This Article is sponsored by
Artificial intelligence was supposed to be the great liberation of the modern worker. At least we came to believe that we would spend fewer hours in front of the computer, and that we would even have more time to think, create, and live.
But a study from the University of California at Berkeley showed just the opposite.
Are we busier than before?
By: Gabriel E. Levy B.
Aruna Ranganathan and Xingqi Maggie Ye, researchers at the Haas School of Business at Berkeley, California, one of the most prestigious universities in the world, spent eight months inside a U.S. technology company of about 200 employees observing, in real time, what happened when workers adopted artificial intelligence tools in their day-to-day lives.
They did more than 40 in-depth interviews with engineers, designers, product managers, and researchers.
The results, published in February 2026 in the Harvard Business Review, were unexpected: instead of working less, people were working more, and with greater intensity.
What the researchers saw
The phenomenon they found has a technical name, workload creep, but its explanation is simple. When an AI tool does something faster that used to take time, it doesn’t use that time gained to rest: it is used to do more things.
The workload does not go down. Grow.
The researchers identified three patterns that repeated themselves consistently. First, the roles began to expand. A designer who previously only designed began to do data analysis. A product manager who didn’t know how to code started writing code. One engineer spent more hours reviewing the AI-generated work of his colleagues. Technology erased boundaries between specialties and pushed each person to cover more ground.
Second, the work leaked out of working hours. Because AI makes it very easy to resume or start a task, employees used it at lunch, in spaces between meetings, in the evenings. The natural boundaries of the day simply disappeared.
Third, permanent multitasking appeared. Workers kept multiple AI processes running at the same time while reviewing code or attending a meeting. The concentration was fragmented. The brain never rested completely.
The result was a cycle that feeds on itself. AI speeds up work, speed expectations go up, tool dependency grows, task scope expands, and total workload ends up being higher. “What appears to be higher productivity in the short term may mask a quiet increase in load and increasing mental strain,” the authors warned.
One worker interviewed summed it up bluntly: he thought that by being more productive with AI, he would save time and be able to work less. The exact opposite happened.
It is not an isolated case
What Berkeley found in that company is no exception. Other research points in the same direction. A survey by freelancing platform Upwork found that 77% of employees who use AI say these tools have added workload to them. 39% spend more time reviewing what AI generates. 71% reported symptoms of burnout.
The Yale budget lab published in November 2025 that, after almost three years since the launch of ChatGPT, the U.S. labor market had not experienced a visible disruption. And an MIT report documented that 95% of generative AI enterprise projects do not produce measurable economic returns, despite investments exceeding $30 billion.
The clash with what was promised
These findings directly contradict the predictions that dominated the public conversation in recent years. In 2023, Goldman Sachs estimated that AI could put 300 million jobs at risk worldwide. McKinsey projected that generative AI would add between $2.6 trillion and $4.4 trillion annually to the global economy.
Dario Amodei himself, CEO of Anthropic, said in 2025 that AI could eliminate half of entry-level jobs in office sectors within five years.
The narrative was clear: automation would come to displace workers.
What is happening, according to the field data, is different. There is no mass displacement. There is more work for the same people.
Is it temporary or is it here to stay?
This is the question that most bothers the academic debate. The Berkeley researchers do not present it as a temporary problem of adaptation.
They warn that, without deliberate management by organizations, the new pace becomes the standard. What today is extra effort tomorrow is simply what is expected.
Economists are divided.
Daron Acemoglu, the 2024 Nobel laureate in economics, argues that AI’s real impact on productivity will be modest over the next ten years and attributes the inflated expectations to venture capital enthusiasm.
Stanford’s Erik Brynjolfsson proposes a more optimistic outlook: Big tech always generates an investment phase before showing its fruits, and argues that there are already signs that this productive phase is beginning.
Ethan Mollick, from Wharton, argues that the real bottleneck is not technological but organizational. AI can do a lot, but companies still don’t quite know how to incorporate it without exhausting their teams.
The question no one was asking
Berkeley’s study shifts the debate into uncomfortable territory.
For years, the discussion revolved around how many jobs would disappear.
But the most pressing question, in light of what the data shows, is another: when AI allows more to be done, who decides how much is enough?
If expectations of speed and volume continue to rise without any counterweight, the promise that technology would make our lives easier ends up becoming its opposite. Not in unemployment, but in workers who are busier, more dispersed and exhausted than before. Productivity is real.
Free time, for now, not so much.
In summary: An eight-month study from the University of Berkeley reveals that artificial intelligence, contrary to what the industry promised, does not reduce the workload, but intensifies it. Workers do more tasks, in more hours and with greater pressure. Research from Yale, MIT, and Upwork points in the same direction, challenging the narrative of mass job displacement that dominated the tech debate in recent years.
References
Ranganathan, A., & Ye, X. M. (2026, February). AI doesn’t reduce work, it intensifies it. Harvard Business Review. https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it
Upwork Inc. (2024). Upwork study finds employee workloads rising despite increased C-suite investment in artificial intelligence. https://investors.upwork.com/news-releases/news-release-details/upwork-study-finds-employee-workloads-rising-despite-increased-c
Yale Budget Lab. (2025, November). Evaluating the impact of AI on the labor market: Current state of affairs. https://budgetlab.yale.edu/research/evaluating-impact-ai-labor-market-current-state-affairs
Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv. https://arxiv.org/abs/2303.10130
Acemoglu, D. (2024). The simple macroeconomics of AI. MIT Economics. https://economics.mit.edu/sites/default/files/2024-05/The%20Simple%20Macroeconomics%20of%20AI.pdf
