AI Burnout Is Real. Designers Scored Second-Worst in Tech.

Nearly every tech worker surveyed says AI makes them better at their job. Not one role would recommend that job to someone starting out.

Both numbers come from Noam Segal’s 2026 Tech Worker Sentiment Survey: roughly 6,000 people across product, engineering, design, research, marketing, data, and sales. 97.2% said AI makes them at least moderately better at what they do. Then every role scored negative on whether they’d recommend it to someone entering the industry. Design came in second-worst.

I recognized my own last quarter in that gap. I shipped more work than I ever have, and somewhere in the middle of it I stopped having opinions about what I was making. Hand off a flow, get the approval, open the next file.

Segal’s survey is one of the largest of its kind, and its most useful finding barely made the coverage.

Losing your job to AI ranked second to last on the list of things tech workers are worried about.

AI burnout is the exhaustion that arrives when AI adoption expands your workload instead of shrinking it, because the expectations attached to the tools grow faster than the tools do. The output goes up and the depletion goes up with it.

Segal’s own one-line version of it:

What sits at the top of the worry list is the expectation to do more for the same pay, and a work pace people described as unsustainable. That is a different problem with a different shape, and almost nobody is naming it correctly.

The fear everyone assumes is the wrong fear

Start with the number that made me stop scrolling. Significant burnout in Segal’s survey went from 44.7% to 55.7% in a single year. Eleven points. Career optimism went the other direction over the same stretch, from 54.8% down to 48.7%, which means fewer than half the people in this industry currently believe their career is heading somewhere good.

Now put that next to the finding nobody expects: 97.2% of those same people said AI makes them at least moderately better at their job. Nearly half said very much or extremely better.

Read those two facts together and the story changes completely. These are people watching a tool work exactly as advertised, then getting handed the surplus. The complaint in that data is about everything that arrived after the tool started working.

Call it compression. The job stayed exactly where it was and got denser. Density is a much harder thing to complain about at work, because on paper you’re doing great. Your velocity is up. Your ticket count is up. There is no line on any dashboard for the fact that you’ve stopped trusting your own read on the work.

I wrote a whole piece a few months back on whether AI is coming for our jobs, and I still stand by the answer. But I was arguing the wrong axis. The question worth asking is what the job turns into once the tools are inside it.

Designers scored second-worst in the entire industry

Here is where it stops being a general tech story and starts being ours specifically.

Segal scored every role on whether people would recommend their own job to someone entering the industry today. Standard net promoter setup. Every single role came back negative. Sales, product, engineering, founders, all of them underwater on the same question.

Design came in at roughly −35. Research came in at roughly −40. Those are the two lowest scores of any function measured.

The identity numbers from the same survey say something similar from another angle. Only 3% of respondents reported no shift at all in their professional identity because of AI. Everyone else landed somewhere: 49% amplified, 27% redefined, 14% destabilized, 5% diminished. Roughly half of this industry feels bigger because of these tools and roughly half feels blurrier. The split runs straight through job titles, which is why it shows up inside teams as much as between them.

Nielsen Norman Group’s State of UX 2026 named the mechanism without naming the burnout. Their line: organizations will “ask more of each role, compressing responsibilities that were once spread across multiple specialists.”

That’s the whole thing in one sentence, from a design research org, published independently of the sentiment survey. Two different groups looking at two different datasets, describing the same squeeze.

Design was already the function most likely to be understaffed relative to engineering. Compressing responsibilities across a team of three lands differently than compressing across a team of thirty. And it stacks on top of a hiring market where entry-level design roles have already thinned out badly, which means the person absorbing the compressed scope is usually the person who would have handed part of it to a junior last year.

Nielsen Norman Group’s prescription for all of this is to go deeper, and they’re blunt about what happens if you don’t.

I don’t disagree with any of that. I’d just point out that it sits in the same report as the compression finding, and depth is the first thing compression takes. Going deeper requires slack, and the compressed version of the job is the version with the slack removed. That’s the bind design is in, and it’s the most plausible read I have on why our number sits second from the bottom.

What compression looks like on an actual Tuesday

None of that is abstract. Here is what it looks like in a normal week.

