evaluative is about the solution space i.e. understanding if the thing we have created (e.g. a product or brand campaign) aligns with those needs. ## Primary research methods ### Qual Some of these will seem familiar to UX researchers, but the main difference is the purpose these methods are used for and the skills used in them, for example projective techniques (where you try to reveal people’s unconscious thoughts on various topics). - **One-to-one interviews:** often semi-structured, focusing on broader attitudes, needs, brand associations or product perceptions. - **Focus groups:** used to explore group dynamics, shared attitudes and reactions to concepts. More common in market research than UX. - **Ethnographic and in-situ research:** observing real-world behaviour in context. - **Shop-alongs:** a subset of ethnography used to explore decision-making in retail environments. - **Customer immersion/video ethnography:** often recorded sessions of people using products in their daily lives, for example the experience of unboxing a product. - **Diary studies:** capturing behaviour and reflections over time, often used for longitudinal insights. - **Triads and mini focus groups:** smaller group sessions (2–4 people) to balance depth and group interaction. ### Quant - **Large-scale surveys:** the workhorse of market research, used for concept testing, segmentation, pricing studies, etc. - **Tracking studies and longitudinal surveys:** repeated measurement of the same thing over time (e.g. brand tracking, NPS tracking, campaign effectiveness). - **Methods for evaluating preferences** in a more structured and rigorous way: - **MaxDiff:** used to prioritise features or messages by asking respondents to select the most and least important from a set, e.g. *“Here’s a list of 10 ingredients. Tell us which ones you most and least want on your sandwich.”* - **Conjoint analysis:** used to test how the relative value of different features by analysing choices across many different combinations, e.g. *“Here are three different sandwiches, each with a different combo of bread, filling, and price. Which do you prefer?”* - **Monadic and sequential monadic testing:** showing participants one or more product concepts (individually or sequentially) to evaluate appeal or performance without bias, e.g. *“Here’s one sandwich. Would you buy it? Now here’s a second one, same question.”* Most UX researchers will be familiar with this, but you might not have heard it called this. - **A/B or multivariate testing:** while often associated with product or marketing experimentation, it’s a valid primary quant method when tightly controlled (e.g. regional rollouts or split tests). ## Secondary research methods Secondary research is about **finding and analysing existing information** to answer your question. This is the area which I think many UX researchers will find the most value in expanding their skills and experience. The internal sources below will be familiar, but most UX researchers aren’t super experienced with external secondary sources. After all, the focus of our job is typically to conduct primary research. ### Internal sources - **Analytics and other BI data:** things like product usage data, web analytics and dashboardss. - **Previous research:** even if they weren’t answering your exact question, there’s often transferable insight from prior research. - **Sales and support conversations:** call transcripts, CRM notes and support tickets. - **Customer feedback and NPS data:** including survey responses and satisfaction data. ### **External sources** - **Market research portals:** these are useful because using existing data is quicker and cheaper than running your own primary research: - **Survey-based platforms** (e.g. [GWI](https://www.gwi.com/?ref=philmorton.co), [YouGov](https://yougov.co.uk/?ref=philmorton.co)): these run large-scale, recurring surveys across many markets to capture consumer attitudes, behaviours, media usage and demographics. - **Industry report publishers** (e.g. [Mintel](https://www.mintel.com/?ref=philmorton.co), [Euromonitor](https://www.euromonitor.com/?ref=philmorton.co), [IBISWorld](https://www.ibisworld.com/?ref=philmorton.co)): have in-house analysts who produce detailed reports on various industries. - **Data aggregators** (e.g. [Statista](https://www.statista.com/?ref=philmorton.co), [MarketResearch.com](http://marketresearch.com/?ref=philmorton.co)): collect and visualise statistics from third-party sources like governments and trade associations. - **Trends platforms:** sites like [TrendWatching](https://www.trendwatching.com/?ref=philmorton.co), [Exploding Topics](https://explodingtopics.com/?ref=philmorton.co) and [Think with Google](https://business.google.com/uk/think/?ref=philmorton.co) analyse data from many sources (e.g. [search](https://trends.google.co.uk/trends/?ref=philmorton.co), social, startups) to identify and categorise trends. - **Government, NGO and trade body reports:** public sources like the [ONS](https://www.ons.gov.uk/?ref=philmorton.co), [World Bank](https://www.worldbank.org/ext/en/home?ref=philmorton.co), [OECD](https://www.oecd.org/?ref=philmorton.co) and industry associations often have free data on things like market sizes and demographics. - **Company websites and investor materials:** annual reports, investor materials and press releases are useful when you’re analysing a single business or competitor set. - **Competitor research:** product reviews, pricing pages, feature comparisons, app store feedback, etc. - **Library databases and public business centres:** resources like IBISWorld or Mintel are often free via local libraries (e.g. the [British Library Business & IP Centre](https://www.bl.uk/bipc/?ref=philmorton.co)). - **Academic research and journals:** sites like [Google Scholar](https://scholar.google.com/?ref=philmorton.co), [JSTOR](https://www.jstor.org/?ref=philmorton.co) and [Pew Research](https://www.pewresearch.org/?ref=philmorton.co) can give you societal and long-term behavioural trends. - **Press and industry coverage:** news stories, trade magazines, expert interviews and conference round-ups can help you get up to speed in a new industry. - **Social listening:** tracking what people say about products, brands, and needs across public conversations - **Social networks and communities** (e.g. Reddit, YouTube, TikTok, Instagram, Facebook, Mumsnet, Quora): you can browse comments, posts and discussions manually or use AI to extract sentiment and themes in a semi-automated way. - **Automated tools** (e.g. [BrandMentions](https://brandmentions.com/?ref=philmorton.co), [Talkwalker](https://www.talkwalker.com/?ref=philmorton.co), [Meltwater](https://www.meltwater.com/en?ref=philmorton.co), [Google Alerts](https://www.google.co.uk/alerts?ref=philmorton.co)): these aggregate mentions across platforms, track sentiment and visualise trends. - **Patent and IP databases:** in technical markets, platforms like [Google Patents](https://patents.google.com/?ref=philmorton.co) or [Espacenet](https://worldwide.espacenet.com/?ref=philmorton.co) might be useful for certain projects. - **ChatGPT and other AI tools:** good for finding sources, especially if you are using [Deep Research](https://www.philmorton.co/getting-the-most-out-of-chatgpt-deep-research/) or similar. ## Types of market research briefs Finally, let’s take a quick look at how these methods are typically used to solve business problems. Here are the main use cases: - **Sizing a market and assessing demand:** *“How big is the market for X? How many people might want this product, and how much would they use or pay for it?”* To do this, you might start with secondary data and statistics, then do a quant survey if you didn’t have sufficient data. - **Identifying and profiling customer segments:** *“Who are our customers (or potential customers)? What distinct groups exist within the market?”* You’d typically use a large-scale quant survey to identify and cluster segments, and follow up with qual interviews to understand the motivations and behaviours behind each group. - **Spotting trends and emerging behaviours:** *“What changes in technology, culture or habits might shape the market in the next 1–3 years?”* This usually starts with trend reports, news coverage and social listening. You might then run expert interviews to explore what these shifts could mean for your business. - **Understanding competitors and market positioning:** *“What are competitors offering, and how do we stand out?”