Why I’m Not Worried About My UX Job in the Era of AI

At Dovetail’s Insight Out conference, one of the speakers, Jess Holbrook, said something that struck a chord. (I am quoting from memory, so apologies if it’s not quite right.) He said, “My goal is to build better products, and that’s why I do UX research.”

This may seem obvious, but it is also profound: UX and design methods are just tools for building better products.

UX Tools Are Constantly Changing

UX methods and tools have always changed.

I’ve been practicing UX for 20 years, spending most of that time at NN/g. When I was hired, my welcome present as a new NN/g researcher included (besides a pile of books) a stopwatch — because back then, time on task could only be collected manually.

I used that watch exactly once. Soon, the earliest unmoderated usability-testing platforms (like UserZoom and UserTesting.com) arrived. We started using them for quantitative user testing (and for qualitative testing also). My stopwatch was no longer needed — these systems would automatically record the time, way more accurately than I could with that watch.

That was just one of many shifts I’ve experienced. Another came with the pandemic. Before then, most of NN/g’s usability testing was done in person — it was hard to find a tool that would allow participants to record their screens remotely. But once Zoom became a standard part of daily work life, we found ourselves rarely running in-person user studies — not because those sessions aren’t valuable or often more effective than remote ones, but simply because remote studies are far more cost-effective and could reach participants located anywhere around the world.

With each change in available tools, our jobs and techniques had to adapt accordingly.

AI Is a Tool

AI is just another tool — or more accurately, a set of tools. We don’t yet fully understand how to use it efficiently.  We are still experimenting with applications ranging from synthetic users to moderation in user studies, from thematic analysis to generating interface components.

As AI tools become stronger and more well-defined, our job is to figure out how they can help us build better products more efficiently and less expensively, and which of our current methods they can replace or augment.

Take usability testing, for example. We don’t do usability testing because we love chatting with users or watching them stumble. We do it because it helps us identify usability issues in an interface. Testing with five users allows us to find 85% of the usability issues in a UI, fix them, and move to the next design iteration.

If we can find 85% of the issues using AI, then let’s do it. Finding even 50% would still be worthwhile, provided we run some user testing to catch what AI might miss.

The point is: tools come and go. We’ve adapted to new ones and moved past old ones many times before. What makes AI feel different is how intimidating it seems. It can feel daunting to keep up with so many new systems that appear every day. And there's so much technical jargon (deep learning, transformers, reinforcement learning, retrieval-augmented generation) that even informal LinkedIn posts can feel inaccessible. This creates the impression that AI is hard to master.

And that’s partly true. But much of it is noise. You don’t need to understand all the inner workings of AI to experiment with it.

New Methods Will Emerge

When a new technology appears, it often prompts the creation of new research methods. For example, customer-journey maps became popular once mobile devices became ubiquitous. People began freely transitioning across devices to accomplish a single goal, so we had to think beyond individual channels and consider that broader goal.

Similarly, we will need new methods for studying and designing AI-based interactions. If generative UIs are dynamically built to suit a user’s specific needs, we will need ways to evaluate them. How will we judge if the AI-generated interface is good or bad? What inputs must we give the AI to produce a usable experience? What data should the AI use to generate that interface?

Even if generative UIs become common, they will likely coexist — at least for a while — with traditional interfaces. AI will also bring new tools for improving those traditional UIs, but we’ll need to study and refine them. We’ll need to compare these new methods to the ones we already trust.

For instance, how does a study moderated by an AI compare to a traditional moderated or unmoderated study? When does AI moderation work well, and when does it fall short? How reliable are the AI-extracted themes and where do we need to go deeper? If AI identifies usability issues in a UI, how many does it find? Are they the most critical ones? Or just the most obvious?

The answers to these questions will inform how we use these tools. They’re questions we’re studying at NN/g.

The UX Skills We Will Need

If your job relies heavily on one specific UX method — like conducting usability tests — then yes, there’s a chance that task may be automated or replaced. Just like stenographers became less necessary once live-transcription technology became reliable.

But if your focus is on understanding people and designing usable interfaces, there will always be work for you.

Maybe tomorrow you’ll do less user testing. Maybe AI will interview users or generate UIs in real time.

But someone will still need to guide it.

Designing better products is a messy, ill-defined problem. It varies with context. Even if we had a single great AI tool for UX, who would define the problem clearly enough for the AI to address it? Who would ask the right questions? And who would judge the output?

That will be our job. These are the skills we use today — and they’re the ones we’ll use tomorrow.

When I started my first UX job, I had done exactly one usability study. But I had run and analyzed dozens of psychology experiments and built computational models of human behavior. None of these were listed in the UX-job descriptions. However, they gave me an understanding of human behavior and the analytical ability needed to ask the right questions.

These are the skills that have served me again and again in my work. These are the skills that will carry us through the age of AI. They’re the ones to cultivate now, no matter what tomorrow looks like.