AI made design work faster. So why isn't anyone happier?

A Nielsen Norman Group survey published last week asked 521 UX practitioners how AI is affecting their work, what it found is rather odd. Almost everyone said AI made their tasks faster, yet the average feeling about using it at work came out at 3.57 out of 5, which is positive but only just. In my experience the gap comes mostly from how AI gets used inside teams (with very little structure, and with leadership expecting the time saved to turn straight into more output), and I think designers who want to stay useful will probably need to get more skilled at their actual craft because of it.

What did the Nielsen Norman Group study find?

The survey ran through August and September this year. On average, respondents rated AI's effect on how long tasks take at 4.2 out of 5 (where 5 means much shorter), and for every task on the list, most of the people using AI for it said it sped things up. Writing code, desk research and analysing research data saw the biggest gains. Anything involving actual users, like interviews or usability testing, saw the least, which I suppose makes sense because a conversation with a customer doesn't really get any shorter.

The more interesting part is in the comments people left at the end. Of those who wrote something down, 69% expressed at least some negative feeling, and the most common reason was unreliable output and the time it takes to check and fix it. Mandates and inflated expectations from leadership came up a lot too. The researchers' own conclusion is that the time saved rarely stays with the person who saved it, because it tends to get absorbed by higher expectations or by cleaning up somebody else's (poorly executed) AI-generated work.

I found that oddly reassuring, mostly because it matches what I've seen over the last couple of years, first as a designer and now running a small product company.

Where AI genuinely helps in my design work

I'll start with the good bits, because there are plenty. AI is really good at getting through repetitive work that isn't especially creative, and drawing wireframes is the obvious example. Being able to put something visual in front of myself, a stakeholder or a user within an hour (rather than a couple of days) means I can find out much earlier whether a direction is viable, and I spend a lot less time polishing ideas that turn out to be duds.

It's also, somewhat unexpectedly, good at spotting the things I've missed. I recently designed a drill-down data exploration feature for Corha, where a user moves from a high-level view down through several layers of detail. It was a fairly big piece of work in itself, and at each level there were bits of data I hadn't considered that turned out to be essential for someone to interpret what they were looking at. Working through it with AI (with a well-formed prompt, which matters a lot) surfaced most of those edge cases early on, and the first round of design options was a much richer experience for it.

In a previous role at an AI-forward startup, we were using tools like Figma Make in their fairly early days, mainly for the prototyping side. That meant we could get simple but properly formed prototypes in front of users and stakeholders very early in the design process. The feedback was better for it, probably because people were reacting to something that felt close to being real.

What happens when teams push it too far

The catch, in my experience, is structure. For general questions AI is great at giving a broad answer from lots of angles, but for proper work and task completion it needs structure in how you prompt it and in how the output gets captured and checked. Without that it tends to lose focus, and the results can be poor or misleading in ways that aren't obvious at first glance.

What makes me a bit apprehensive is that people who couldn't judge a piece of design work before can now produce one. Event and small business flyers are probably the clearest example of this; there are a lot more of them around and many are, honestly, terrible, because the person making them can't tell what's wrong. The same thing happens inside product teams, just less visibly. Several of NN/g's respondents described spending their time fixing unusable prototypes that someone else had generated in minutes.

In that previous role, the CEO and I were fairly deliberate about managing expectations. We were both quick to point out when a model or tool had produced something poorly formed, and nobody was pretending a generated prototype was a finished design. I don't think that's especially common, though. Anecdotally, I've seen senior people in a few organisations get so excited about AI that they push it into everything, and the important things (thinking time, talking to users, checking the work) are the first to get squeezed so the team can go faster. The quality of the output drops to some degree. Probably more importantly, leaders can start to lose the respect of their teams, because people are being asked to put their names to work they aren't happy with. I think we're in the early stages of seeing the results of this, and the survey kind of backs that up.

How should designers and founders use AI?

My view is that AI should be treated like any other tool, and Excel is the comparison I keep coming back to. You can run structured calculations over enormous data sets in Excel far faster than you ever could with a calculator, but only because the work is structured properly and the person using it stays alert to an arithmetic blunder hiding somewhere. Nobody (sensible) hands a spreadsheet to the board without checking it, and AI output deserves the same treatment, just with a much wider range of ways to go wrong.

That has a slightly counterintuitive consequence. People in specialist roles like UX design probably need to become more knowledgeable about their discipline because of AI, since spotting a poor output depends on knowing what a good one looks like. That kind of vigilance has to be wired into you, and in most cases it only comes from having done the work yourself.

So my advice, if you're a designer, is fairly simple. Keep designing. Not everything needs to go through AI (although it can genuinely help in plenty of situations), and it's worth continuing to learn how to do things by hand and how they tick beneath the surface. That's what lets you be a good orchestrator and moderator of whatever your tools produce.

For founders and product leaders, the lesson here is probably to measure whether the quality of the work is going up as well as the speed. NN/g suggests looking at the whole system, meaning the total effort across everyone involved in shipping something of acceptable quality, and I think that's right. A prototype generated in ten minutes that then takes a designer three days to make usable has mostly just moved the work onto someone else's desk.

Source: AI in UX: Work Is Faster, But Practitioners Have Mixed Feelings, Nielsen Norman Group, 23 October 2026.






Is AI making UX designers more productive? Mostly, yes. In Nielsen Norman Group's 2026 survey of 521 UX practitioners, most people using AI for a task said it made that task faster, with the biggest gains in desk research, analysing research data and writing code. Tasks involving real users, like interviews and usability testing, saw the smallest gains.

Why do designers feel negative about AI if it saves them time? The most common complaint was unreliable output and the time spent checking and fixing it. Mandated use and inflated expectations from leadership also came up often, and in many cases the time saved went into producing more work or fixing other people's AI-generated work.

How should designers use AI without losing quality? Treat it like any other tool, with structure in how you prompt it and a habit of checking what comes out. It works best for repetitive work like wireframes and for catching edge cases, and designers still need to know their craft well enough to spot a poor output.

Will AI replace UX designers? Most practitioners don't think so. In the NN/g survey, 65% expected their role to stay as important or grow, though designers rated their job security lower than product managers did. In my experience the designers who keep designing, and keep learning how things work beneath the surface, will be the ones best placed to direct AI tools well.

Luke Hawkins is the founder of Corha and has spent ten years as a product designer across early-stage SaaS and scale-ups. Corha helps small teams and startups bring everything their customers are telling them into one place, so the whole team can act on it. You can try it free at corha.app.