* Secondary research is the main method for this. You might also run qual interviews or quant surveys to compare brand perceptions or feature preferences. - **Testing concepts:** *“Which of these ideas is most appealing? What needs to change before launch?”* Similar to concept testing in UX research, but this might be for marketing campaigns rather than just product or service design. - **Prioritising features or benefits:** *“What should we focus on first? Which messages or features matter most to customers?”* Methods like MaxDiff or conjoint analysis help you structure trade-offs and identify what really drives preference, especially when you’ve got too many options to build or say at once. - **Measuring brand awareness and perception:** *“Do people know about us? What do they think of our brand?”*Run quant surveys to measure unprompted and prompted brand awareness, as well as associations or perceptions. You might also track changes over time with a brand tracker or sentiment analysis. - **Tracking satisfaction and loyalty:** *“Are our customers happy? Will they stay or recommend us?”* Use quant surveys like CSAT or NPS to measure satisfaction over time. Supplement with qual interviews or open-text responses to understand pain points and what’s driving scores up or down. - **Understanding price sensitivity and finding the right pricing model:** *“What should we charge? How much are people willing to pay? Will they pay more for certain features?”* Pricing research often uses structured quant methods like conjoint analysis (to test different bundles and price points) or Van Westendorp pricing (to identify acceptable price ranges). ## Be curious What I’ve learned from exploring these methods recently is that **market research isn’t *something that someone else does***. UX researchers can do this stuff too! You might not have experience using these techniques, but they are adjacent enough that you could pick them up with a bit of guidance and coaching. With more people working on strategic briefs, I think **it’s inevitable that we’ll see more people starting to blend UX research and market research.** It’s a good time to be curious and start learning what those people in the ‘other’ research team do. ### Why researchers love writing reports URL: https://www.philmorton.co/why-researchers-love-writing-reports/ Last updated: 2025-05-27T13:38:36.000Z If you ask me to show you the work I’m proudest of, there’s a good chance I’d show you a research report. Not all researchers love writing reports, but many do. Yet **this is at odds with where the industry is going**: faster product cycles, less formal documentation, outcomes over outputs. Speak to stakeholders and they often get frustrated with this - *why do they keep making 80-page reports?* So why do researchers love writing reports and what are the alternatives? ## Make something heavy In her brilliant essay [*Make Something Heavy*](https://www.workingtheorys.com/p/make-something-heavy?ref=philmorton.co), Anu Atluru outlines how the modern world pushes creators to make ‘light’ things: Reels, TikToks, tweets, memes, etc.: > **The modern makers’ machine does not want you to create heavy things.* It runs on the internet—powered by social media, fueled by mass appeal, and addicted to speed. It thrives on spikes, scrolls, and screenshots. It resists weight and avoids friction. It does not care for patience, deliberation, or anything but production.* > > *It doesn’t care what you create, only that you keep creating. Make more. Make faster. Make lighter. (Make slop if you have too.) Make something that can be consumed in a breath and discarded just as quickly. Heavy things take time. And here, time is a tax. And so, we oblige—everyone does. *We create more than ever, but it weighs nothing.** Creators often feel unsatisfied by this constant churn: > **You don’t feel like a true creator because you haven’t made anything heavy, and deep down, you know light things don’t count.* Your output is high, but your imprint is low. You ship, but you do not build. You call yourself a creator, but what have you made that could survive a month offline? A year? A decade? If you stopped posting tomorrow, would anything remain? Creating for 24-hour cycles isn’t freedom, leverage, or legacy—it’s just renting out your time.* What we really crave is to make something ‘heavy’ - something lasting, substantial, and meaningful. > **No one wants to stay in light mode forever.* Sooner or later, everyone gravitates toward heavy mode—toward making something with weight. Your life’s work will be heavy.* The quotes don’t really do this incredible article justice - go and read the whole thing! ## The modern product process pushes researchers to make lighter things Although *Make Something Heavy* is written about creators, it’s easy to draw parallels to other disciplines, including UX researchers. The trend towards ‘just enough research’, continuous discovery and serving faster product cycles means that **researchers end up creating a lot of ‘light’ work**: - Summaries of research in email and Slack channels - Spreadsheets of hypotheses and evidence - Snippets of insight in Dovetail - Miro and FigJam boards - Research recordings But this stuff is often discarded once the purpose of the research has passed. You could argue that researchers should feel that the product itself is their work, but the trend towards **democratising research means that they are often not fully embedded within product squads and don’t feel ownership of the work being done.** There is a joy that comes from making something tangible. Engineers write code and make apps. Designers make an interface that everyone can see and touch. What do researchers get to make? When research is democratised and researchers spend time coaching others, they end up creating little at all. ## Research reports are heavy This can leave people feeling that there’s something missing. **When you look back on your career, what do you have to show for it?** Research reports are heavy, a substantial piece of work that you spent time to deliberately craft. They preserve the work you did in a way that other lighter formats can’t. ## Writing is thinking The other big appeal to writing a report is that *writing is thinking*. The act of creating a report helps you synthesise what you’ve learned and communicate it to others. **Tools like Miro and FigJam encourage divergent thinking**: with an infinite canvas, you can keep expanding your analysis and thinking indefinitely, never reaching a conclusion. **The constraints of a report slide encourage the opposite**: a three-page executive summary forces you to converge and summarise what the research means. Researchers often find this hard. It’s much easier to talk through a Miro board than write a 100-word summary of 24 hour-long interviews. But when done well, the impact you can create from communicating insight effectively, in a way that changes how other people see customers’ reality, is hard to beat. ## If not a report, then what? Despite their appeal, reports are not always the right research output. So what else can researchers create that feels ‘heavy’ and can have a similar impact? - **Videos** are perhaps the most underused tool because people lack the storytelling and technical skills to create them, but they can be hugely effective. Research videos can go viral and they take craft to create, making them feel heavy. - **Journey maps** and other visualisations of the problem space might be a snapshot in time, but they feel substantial and take a huge amount of synthesis to create. They’re useful for others and you can look back proudly at them. - **Personas** get a bad wrap, usually unfairly [because they are poorly executed](https://foolproof.co.uk/journal/what-makes-for-a-good-set-of-personas?ref=philmorton.co). But creating a summary of who customers are and what they need is incredibly valuable. - **Other long-lasting materials that generate empathy for customers:** design principles, presentations that bring customers to life, cheat sheets and so on. The ultimate goal of a researcher is to [generate empathy for customers in the minds of the people making decisions about the experience](https://www.philmorton.co/why-your-research-isnt-landing-and-how-to-fix-it/). Reports are a great way to do this and give researchers the feeling of having made something substantial, but they’re not always appropriate for the task at hand. Think creatively about how to summarise and share insight, and you can find ways to give researchers job satisfaction while aligning with the needs of everyone else on the team. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The 10 habits of highly productive people URL: https://www.philmorton.co/the-10-habits-of-highly-productive-people/ Last updated: 2025-05-07T09:50:48.000Z *I wrote the first version of this a decade ago, but it’s something I keep referring back to when I’m coaching people on productivity and time management.* ## Personal productivity is a skill, not a personality trait Work throws a lot of stuff at us. Tasks, meetings, emails, Slack messages, accountability, responsibility, expectations, side projects, last minute requests, things you’ve been meaning to do but just can’t find the time to get them done but I promise I will get round to them at some point
