The most attractive narrative in American work culture today is not that AI will take your job. This AI will save you from it.

That’s the version the industry has spent the last three years selling to millions of nervous people eager to buy it. Yes, some white jobs will disappear. But for most other roles, the argument goes, AI is a force multiplier. You will become a more capable, more in-demand lawyer, consultant, writer, coder, financial analyst — and more. The tools work for you, you work less, everyone wins.

But a new study published in the Harvard Business Review follows that premise to its actual conclusion, and what it finds is that there is no revolution in productivity. It knows that companies are at risk of becoming burnout machines.

As part of what they described as “in-progress research,” UC Berkeley researchers spent eight months inside a 200-person tech company looking at what happened when workers truly embraced AI. What they found in over 40 “in-depth” interviews is that no one is forced into this company. No one is told to hit new targets. People just started doing more because the tools made it more possible. But since they can do these things, the work starts to bleed into lunch breaks and nights. Employees’ to-do lists expand to fill every hour the AI ​​frees up, and then continues.

As one engineer said, “You think maybe, oh, because you can be more productive with AI, then you can save time, you can do less work.

Over on the tech industry forum Hacker News, one commenter has same reactionwrote, “I feel it. Since my team jumped to an AI in all working styles, expectations tripled, stress tripled and actual productivity only increased by 10%. It seems like leadership is putting a lot of pressure on everyone to prove that their investment in AI is worth it and we all feel the pressure to try to show them that it’s true that we have to work longer hours to it can be done.

Fun and alarming too. The argument about AI and work often stops at the same question – are the gains real? But few stop to ask what would happen if they did.

The researchers’ new findings are not entirely new. A separate test last summer found experienced developers using AI tools 19% of the tasks while assuming they are 20% faster. At the same time, a study by the National Bureau of Economic Research that tracked the adoption of AI in thousands of workplaces found that productivity gains amounted to Only 3% time savingswithout significant impact on income or working hours in any job. Both studies are divided.

This one will be harder to dismiss because it doesn’t challenge the premise that AI can augment what employees can do on their own. It confirms this, then shows where all the enlargement actually leads, which is “fatigue, burnout, and a growing feeling that the job is harder to leave, especially when the organization’s expectations for speed and responsiveness increase,” according to the researchers.

The industry’s bet is that helping people do more is the answer to everything. This could be the beginning of a different problem entirely.