Nielsen Norman Group survey shows AI speeds up UX work, but most comments are negative

On October 23, 2026, Nielsen Norman Group published the results of a survey of 521 UX professionals on the use of artificial intelligence at work. The study, conducted by Maria Rosala and Rachel Banawa, reaches a conclusion that should unsettle any product leadership team: AI saves time on nearly every task measured, but that doesn't translate into enthusiasm. Average sentiment came in at 3.57 out of 5, only slightly above neutral, and 69% of those who chose to leave an open-ended comment at the end of the survey expressed some degree of dissatisfaction.

For people who design products, the interesting data point isn't "does AI work or not." It's where it works, where it doesn't, and why the perception of those who use it remains lukewarm even in the face of real, measured speed gains.

Where AI steps in, and where it doesn't come close to the user

Access to generative tools is now nearly universal: 99% of respondents have access to some form of AI at work, and 66% use at least one agentic tool (an agentic product like Claude Cowork or a coding agent like Cursor). But usage is far from even across tasks.

The pattern is clear: AI speeds up work on artifacts and data, not the work of being with the user. Logging into a test session, conducting an interview, evaluating an interface with human contextual judgment: none of that shrinks much with an assistant. Only 11% of respondents reported using AI interviewers, and the same share said they use synthetic users, a practice that remains marginal even in 2026.

This confirms something researchers already felt firsthand: AI helps prepare an interview script, it doesn't replace listening. The session itself still requires time to schedule, recruit, and conduct, and contextual analysis remains human work.

Agentic AI doesn't reach every company, or every role, equally

Agentic AI use varies both by type of organization and by role within the product team. Startups use it far more than NGOs and government, a gap that exceeds two to one.

NN/G's explanation makes sense: government tends to be more cautious about privacy and security, NGOs run into cost and technical know-how constraints, and startups have fewer approval layers, plus they attract people who already want to work with AI embedded in everything. One respondent told the survey team, in a follow-up interview:

I really wanted to go into a company that will be building stuff from the ground up, where AI is embedded everywhere.

Respondent from the Nielsen Norman Group survey, in a follow-up interview

By role, the pattern is also uneven: product managers and ops professionals (such as DesignOps) use agentic AI more often than content designers and researchers. NN/G suggests that PMs and ops spend their day orchestrating people and data across systems, exactly where an agent delivers the most value today; researchers and content designers produce high-quality artifacts (a research synthesis, interface copy) for which a generic chatbot already does the job.

For anyone leading cross-functional squads, this is a practical warning: setting "use more AI" as a single blanket goal ignores that the right tool changes depending on the role. Holding a researcher to the same agentic adoption rate as a PM means measuring productivity with the wrong yardstick.

The math that doesn't add up: output quality and privacy

Only 9% of respondents said they had no meaningful barrier to adopting AI at work. The two most cited obstacles were the quality of generated output (54%) and data privacy or security (50%), followed by ethical or sustainability concerns (34%).

One respondent who works on niche apps summed up the reliability problem well:

It's been hard to incorporate AI into our workflows because we design specialty apps, and the quality of generating designs from AI is very poor / immature.

Respondent from the Nielsen Norman Group survey

Interestingly, lack of leadership support was the least cited barrier. For many respondents, the problem is the opposite: too much top-down pressure to use the tool, even when it still fails at specific product tasks. This creates what one respondent called "slop" (generated junk) that someone more careful then has to review and fix, doubling the workload for whoever cares about the final result.

Why sentiment stays lukewarm

Here's the point most relevant to product designers: AI shortened every task measured in the survey (average rating of 4.2 out of 5, where 3 is "no difference" and 5 is "much shorter"), but that didn't lift overall mood.

Of the 521 respondents, 226 (43%) left an open-ended comment at the end of the survey, a high response rate for an optional question. Of those comments, 69% carried some negative sentiment, 41% were purely negative, and only 20% were purely positive. The most cited reason for discontent wasn't philosophical: it was unreliable output and the effort of checking and correcting what AI delivers.

AI is incredible because it can make 80000 mistakes in the time that it takes for someone to correct 5 of them. It also makes it easier for businesspeople to create complex slop-flows that can take weeks to months to untangle, that would have taken a designer a month to make well without AI.

Respondent from the Nielsen Norman Group survey

Not every account was negative. Nearly half of the comments (49%) included some positive element, and the most common reason was simply speed:

With AI I can complete work on most days within hours now. This frees up so much more valuable time.

Respondent from the Nielsen Norman Group survey

One detail worth bringing into any team: of the seven respondents who work alone or on very small teams, five left purely positive comments. For those with no one to bounce a design decision off of in the hallway, AI has become a thinking partner, not a substitute for collective judgment. The benefit is real, but it's conditional on who controls how the tool is used, and smaller for those working under a mandate, tight deadlines, or reviewing someone else's output.

Job security: researchers more skeptical than designers

A majority (65%) of respondents expect their role to remain important or gain more responsibility because of AI; only 18% think their job will shrink or disappear. But that confidence isn't equal across functions: researchers were statistically less likely to believe their role will stay relevant and more likely to believe it will shrink, compared to designers. In a direct comparison of purely negative comments, researchers also led (51% versus 33% for designers).

In short: the honest reading of the survey isn't "adopt AI" versus "resist it." It's recognizing that the speed gain is real and uneven: greater on artifact and data tasks, nearly nonexistent on tasks involving flesh-and-blood users, and always paired with the invisible cost of checking, correcting, and absorbing deadline pressure inflated by a tool that still makes mistakes. Whoever leads product decisions gains more by treating this as an explicit trade-off than as an adoption target to be hit.

Source

Nielsen Norman Group: "AI in UX: Work Is Faster, But Practitioners Have Mixed Feelings", by Maria Rosala and Rachel Banawa, October 23, 2026.