 It’s ok to find this difficult. No-one teaches us how to deal with it. **Without the skills to deal with all the demands that work places on you, it’s easy to feel overwhelmed.** Some people just get swept along on a tide of meetings, with no real control over how they’re spending their time. ‘Productivity’ isn’t about getting more done with your time. It’s about having control over your time and spending it intentionally. Done well, it brings a sense of calm to your life. Some people think “I’m just not an organised person”, but being organised is not a personality trait. Personal productivity is a skill that everyone can learn, just like public speaking or any other skill we need at work. ## The 10 habits ### 1\. Plan ahead, so you don’t worry if you will have enough time or not ![An illustration of a calendar that has time blocked out](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---1-1.png) Don’t guess at the answer to “do I have enough time?” or “when will you able to do it by?” - Get a clear understanding of what you will be doing. - Visualise it (e.g. using your calendar). - Block your calendar to protect your time. - Communicate to others in your shared calendar. - When the plan changes, you can see the trade-offs you will have to make. Blocking out your calendar for deep work is one of the central ideas in [Make Time](https://www.amazon.co.uk/Make-Time-focus-matters-every/dp/0593079582/?ref=philmorton.co), [Deep Work](https://www.amazon.co.uk/Deep-Work-Focused-Success-Distracted/dp/0349411905/?ref=philmorton.co), [Feel-Good Productivity](https://www.amazon.co.uk/Feel-Good-Productivity-More-What-Matters/dp/1847943756/?ref=philmorton.co) and host of other books. ### 2\. Stay flexible when plans change ![An illustration of changing calendar](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---2-1.png) Sh\*t is going to hit the fan. Don’t let your plan stop you from adapting to emerging events. - Keep adjusting your plan and reprioritising. - Negotiate with others, don’t just say ‘yes’ or ‘no’ to requests, be creative. - Be clear what needs to stay the same and what can move. ### 3\. Write tasks down, so you don’t have to remember them ![An illustration of a to-do list](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---3-1.png) Don’t waste your brain on remembering things. It’s not very good at it and you need your mind for other things. - Keep one list of things you need to do. - If it’s a big task, block out time to do it. ### 4\. Automate recurring tasks, so you never forget them ![An illustration of a to-do item that repeats every Friday](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---4-2.png) Don’t let other people down. Computers are much better at remembering things than you are. - If things happen on a schedule, have a way to automatically add new tasks to your to-do list. - You can do this with Things, Trello, Notion and even Microsoft To-do. ### 5\. Prioritise your tasks so you focus on the right things ![A 2 by 2 prioritisation matrix. On the x axis is urgency and on the y axis is importance.](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---5-1.png) ### 6\. Communicate with others, so you can manage their expectations ![An illustration of a calendar event, "Finalising Exco presentation (DO NOT BOOK OVER)"](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---6-1.png) People can’t read your mind. Be vocal about your busyness, priorities and trade-offs. - Let people know when you realistically think you can get something done. - Use your calendar so people can see what you are working on. - If someone wants something done sooner, explain what the consequences would be. ### **7\.** Set boundaries to protect your time ![An illustration of settings in Outlook, showing working hours.](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---7-1.png) Work expands to fill the time and it will always feel like there’s more to do. - Set your own boundaries. - You can’t always stick to them, but you can try. - Having hard deadlines will make you focus on what really need to be done and force you to prioritise. - 40 hours per week is actually a lot of time: think about how many books you could read per week if you dedicated all of that time to one thing! ### 8\. Design your environment, so you don’t waste time ![An illustration of favourite folders pinned to the sidebar of Finder](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---8-1.png) Think about how to **apply experience design skills to your own work environment** and workflows. - Observe how you work and what you do regularly. - Think of better ways to do your work, for example
 - Put things in a place you can find them. - Add shortcuts for things that you commonly do. ### 9\. Get to know your technology, so you know how it can help you ![An illustration of a preferences modal window](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---9-1.png) Don’t let technology control you. Learn how to get the most from what it’s best at. - Explore settings in Slack, Zoom, Miro, Outlook, etc. - Learn from other people and how they use technology. ### 10\. Keep learning, reflecting and iterating, so you’re always improving ![A three step process: Reflect, Create and Try it out.](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/10-habits---10-1.png) There are three steps to experimenting with this: 1. **Reflect** on what you do and how you work. 2. **Create** new ways to work that are tailor to your needs and preferences. 3. **Try it out** for a while, even if it’s just a week. What you need to do to be productive will change over time as your life changes and your job changes. - Invest some time in reflecting on what you are spending your time doing. - Try new tips and techniques for a week. If they don’t work, try something else. - Keep iterating like you would design work. ## Conclusion **These 10 habits are universal, but how you apply them is different for everyone** \- there’s no perfect system or a system that will stay the same forever. You need to apply the design process to yourself. It takes time to develop a system or set of habits, so try things out and iterate. Over time, you’ll build a way of working that fits you. ## Further reading [Make Time: How to Focus on What Matters Every DayFrom the New York Times bestselling authors of Sprint, 
![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon-1.ico)GoodreadsJake Knapp![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/37880811.jpg)](https://www.goodreads.com/book/show/37880811-make-time?ref=philmorton.co) [Atomic Habits: An Easy & Proven Way to Build Good Habit
No matter your goals, Atomic Habits offers a proven fra
![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon-4.ico)GoodreadsJames Clear![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/40121378.jpg)](https://www.goodreads.com/book/show/40121378-atomic-habits?ref=philmorton.co) [Getting Things Done: The Art of Stress-Free Productivit
In today’s world, yesterday’s methods just don’t work. 
![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon-2.ico)GoodreadsDavid Allen![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/1633.jpg)](https://www.goodreads.com/book/show/1633.Getting%5FThings%5FDone?ref=philmorton.co) [Feel-Good Productivity: How to Do More of What Matters 
The secret to productivity isn’t discipline. It’s joy. 