It’s producing three concept directions in the time you used to produce one, and presenting all three because you can, and nobody in the room having a strong opinion about any of them because none of them cost anything. It’s a research synthesis that took four hours instead of three days and reads fine, and you can’t shake the feeling that you skipped the part where you actually understood it. It’s writing the spec faster than the team can absorb it.

The output is real. The gap is that judgment moves at its own speed no matter how fast production gets, and the calendar only budgets for production.

Upwork’s research put numbers on that gap. 96% of C-suite leaders expected AI to boost productivity. Meanwhile 77% of employees using AI said the tools had added to their workload, and 47% said they didn’t know how to deliver the productivity gains their leadership expected of them. That last number is the one I’d put on a wall. Half the people using these tools have been handed a target with no mechanism attached.

Kelly Monahan, managing director of the Upwork Research Institute, named the cause: new technology dropped into “outdated work models and systems,” where it can’t deliver the productivity value leadership expects of it. The tools changed and the scaffolding around them stayed put. Individual people are absorbing the difference.

Segal’s open-ended responses got at the qualitative version, and it’s bleaker than the charts.

That’s his summary of what people wrote in unprompted, in a survey that never asked about it. Cognitive rot came up often enough to be a theme.

I recognized that immediately. My throughput went up and my confidence in my own taste went down. Those moved at the same time, in the same direction, and I spent a quarter assuming that was a personal failing.

It isn’t design-specific either. The comments under Segal’s podcast episode filled up with people naming the same thing from other seats. One senior product manager wrote that what landed for them was “how management and ICs experience AI,” and finished the comment with “I feel very squeezed.” That’s a job with barely any overlap with mine, describing the identical shape.

Separate this from the two things it gets confused with. Pulling back from a job that hasn’t earned more from you is a deliberate scope decision you make on purpose. Compression runs the other way. Scope expands underneath you while nobody updates the description of your job, and you keep saying yes because each individual increment looks trivial. The grind of applying into a market that won’t answer at least has an end date. Compression is just Tuesday.

Reflective Coda

The most useful finding in the survey has nothing to do with AI. Only 25% of respondents rated their manager as highly effective, while 36% rated theirs as ineffective, and Segal’s read on the data is that managers are the single biggest lever on well-being in the whole dataset. Which puts the fix for this in a conversation with a person, well upstream of any tool or workflow.

So name the thing accurately before you take it anywhere. If you’re afraid AI is going to take your job, that’s a skills-and-positioning problem and it has a well-documented set of moves. If you’re producing more than you ever have and quietly losing your read on whether it’s good, that’s a scope problem, and no amount of upskilling touches it. The two feel similar from the inside. The fixes have nothing in common.

The scope version has to be raised as scope, in the language your manager already uses for scope. Framing it as burnout turns it into a wellness conversation and hands the problem straight back to you. Your role absorbed work that used to belong to other people, the tools made that invisible on every dashboard your manager looks at, and someone has to say that out loud. The 55.7% figure exists because most people don’t.

I still ship more than I used to. I’ve stopped treating the throughput as evidence that anything is working.

What I do differently now is small and unglamorous. I pick one thing per cycle that I refuse to accelerate, and I sit with it at the old speed, badly, until I have an opinion about it I can defend. That single deliberate exception is the only reliable way I’ve found to tell whether the rest of the work is any good. Speed is the only part of this job the tools actually changed. Pretending they changed judgment too is what’s burning people out.

Sources

• How tech workers actually feel about AI in 2026: Annual AI sentiment survey (Noam Segal, Lenny’s Newsletter, Jul 2026)
• Why the AI honeymoon is ending (and tech workers are feeling it) (Noam Segal, Lenny’s Podcast, Jul 2026)
• Noam Segal survey: No tech job earns a positive recommendation score in 2026 (BigGo Finance, Jul 2026)
• Tech Workforce Splits: Burnout Surges, Optimism Fades Amid AI Boom (StartupHub.ai, Jul 2026)
• State of UX 2026: Design Deeper to Differentiate (Kate Moran, Raluca Budiu & Sarah Gibbons, Nielsen Norman Group, 2026)
• Upwork Study Finds Employee Workloads Rising Despite Increased C-Suite Investment in Artificial Intelligence (Upwork Research Institute, Jul 2024)
• 77% Of Employees Report AI Has Increased Workloads And Hampered Productivity, Study Finds (Forbes, Jul 2024)