The pitch was simple: AI handles the grunt work, you go home early. Eight months of observation at one company says the opposite happened.
Aruna Ranganathan and Xingqi Maggie Ye, researchers at UC Berkeley's Haas School of Business, spent two days a week inside a 200-person U.S. tech firm from April through December 2025. They sat in on meetings, tracked internal communications, and conducted 40 in-depth interviews across engineering, product, design, research, and operations. The company offered enterprise AI subscriptions but didn't mandate use. Adoption was voluntary. The results were not what the brochure promised.
Nobody worked less. The AI made tasks faster, which made more tasks feel possible, which made the to-do list grow until it consumed every minute the AI had freed up — and then kept going.
The Three Ways It Got Worse
The researchers identified three mechanisms driving what they call "workload creep."
First, task expansion. Product managers started writing code. Researchers took on engineering work. Not because anyone asked them to — because AI made unfamiliar work feel newly accessible. The boundaries of each job widened until every role contained pieces of every other role.
Second, the death of downtime. Workers filled loading screens, lunch breaks, and meeting transitions with "quick prompts." The natural pauses that once separated tasks — the moments when the brain switches context or simply rests — got colonized by work that now felt too easy to skip.
Third, chronic multitasking. With AI handling parts of multiple threads simultaneously, workers managed more concurrent tasks than before. The cognitive load didn't decrease. It redistributed across more surfaces.
One engineer summarized the experience to the researchers: "You had thought that maybe, 'Oh, because you could be more productive with AI, then you save some time, you can work less.' But then really, you don't work less. You just work the same amount or even more."
The Burnout Gradient
By month six, the study found reports of burnout, anxiety, and decision paralysis had spiked. The people burning out fastest weren't the skeptics who avoided AI. They were the power users who embraced it most aggressively.
The pattern extends beyond one company. A LeadDev survey found 22% of developers at critical burnout levels, with nearly a quarter more moderately burned out. Across industries, digital exhaustion has hit 84% of workers, and 77% report unmanageable workloads — even as 70% now use AI at least weekly. Thirty-eight percent of employees say they feel overwhelmed about having to use AI at work.
The correlation is hard to ignore: the people doing the most with AI are the ones closest to breaking.
The Macro Picture Is Worse
Here's where the individual experience meets the aggregate data. A National Bureau of Economic Research study surveying 6,000 executives across the U.S., U.K., Germany, and Australia found that nearly 90% of firms reported no measurable impact from AI on employment or productivity over the past three years.
The average executive uses AI 1.5 hours per week. A quarter don't use it at all. The predicted productivity gain over the next three years: 1.4%.
Apollo's chief economist Torsten Slok put it bluntly: "AI is everywhere except in the incoming macroeconomic data." He was echoing Robert Solow's famous 1987 observation about computers: "You can see the computer age everywhere but in the productivity statistics." MIT economist Daron Acemoglu called the numbers "just disappointing relative to the promises that people in the industry are making."
So at the individual level, AI makes people work more. At the company level, it doesn't show up in productivity. Both things can be true if the extra work is low-quality busywork that expanded to fill the time AI created.
The Mechanism Nobody Talks About
The Berkeley researchers identified something that doesn't appear in any vendor's pitch deck: AI doesn't just automate tasks. It changes the worker's relationship to work itself.
When a task takes thirty seconds instead of thirty minutes, the psychological barrier to starting it vanishes. That sounds like a feature. But barriers serve a function. They force prioritization. They create natural stopping points. They give people a reason to say "that's not my job." Remove the friction and every possible task becomes an obligation.
The researchers recommend "intentional pauses" and "sequencing workflow" — essentially rebuilding by hand the friction that AI removed. Which raises a question nobody in the industry wants to answer: if the solution to AI-driven burnout is deliberately slowing down, what exactly did the AI accomplish?
Rebecca Silverstein, a licensed clinical social worker at Elevate Point, was more direct: "Just focusing on that productivity mindset, in the long term, is super harmful."
The companies selling AI promise more output with less effort. The data says they're delivering more effort with ambiguous output. The workers absorbing the difference didn't sign up for either version.
Top comments (0)