![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon-3.ico)GoodreadsAli Abdaal![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/142402923.jpg)](https://www.goodreads.com/book/show/142402923-feel-good-productivity?ref=philmorton.co) [Four Thousand Weeks: Time Management for MortalsThe average human lifespan is absurdly, insultingly bri
![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon-5.ico)GoodreadsOliver Burkeman![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/54785515.jpg)](https://www.goodreads.com/book/show/54785515-four-thousand-weeks?ref=philmorton.co) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Critiquing design is a researcher’s job URL: https://www.philmorton.co/critiquing-design-is-a-researchers-job/ Last updated: 2025-05-02T14:26:05.000Z **Researchers are some of the best people at critiquing design work**, but sometimes I see people who
 - Don’t realise that their viewpoint is valuable. - Don’t have the skills or experience to critique design work. - Are hesitant to offer feedback that’s not based on evidence they’ve gathered from the same project. When researchers don’t speak up and contribute during design reviews, teams are missing out on a valuable perspective. ## Why researchers make great design critics Of course they conduct primary research that is directly applicable to what they’re working on. But almost as important is **the accumulated years of experience they have watching people use software**. When you’ve done hundreds or thousands of hours of user interviews, it gives you an instinct for what will and won’t work. This expertise isn’t a replacement for research, but it makes researchers pretty good at guessing what real people will do when they encounter a design. Empathy for customers is essential when critiquing design work, and the ultimate goal of any researcher is to generate this empathy in the minds of those who are making decisions about the experience. ## How to contribute in a constructive way - Speaking of **empathy**, the first step is to have some for the designer you’re working with. Consider the environment they are working in. Are they working with tight deadlines, big constraints, missing information or something else significant? - **Ask non-judgmental questions** (“What were some of the options you considered before settling on this one?”, “Can you walk me through how you approached this part of the design?”) - **Share your experience** of what you’ve seen before, from other research you’ve conducted. It might not be in the same exact context, but you don’t need to present irrefutable evidence here. - **Be as specific as possible.** Vague feedback like “this feels off” isn’t helpful. Instead, explain your reasoning and what you’re basing it on. - **Avoid jumping to solutions**, unless invited. Critique isn’t about redesigning. Stay focused on surfacing risks, raising questions and offering perspective rather than prescribing fixes. ## Magic happens when designers and researchers work together The most successful design teams I’ve seen have always been those where the designer and researcher are working hand-in-hand. - They work on the design together. - They work on the research together. - They present their work together. Most importantly, **they feel joint ownership of the design work**. It’s not the designer’s work who the researcher has helpfully given input into. It’s *their* work. Some researchers just stick to their specialism and concentrate on their research. Who can blame them - there’s a lot to do in that alone. But when researchers isolate themselves to their own domain, the rest of the team is missing out on their full skillset and expertise. The best design outcomes always come from **researchers and designers working more closely together**. ### Getting the most out of ChatGPT deep research URL: https://www.philmorton.co/getting-the-most-out-of-chatgpt-deep-research/ Last updated: 2025-04-23T09:55:21.000Z [ChatGPT deep research](https://openai.com/index/introducing-deep-research/?ref=philmorton.co) has been out for a couple of months now. What is it good for and what are its limitations? ## What is deep research? ![A screenshot of ChatGPT deep research](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/Deep-research.png) As the OpenAI website explains: > *Deep research is OpenAI's next agent that can do work for you independently—you give it a prompt, and ChatGPT will find, analyze, and synthesize hundreds of online sources to create a comprehensive report at the level of a research analyst.* > *
* > *Deep research may take anywhere from 5 to 30 minutes to complete its work, taking the time needed to dive deep into the web. In the meantime, you can step away or work on other tasks—you’ll get a notification once the research is complete.* Essentially, **the more time the AI has to complete a task, the better the results will be**. Deep research gives your prompt more compute, specifically to spend browsing the web and summarising what it finds. The catch is that you only get 10 queries per month on the $20/month Plus plan (or 120 per month on the $200 Pro plan!) This makes it feel like a more specialist tool than anything else on ChatGPT - **it’s more of a scalpel than a Swiss Army knife**. It’s powerful, but narrower in scope, so getting good results depends on how well you write your prompt. ## What it’s good for - **Niche topics** where you know the information is out there, but it’s time-consuming to find, collate or analyse. - Example use case (not the full prompt): *Make a list of 30 newsletters in the areas of product, leadership, design and user research. Check the dates that the last three issues were published, then analyse this data to determine the most common day of the week that newsletters are published.* - **Very specific queries with multiple parameters**, and a standard LLM would not have enough time to work through the task. - Example use case: *Compile a shortlist of 10 rear-mounted bike racks that would fit a 26 inch wheel on a Nihola Family tricycle. I need to mount a battery on top of it, so it needs to be flat and at least 37cm long. It should be for sale in the UK at a price lower than ÂŁ100.* - **Summarising dozens of sources** or multiple viewpoints. - Example use case: *What have Marty Cagan, Teresa Torres and other product management thought leaders written about the role and value of dedicated user researchers? What does the UX research community think about their viewpoints? Given the democratisation of research and the push for non-specialists to conduct research, what is the future role of dedicated UX researchers?* - **Learning about a new topic** in detail. - Example use case: *I want to understand the modern digital experience stack used by enterprise-grade websites and apps. Please break it down into categories and explain the role of each category in delivering, measuring, and optimising the digital customer experience.* ## Limitations - **You only get 10 queries a month** (on the Plus plan), so you can’t use trial-and-error quite as easily to get to the results you need. If it doesn’t give you what you want, you need to put ‘another coin in the machine’. - **There’s a long feedback loop because it’s slower**, so you can be waiting 20 minutes before you realise that it didn’t understand what you wanted. - It always asks clarifying questions before it begins its task, but **once it’s started then there’s no mid-task course correction**. - **It’s only as good as what’s on the public web.** If the data it needs is behind a paywall, in a book, in a YouTube video transcript, on Facebook, etc. then it will struggle. - Because of the above, **the quality of sources may vary**, depending on the domain. - **It magnifies the problems of the standard ChatGPT**, i.e. vague prompting and poor sources will lead to disappointing results ## Tips for getting the most out of it As with anything, the more you use it, the better you get at it. Here’s what I’ve learned so far: - **The longer and more prescriptive the prompt, the better.** As Dare Obasanjo [observed](https://www.threads.net/@carnage4life/post/DGtFwrrvrpy?xmt=AQGzn7RQ5nhxOo6uveU%5F-trUsxUFEnqIPDg1K-ckM1Rjaw&ref=philmorton.co), *“LLMs like ChatGPT reward detailed prompts with better responses, while Google search tends to degrade with longer queries.”