82% of Tech Workers Say AI Makes Them More Productive. Burnout Jumped 11 Points Anyway.

Getting faster at your job and burning out are not opposites. According to Lenny Rachitsky's survey of ~6,000 tech workers, published July 7, 82% say AI makes them more productive. Burnout increased from 44.7% to 55.7% year-over-year. The gap between those two numbers is where a lot of people are quietly living right now.

What the survey found

Rachitsky's data splits respondents into four clusters: 41% feel energised by AI, 35% feel conflicted, 12% feel disoriented, and 12% feel resentful. What unifies the bottom three groups is not their opinion of the tools. It's a specific fear. Job loss came in at 22%. The most common fear, at 51%, was being expected to do more work for the same pay.

The productivity gains are real — the survey confirms them. But they're landing in workplaces that are treating increased output capacity as a reason to raise baseline expectations, not as slack to return to the worker. That's not a theory about how AI should work. It's what 55.7% of respondents are experiencing while simultaneously reporting they're getting more done.

What it means on Monday morning

Supervision is usually framed as a quality question: how closely should you check AI output before it goes out? But there's a second version of that question that most guidance skips entirely: who decides how the time you've recovered from AI tasks gets used?

If your manager or client can see that a task that used to take three hours now takes forty-five minutes, one of two things happens. Either that time gets returned to you — less overtime, more room for harder problems, clearer thinking before decisions. Or the slot fills immediately with a new request, and the implicit baseline for your daily output quietly shifts upward. Most people in the Rachitsky data never made an explicit choice between those two outcomes. The workload just expanded to fill the space.

The supervision skill here is the same as it is with output quality: make the decision explicitly before it gets made for you.

Elena: two years of efficiency, one outcome she didn't plan for

Elena is a content strategist at a 65-person B2B software company. Over the past 18 months, AI has roughly doubled what she produces: first drafts, research summaries, client briefs, campaign frameworks. She's rated a high performer. Her deliverables list doubled with the rating.

She would not have described herself as burned out eighteen months ago. She would now.

Elena is not missing a better tool. She's missing the conversation about what the productivity gain was supposed to be for. Her AI adoption made her more valuable to her employer. What she didn't set was any expectation about what she'd get in return.

Maya: same tools, different outcome

Maya is an operations manager at a 130-person healthcare staffing company. She's in the 41% energised group. She got there by having a specific conversation with her director before rolling out AI for reporting tasks: she expected to recover four to six hours per week, and she wanted half of that reserved for deeper client analysis rather than additional reports. They agreed on it in writing. Three months later, she's producing better work for fewer clients, and the improvement is visible to both of them.

The difference between Elena and Maya is not the AI tool. It's that Maya made the workload question explicit before her director made it for her.

The one-sentence version

Getting better at AI without deciding what to do with the recovered time is how a productivity tool becomes a faster hamster wheel.

<BlogPracticeSection />