The Existential Crisis of UX Designers in the Age of AI

AI is having a bit of a moment in design.

You can now type “design me an onboarding flow” into a tool and get something that looks clean, polished, and vaguely familiar in a few minutes. It’s fast, it’s impressive, and it can make a lot of teams quietly wonder: do we still need UX designers at all?

My view: if your ambition is to ship something that looks nice and works “well enough”, AI will absolutely get you there.

If your ambition is to create an experience that is authentic, compelling, and genuinely different, you still need designers.

How AI “thinks” about UX

When we ask AI to help us design, we’re usually asking it to look at content in a fairly narrow way.

“Design me an onboarding flow for a budgeting app” basically tells the model: go and find patterns from other onboarding flows and remix them into something that looks like an onboarding flow.

That has some clear benefits:

• It condenses days of desk research into a handful of prompts.
• It surfaces common patterns, layouts and copy styles very quickly.
• It can help non‑designers get from a blank page to something testable.

But underneath it all, AI is still pattern‑matching on what it has seen online – flows, design systems, UI kits, screenshots, articles about “best practice”. It is not looking at the messy, human context around those designs. It doesn’t know whether users were stressed, confused, delighted or quietly ignoring half of what was on the screen.

It can design an experience.

It doesn’t yet know if it’s the right one.

How designers actually work

As product and service designers, we rarely design from a purely visual reference library. We design from experience.

We:

• Ship something, even if it’s rough.
• Watch how real people actually use it.
• Ask questions.
• Make tweaks, see what changes, then do it again.

Over time, we build a sense of what really works, not just what looks like it should. We remember the project where the “clean” pattern failed because users were anxious. The onboarding that technically followed all the rules but collapsed because people didn’t trust the brand yet.

So when we design a new flow, we’re not just pulling from Dribbble or design blogs. We’re pulling from:

• Past experiments and A/B tests.
• Real conversations in research sessions.
• Stakeholder constraints, business realities, odd edge cases.
• Things we’ve seen in the wild across different products and sectors.

That lived, sometimes slightly chaotic mental library is what lets us diverge. We can look at a problem and think, “The obvious pattern is X, but given these users and this context, we might actually need to do Y.” That kind of “odd” decision is exactly what an AI, trained to follow patterns, would struggle to make.

Data tells us what, not why

Analytics are brilliant at telling us what is happening. They tell us where people drop off, which button they click, how long they linger on a screen, which version converted better.

But they rarely tell us why.

• Why did users abandon the form on step 3?
• Why did they choose that slightly worse plan over the “best value” one?
• Why did they ignore the shiny new feature we were so proud of?

To get to the why, we still sit in user testing sessions and research interviews. We watch faces, hear the awkward pauses, and ask follow‑up questions that weren’t on the original discussion guide.

That qualitative layer is where a lot of the important design decisions come from:

• “They’re not confused by the wording; they don’t trust us yet.”
• “They’re not skipping this step because it’s boring; they’re in a rush and on their phone at a bus stop.”
• “They’re not lost; they just don’t think this is for them.”

AI can help summarise research notes and cluster comments, but it doesn’t walk out of a session feeling that slight knot in your stomach that says, we’ve missed something important. That feeling often drives the next iteration.

We’ve already seen the first wave

None of this is happening in a vacuum.

We’ve already seen how AI and data are reshaping UX in more subtle ways.

Take banking as an example.

Some organisations, like HSBC, are already using data to adapt experiences based on your behaviour. If you’ve paid a cheque in at a physical branch, the app can later highlight the ability to deposit cheques in‑app instead. It’s a small, targeted nudge based on what you’ve actually done, not just who you are on paper.

In 2024, I wrote in more detail about Hyper‑personal AI UX and the dawn of Dynamic Knowledge Rendering – the idea that the same underlying knowledge can be rendered differently for different people, in different contexts, to give each person what they need in that moment.

So personalisation isn’t new.

What’s new is what happens when AI can not only personalise what you see, but start to change how the product itself behaves.

The next step: self‑changing products

Let’s push this a bit further.

We’re starting to see hints of systems that can:

• Watch user behaviour in real time.
• Automatically generate alternative designs or flows.
• Set up and run A/B tests.
• Roll out the winning version.

Now imagine that loop running with almost no human in the middle. The platform quietly observes, generates variants, tests them on segments of users, and ships improvements on its own. Screens, copy and journeys evolve automatically.

From a business point of view, that sounds exciting:

• Faster experimentation.
• Less manual work.
• Continual optimisation.

From a designer’s point of view, this is where the existential wobble starts. If the system can watch, test and tweak on its own, what exactly is our job?

Where designers still matter

For me, the answer depends on the ambition of the product.

If you want something that:

• Looks modern.
• Matches common patterns.
• Performs “okay” on surface metrics.

…AI can help you get there faster than ever.

But if you want something that is:

• Authentic to your brand and your users.
• Genuinely compelling, not just optimised for clicks.
• Distinctive in a world of very similar AI‑generated interfaces.

…you still need designers.

Designers do a few things AI doesn’t:

• Frame the problem: Are we optimising the right thing? Are we measuring what actually matters to users and the business, not just what’s easy to track?
• Hold the bigger story: How should this product feel? What does “a good experience” mean for these people in this context?
• Navigate trade‑offs: Sometimes the most “efficient” flow is not the most ethical, inclusive or long‑term healthy one.
• Bring lived experience: We connect dots across projects, sectors, and life outside the screen.

AI can build screens. Designers still build experiences.

AI as a co‑creator, not a replacement

All of this is why I prefer to think of AI as a co‑creator rather than a competitor. When I’m building products with the help of AI, it’s a similar dynamic.

I’m bringing over a decade of digital product experience, plus research into the different stakeholders and their needs, and using AI to:

• Explore patterns and options faster.
• Generate variants I might not have had time to mock up myself.
• Stress‑test ideas before putting them in front of users.

The point isn’t to let AI decide what the product should be.

The point is to use it to move faster through the mechanical parts of the work, so there’s more time and energy for the human parts: understanding people, asking better questions, and making more thoughtful decisions.

So, should designers be worried?

If we stay purely in the business of pushing pixels and assembling standard patterns, then yes – AI will eat a lot of that work.

But if we lean into:

• Framing the right problems.
• Understanding the messy “why” behind the data.
• Making sense of conflicting needs across users, teams and constraints.
• Using AI as a powerful assistant rather than a threat.

…then AI doesn’t erase the role of UX designers. It just forces us to drop the illusion that the job was ever mainly about screens in the first place.

The existential crisis isn’t “Will AI replace me?”

It’s “Am I doing work that’s uniquely human enough that AI can’t?”

That, I think, is a much more interesting question for all of us in UX right now.

‍