* Specify the role it is playing, give it context, clear task instructions and expectations of the output. - **Outline the types and age of sources it should use.** Deep research using sources that you think are not authoritative enough is usually the biggest reason for getting a disappointing result. Ask it to prioritise or only use sources that meet a certain criteria. - These prompts take longer to write, so **get ChatGPT to write the prompt for you**. Use the voice dictation feature to tell it what you want and let it craft it. You will likely need to go through a couple of rounds of iteration before the prompt is ready, but it’s faster than writing it yourself. - **Test prompts with regular ChatGPT (especially the o3 model) before using deep research.** If you don’t need a big report, o3’s response can often be enough. In a recent query, it took over 3 minutes to make 20 searches and review 128 sources, without using up my monthly deep research limit. - **Use an emoji like 🔭 to remember which are your deep research conversations** so you can easily find them later. Deep research isn’t for everything, but when you’re dealing with a complex, time-consuming task that requires real-world data from across the web, it’s a powerful tool to have. Just know that it rewards preparation. If you treat it less like a chat and more like a project brief, you’ll get much more out of your 10 queries a month. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### The case for designing in mid-fidelity URL: https://www.philmorton.co/the-case-for-designing-in-mid-fidelity/ Last updated: 2025-04-16T10:00:26.000Z In a world with design systems and [AI that can create fully-functioning apps in minutes](https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/), why would anyone bother with mid-fidelity design? I argue that mid-fidelity is too easily overlooked. For certain stages of the design process, it offers a sweet spot between low and high-fidelity which allows you to answer “*Are we building the right thing?”* without falling into the trap of focusing on *“How do we build it the right way?”* ## What is mid-fidelity design? ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/Screenshot-2025-03-19-at-15.24.54.png) The Apple Fitness+ Explore screen at the three levels of fidelity. A few quick definitions: - **Low-fidelity:** Anything from a simple sketch to a greyscale wireframe with little to no content. Rough and with minimal detail, this is great for exploring ideas and getting feedback quickly. - **Mid-fidelity:** Has more detail and polish than low-fidelity, but stops short of full visual detail. These designs are not pixel perfect and are intentionally vague in certain areas. - **High-fidelity:** Detailed and polished designs which look like the finished product, ready for hand-off to developers to build. ## The advantages of mid-fidelity - **Speed:** Mid-fidelity is quick to create because you’re not going for fully realised, pixel-perfect designs. - **More people can create them:** Figma can be an intimidating tool for people who are not in it every day. To create mid-fidelity designs, you can use any tool that has text, images and shapes: Miro, FigJam, even (dare I say) PowerPoint. - **Encourages a specific type of feedback:** Because it doesn’t look polished, your stakeholders and research participants are more likely to give you feedback on what it is and how it works, rather than what it looks like. - **They’re throwaway:** Because you’re less attached to mid-fidelity designs, you’re more likely to experiment with them. - **Supports ambiguity:** Mid-fidelity designs don’t show how everything works, which can be useful in certain circumstances. Speaking of which
 ## When it makes sense to use ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/04/Screenshot-2025-03-19-at-16.00.18.png) Mid-fidelity is well suited for articulating a product vision in a storyboard format. There are two main use cases where I think mid-fidelity is most valuable: - **Articulating a product vision** in a proof-of-concept or storyboard. When you want to bring to life what the experience will be like in the future in a compelling way, but don’t want to be too specific about exactly how this will be achieved. Mid-fidelity is perfect for aligning stakeholders on the overall direction without having them pick apart every detail (which you haven’t figured out yet). - **Early-stage concepts** when you need to quickly test and iterate multiple ideas for a new product or feature, to reduce risk and uncertainty. Mid-fidelity helps you, your stakeholders and research participants focus on *what* the concept is and *how* it works, not what it looks like. ## Building the right thing, before building the thing right The next time you’re in the early stages of product discovery and there’s still a lot of uncertainty about what to build, consider working in mid-fidelity. More concrete than low-fidelity but without the temptation to get into the details that comes with high-fidelity, it’s a tool that’s well suited to the stage of the process when you don’t have everything worked out yet. You'll be surprised how much clarity you can create without pixel-perfect designs. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Focus on what causes success, not the success itself URL: https://www.philmorton.co/focus-on-what-causes-success-not-the-success-itself/ Last updated: 2025-04-09T10:00:20.000Z I went on a leadership training course a few years ago and in the breaks, they taught us to juggle. At first, I thought it was just a bit of fun, but it turned out to be **one of the best demonstrations of how people learn** that I’ve seen. It taught us to focus on what causes success, not the success itself. ## How novices try to juggle When you first learn to juggle, you inevitably **focus on catching the ball** that you’ve thrown (the result you’re aiming for). But here’s the problem: - When you move your arms out of position to catch a badly thrown ball, your hand is in the wrong position to make the next throw. - This means you make another bad throw. - The cycle continues until you drop all the balls. If your hands are always chasing badly thrown balls, it’s impossible to maintain a rhythm. ## Focusing on throwing the ball instead Instead, when you focus on the throw **(what causes success)**, you learn much more quickly. They got us to practice just throwing accurately and not even try to catch the ball. When you focus on accurate throwing, it’s so much easier to catch the ball and move it to your other hand to be thrown again. ## Applying this to personal development People get stuck trying to improve outcomes (e.g. be more confident in front of stakeholders) by focusing solely on that outcome. ❌ You don’t become more confident by just thinking “this time I’m going to be more confident!” ✅ You become more confident by improving the things that make people confident (knowledge of the subject matter, being prepared, practicing, having support, etc.) **We often fixate on outcomes, but it’s the cause of them we should focus on.** ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Why your research isn't landing (and how to fix it) URL: https://www.philmorton.co/why-your-research-isnt-landing-and-how-to-fix-it/ Last updated: 2025-04-03T15:24:32.000Z The most common blindspot that UX researchers have is how they’re communicating the insights they’ve discovered. Too many people struggle to present their findings in a compelling way. They write reports that lack impact because they: - Don’t link what has been found back to the business challenge. - Have many observations but little meaningful interpretation. - Are full of overly dense slides that no-one wants to read. - Lack a clear narrative and are devoid of storytelling. On top of this, there’s too much focus on reports as the sole way to communicate what we’re learning about the customers we speak to. ## Research is only as valuable as the impact it creates The goal of a researcher isn’t to “do the research”. The goal of a researcher is to **generate empathy for customers in the minds of the people making decisions about the experience**. If what we learn in our research - however brilliant - is not understood by our audience and does not persuade them to act, it has no value. Teams that fail to grasp this lose credibility with their peers and eventually, their budget to continue this important work. ## Communicating research more effectively There are whole training courses on this ([let me know](mailto:phil.morton@foolproof.co.uk) if you’d like me to run one for your team) but essentially this boils down to thinking about the UX of your research output like we think about the UX of a website or app. For a research report: - **Start by empathising with your audience:** think about the objectives of the research and how it links to the wider business problem that you are helping to solve. Research does not happen in isolation. - **Give people the answers to their big questions:** have an executive summary that answers the main research questions up front. Don't bury the lede. - **Use storytelling techniques to make it memorable:** set the scene, build and resolve tension, have a simple narrative that is used to frame everything you’ve found. - **Design it to be scanned:** every researcher knows that people don’t read websites and apps
 the same applies to your work! Make your slide titles descriptive so that you can get the message without reading the whole page. - **Put excess detail in an appendix:** think about what is essential to tell your story and what you are adding in for the sake of documenting every ‘interesting’ finding that you uncovered. ## How to make your research go viral Once you’ve got your main research output in shape, that’s when the fun really begins. If you want to have an outsized impact and get the work seen by many more people, you need to make your work go viral within your business. Spoiler alert: 80-page PDFs don’t usually go viral. If you think like a marketer, there are several ways to do this: - **Video** is the big one. Two examples here: - In research for a games console manufacturer, we found a huge usability issue. we made a 2 min video clip of people struggling with this particular feature and it was so compelling that it went viral in the company. When this issue was fixed, revenue for this product went up 400%. - We reviewed the application process for a prominent bank account. Colleagues of mine created a video of what the process is actually like to go through by showing every step, every form and every physical letter. It had high production values (although filmed on a budget) and it was shared across the bank, including with the CEO. - **Infographics and posters:** you don’t have to be an illustrator to do this now that we have Canva, Miro, ChatGPT 4o, etc. - **Slack, Teams, etc:** share individual findings (not a summary of the whole research) in an interesting digestible format e.g. run a little quiz or guessing game. There are many more examples but the key message is that you do not need to be an artist to do this. If you can create an Instagram Reel in CapCut or a birthday party invitation in Canva, you can do this. It just takes a bit of imagination. ## I’m the problem, it’s me So the next time you hear a researcher say
 - *They never listen to me!* - *How do we get more visibility?* - *No-one cares what customers think.* 
maybe it’s not their stakeholders. Maybe it’s the way they’re communicating what they’re learning. Research outputs are not just documentation - they bring customers’ reality to life in a way that everyone can understand. Small changes like using better visuals, telling a compelling story, and focusing on key insights can make a big difference. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### How AI coding is reshaping product teams URL: https://www.philmorton.co/how-ai-coding-is-reshaping-product-teams/ Last updated: 2025-03-26T13:40:42.000Z Coding is the area in which AI is having the biggest immediate impact. [63% of professional developers use it](https://survey.stackoverflow.co/2024/ai?ref=philmorton.co#sentiment-and-usage-ai-sel-prof) (and that’s data from May 2024). There are two types of tools emerging: ### 1\. AI coding assistants Tools like [GitHub Copilot](https://github.com/features/copilot?ref=philmorton.co), [Cline](https://cline.bot/?ref=philmorton.co) and [Cursor](https://www.cursor.com/?ref=philmorton.co) support developers in their day-to-day work, helping to suggest code, debug issues, generate tests and so on. ### 2\. Prompt-to-code tools Tools like [Bolt](https://bolt.new/?ref=philmorton.co), [Lovable](https://lovable.dev/?ref=philmorton.co) and [v0](https://v0.dev/?ref=philmorton.co) can create a fully functioning web or mobile app from a prompt in seconds. Most also allow you to import a Figma file as a starting point. ![A screenshot of Lovable creating a Goodreads clone.](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/Screenshot-2025-03-14-at-15.37.28.png) Lovable made me a functional Goodreads clone in a couple of minutes. These are the tools you can use for [‘vibe coding’](https://en.wikipedia.org/wiki/Vibe%5Fcoding?ref=philmorton.co) \- just type or talk to the AI and it’ll make whatever you fancy. Companies making these tools are amongst the fastest growing startups in the world (ever). What does all of this mean for product development? ## Consequence 1: squads won’t need as many developers In [A Vision for Product Teams](https://www.svpg.com/a-vision-for-product-teams/?ref=philmorton.co), Marty Cagan predicts that *“product discovery will become the main activity of product teams, and gen AI-based tools will automate most of the delivery”*. > *
over the next 3-10 years we’ll continue to see the average product team drop from something like 8 people (6 engineers, a product manager and a product designer), down to 3 (a product manager, a product designer, and an engineer).* Smaller teams work better due to the reduced communication overhead. With fewer people per team, an organisation could also work on more problems with the same headcount as before (or simply cut costs). ## Consequence 2: the democratisation of coding Tools like Lovable give non-technical people the ability to create prototypes at a much higher fidelity than before. Think of it like the democratisation of research, but for code. Instead of a clickable Figma prototype - where you can’t even type in input fields - you can have a fully functional website or app. Think of how much better feedback will be from user research if people can use something that actually works. This will mean that when you’re [testing business ideas](https://www.strategyzer.com/library/testing-business-ideas-book?ref=philmorton.co), you’ll be able to **reduce uncertainty and risk** to a much greater extent than before. You can validate and iterate early stage concepts much further, before you have to commit to spending developer resource to make the thing for real. ## What does it mean for skills and careers? If product squads are getting smaller and taking discovery further, then the skills people need to succeed in those teams are slightly different to today. **Being a generalist is going to be increasingly valuable**, as Anton Osika, CEO of Lovable [mentioned on Lenny’s Podcast](https://open.spotify.com/episode/7DOJOSKKir2NW38IrGHT4G?si=125939e91d874575&ref=philmorton.co): > *Doing a bit of everything, being a generalist is, I think, much more important than it used to be. If I’m putting together a product team today, I will really obsess about getting as many skillsets as possible for each person I hire.* In a future where AI can do more specialised work, it makes sense to **broaden your skills** and be able to contribute to more of the product development process rather than just research or design. If you're a researcher, learning more about market analysis, experimentation or product strategy could set you apart. If you're a designer, picking up skills in research or growth strategy will make you even more valuable. Whatever your starting point, broadening your toolkit is the best hedge against teams getting leaner and roles becoming more fluid. [Expertise is still essential in the age of AI](https://www.philmorton.co/ai-wont-make-you-an-expert-but-it-will-make-experts-better/), which feels like a contradiction to the above, but I think it’s more important than ever to be ‘T-shaped’. Broadening your skillset through continuous learning will only serve you in good stead as product teams evolve. **The most indispensable people** will be those who can flex across roles and think beyond their job title. ## Further reading (and listening) [A Vision For Product Teams | Silicon Valley Product GroupA partnership dedicated to teaching best practices to product teams and product leaders![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/touch.png)Silicon Valley Product GroupMarty Cagan![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/svpg-social.jpg)](https://www.svpg.com/a-vision-for-product-teams/?ref=philmorton.co) [Building Lovable: $10M ARR in 60 days with 15 people | Anton Osika (CEO and co-founder)Lenny’s Podcast: Product | Growth | Career · Episode![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon32.b64ecc03.png)Spotify![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/ab6765630000ba8a384f8db2e26be89bff3bb9b4)](https://open.spotify.com/episode/7DOJOSKKir2NW38IrGHT4G?si=b81ef6cbd4914b6e&ref=philmorton.co) [Inside Bolt: From near-death to \~$40m ARR in 5 months—one of the fastest-growing products in history | Eric Simons (founder & CEO of StackBlitz)Lenny’s Podcast: Product | Growth | Career · Episode![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon32.b64ecc03-1.png)Spotify![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/ab6765630000ba8aa07b948be7568fbac074f7e2)](https://open.spotify.com/episode/3nBCnIGltiFsujZsrxxEB3?si=d82ac4b9c424488b&ref=philmorton.co) [The End of Programming as We Know It![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/favicon.ico)O’Reilly MediaTim O’Reilly![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/binary-1187198_1920_crop-cf3ecf0e521f99a1bb5c5565755c9c4d-1.jpg)](https://www.oreilly.com/radar/the-end-of-programming-as-we-know-it/?ref=philmorton.co) [The 70% problem: Hard truths about AI-assisted codingA field guide and why we need to rethink our expectations![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/icon/https-3A-2F-2Fsubstack-post-media.s3.amazonaws.com-2Fpublic-2Fimages-2F5461e150-2b93-418a-b0b7-0499a1a2ac22-2Fapple-touch-icon-180x180.png)ElevateAddy Osmani![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/thumbnail/https-3A-2F-2Fsubstack-post-media.s3.amazonaws.com-2Fpublic-2Fimages-2F0e49ab22-0fac-4959-afa6-b6e226056db4_6072x6072.jpeg)](https://addyo.substack.com/p/the-70-problem-hard-truths-about) ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### AI won’t make you an expert, but it will make experts better URL: https://www.philmorton.co/ai-wont-make-you-an-expert-but-it-will-make-experts-better/ Last updated: 2025-03-21T10:50:16.000Z Imagine you’re working on a project with two colleagues. One is a seasoned expert, the other is just starting out. Both have access to the same AI tools. Who do you think would benefit more? A couple of articles I’ve read recently suggest that it’s the more experienced colleague. ## AI amplifies expertise rather than compensating for inexperience [This article from the Economist](https://www.economist.com/finance-and-economics/2025/02/13/how-ai-will-divide-the-best-from-the-rest?ref=philmorton.co) argues that rather than AI being an equaliser, it will actually increase inequality: > *In complex tasks such as research and management, new evidence indicates that high performers are best positioned to work with AI. Evaluating the output of models requires expertise and good judgment. Rather than narrowing disparities, AI is likely to widen workforce divides, much like past technological revolutions.* Essentially, experts can filter AI’s suggestions effectively, whereas novices struggle to discern good from bad outputs. The wonderfully titled [*New Junior Developers Can’t Actually Code*](https://nmn.gl/blog/ai-and-learning?ref=philmorton.co) by Namanyay Goel has a similar theme: > *Every junior dev I talk to has Copilot or Claude or GPT running 24/7\. They’re shipping code faster than ever. But when I dig deeper into their understanding of what they’re shipping? That’s where things get concerning.* > > *Sure, the code works, but ask why it works that way instead of another way? Crickets. Ask about edge cases? Blank stares.* > > *The foundational knowledge that used to come from struggling through problems is just
 missing.* > > *We’re trading deep understanding for quick fixes, and while it feels great in the moment, we’re going to pay for this later.* Without struggling through problems manually, people miss out on deep learning and problem-solving skills. ## The key skill in the AI era: knowing when to use it AI is great when it’s used for: - Speeding up work that you already have expertise in. - Getting a quick answer to something that is not your core competency. But if you rely on it too heavily for your core work, you risk skipping the learning process and delivering mediocre work. Imagine a UX researcher who has never analysed research without AI. Are they really going to be able to judge the quality of AI-generated insights and craft a memorable narrative for the research? ## AI as a power tool, not a cheat code Although it would be handy to skip all the learning and get all of the answers from AI, the best performers in the future are going to be the ones that blend expertise and AI skills. When you’re using it, you need to ask yourself: - How important is the quality of the output? - What’s the risk of getting it wrong? - Am I able to judge the quality of the output? The only way to develop this intuition is to use AI, but just don’t do it at the expense of developing your foundational knowledge. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Using AI to write more realistic prototype copy URL: https://www.philmorton.co/using-ai-to-write-more-realistic-prototype-copy/ Last updated: 2025-03-17T17:19:05.000Z I recently finished a project where I used AI to write 90% of the copy for my mid-fidelity concepts and prototype. In the ‘olden days’ of 2 years ago, you’d have to write all of the copy in your designs yourself, or resort to lorem ipsum. But what if there was a better way? ## The problem with writing copy yourself Typically a product/UX designer is the first person who writes the copy in a concept or design. Content designers/UX writers don’t always get involved in the early stages of design work (although they should!) This is a pain because: - **It’s time-consuming:** Writing copy manually takes time, especially when you're not a deep subject matter expert. - **The alternative (lorem ipsum) is unrealistic:** Content makes up a big part of many experiences, so when people see this in research, they find it confusing and can’t give good feedback on your design work. ## Using AI as your copywriter In my project, I used AI to generate copy for my mid-fidelity concepts. Just to be clear, this was throwaway copy for research, rather than what would go into production. Using AI rather than writing it myself meant I was able to: - **Speed up the process:** I could create and iterate the concepts much faster. - **Write in their tone of voice:** You can give AI the tone of voice guidelines or sample copy so that it can do a decent job of sounding on-brand. I think this is most useful when you have large amounts of text in your designs, rather than small pieces of UI copy. ## Making your data realistic The other thing that AI can help designers with is making sure that prototypes have realistic scenarios shown in them. I’ve seen so many prototypes that have unrealistic data because the designer isn’t an expert in the subject matter or they didn’t have the time to really think about the content/data they’re showing. In this project, I was designing a recommendation engine. With AI, I was able to create realistic recommendations without having to spend weeks becoming a deep subject matter expert. ## An example prompt Imagine that we’re designing a product wizard for the [John Deere](https://www.deere.com/en/?ref=philmorton.co) website. You tell it what you need and it suggests suitable products, using a gen-AI explanation to make the recommendation more personalised. Without deep expertise in agricultural machinery, you might struggle to write realistic copy for your prototype. This is where we can get AI to help us. Let’s imagine we’re working on the last step of the product wizard, which provides the recommendation: #### Write realistic product recommendations **Act as a digital marketing copywriter specialising in agricultural machinery, with over 20 years of experience.* **Imagine a John Deere customer who runs a mid-sized dairy farm is interested in purchasing a new tractor to improve their feed management. Specifically, they want equipment that can manage precise feeding schedules, ensure balanced nutrition for their herd, reduce feed waste, and streamline feed mixing and delivery processes.* **Recommend 3 suitable John Deere products for this scenario. For each recommendation, write:* - **Product name* - **What it is: (brief description of the product)* - **How it works: (explain briefly the key functionality)* - **How it helps: (directly connect the product's capabilities to the farmer’s stated needs)* - **Why we’ve recommended it for you: (highlight why this product is particularly suitable for a dairy farmer focused on feed management)* - **Price in USD* - **A link to the product on the John Deere website, where I can validate the information you have given me.* **Finally, make a recommendation of which of the three options is most suitable and explain your reasoning, referencing the customer’s input.* This will give you much more realistic copy for your prototype than lorem ipsum, and will take you a lot less time than if you had to do the necessary research yourself (of course you still need to find a way to check the output). ## Final thought While AI isn't a replacement for professional UX writers/content designers, it can make a big difference to how fast you can create and iterate early stage concepts and designs. ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### You should probably be doing more research with couples URL: https://www.philmorton.co/you-should-probably-be-doing-more-research-with-couples/ Last updated: 2025-05-02T14:25:04.000Z The vast majority of qual UX research is one-to-one. But in reality, many products and services aren’t bought or used by just one person: it’s a couple making the decision. Just consider the data: - [71% of holiday bookings are made by couple or families](https://www.travolution.com/news/travel-sectors/largest-accommodation-survey-reveals-travellers-openness-to-ai/?ref=philmorton.co#:~:text=France%20%2810,and%20the%20US%20%287) - [Partner influence is the biggest factor in car purchases](https://www.am-online.com/news/digital-marketing/2015/09/24/spousal-influence-trumps-price-on-car-purchasing-decisions?ref=philmorton.co#:~:text=The%20company%E2%80%99s%20own%20research%20showed,and%20make%2Fmodel%20at%2015) - [More than three-quarters of couples share passwords](https://www.expressvpn.com/blog/we-asked-couples-in-4-countries-about-password-sharing/?srsltid=AfmBOop-MuFX4vsumFjwIemQrZtjc3dpl8wu%5F70K5hZsO5656tTuXGF4&ref=philmorton.co#:~:text=But%20this%20rule%20certainly%20doesn%E2%80%99t,a%20password%20with%20a%20partner) - [63% of first-time buyer mortgages are joint](https://www.morningstar.co.uk/uk/news/AN%5F1674634047670754900/nearly-two-thirds-of-uk-first-time-buyer-mortgages-are-in-joint-names.aspx?ref=philmorton.co#:~:text=Some%2063%25%20of%20first,as%20sole%20applications%2C%20Halifax%20said) - [61% of couples have a joint bank account](https://www.tsb.co.uk/news-releases/were-not-so-romantic-when-it-comes-to-money.html?ref=philmorton.co#:~:text=The%20majority%20of%20couples%20have,and%20financial%20freedom%20when%20spending) - [40% of life insurance policies are joint](https://www.unbiased.co.uk/discover/insurance/life-insurance/what-is-a-joint-life-insurance-policy?ref=philmorton.co#:~:text=sum%20payout%20%20if%20either,away%20during%20the%20policy%20term) We do so much one-to-one research because it’s familiar, easy to recruit for, and all of the tooling is set up to support it. But if we’re only doing research with one person, we’re often missing half the story. ## What it looks like in practice A few years ago, we worked with a UK mortgage provider that was launching two novel products. Our job was to redesign their website to ensure people could easily understand these new offerings, especially since most mortgages are taken out jointly. So we conducted one-to-two research: 1. We started the interview with both people in the room. 2. One of them explored the prototype while the other waited outside. 3. We brought them back together and the person who had been using the prototype had to explain how the mortgage worked to their partner. We quickly saw that even when one person ‘understood’ the mortgage, they struggled to explain it clearly to their partner. Seeing the back-and-forth between partners revealed gaps and misunderstandings that wouldn’t have surfaced in a one-on-one interview. ## Tips for research with couples If you’re interested in trying this out, here are a few things to bear in mind: - **You’ll likely need more time.** Consider 60-90 min interviews. - **Do the research in person, if you can.** You can make it work remotely, but you might not be able to see both people on camera unless they are sat very closely together. - **Create a clear structure for the session.** Don’t just run it like a normal qual interview that happens to have two participants in it. Like the case study above, use it to run the session in a different way to the standard one-to-one session. - **Encourage discussion between the couple.** Don’t just ask questions to each in turn; you want to see how they interact with each other. How do they explain things to each other? How do they decide what to do? - **Don’t let one person dominate the conversation.** Even though one person might lead on researching a purchase or using a product, you still want to hear from their partner if they are involved in decision-making. - **Share video highlights.** Because it’s an unusual type of research, this is a perfect opportunity to engage your stakeholders by showing them how people are *really* buying or using your product/service. Few researchers seem to use one-to-two research, but it’s something more people should have in their toolbox. It’s more work to set up, but hardly anyone else is doing it, so you’ll uncover insights that your competitors won’t. Have you tried this before? Let me know by hitting reply or [on LinkedIn](https://www.linkedin.com/in/philipsamorton/?ref=philmorton.co). ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk). ### Welcome to Desk Notes! URL: https://www.philmorton.co/welcome-to-desk-notes/ Last updated: 2025-03-18T10:08:23.000Z Welcome to my new website and the very first issue of **Desk Notes**! [I’ve been writing on LinkedIn](https://www.linkedin.com/in/philipsamorton/?ref=philmorton.co) for the past few years, sharing thoughts on research, design, leadership and career development. But LinkedIn has its limitations - posts disappear over time and while brevity is great, sometimes you need more space to go into more detail. That’s why I created this site and newsletter - to give my writing a more permanent home and to make it easier to revisit useful ideas. ## Why Desk Notes? I had a mentoring catch-up recently and wanted to share some tools and resources, but it was difficult to content that was hidden away deep in a shared drive and find older posts on LinkedIn. Desk Notes is my solution to that. This newsletter is built around the idea of learning by osmosis - the kind of insights you pick up just by sitting next to someone (me!) at work. I want to share the knowledge and experience I've gained over the last 15 years, to help other people. Each issue will be a mix of: - Lessons from experience, across research, design and leadership. - Tools, frameworks and mental models that I've found useful. - Book recommendations, useful links and industry trends. ## What’s next? I'm planning to post weekly and to get off to a good start, I've got the first few newsletters lined up already. Beyond that, I'd love to know what kinds of topics would be most useful to you. Hit reply and let me know! đŸ“© ![](https://storage.ghost.io/c/32/f8/32f8d7f6-bab3-4cc4-8608-ddbeeef397cc/content/images/2025/03/IMG_1179.jpeg) ## Sign up for Desk Notes My free, weekly(ish) newsletter where I share insights about product design, research and leadership. Subscribe Email sent! Check your inbox to complete your signup. Check your email for the confirmation link (might end up in spam/junk).