“It was like I had a dozen browser tabs open in my head, all fighting for attention.” That line, from a senior engineering manager in a recent workplace study, captures a problem that arrived wrapped in the language of efficiency. Generative AI was supposed to remove drudgery. In practice, for many office workers, it has added a layer of supervision, correction, prompting, and second-guessing that looks less like delegation and more like managing a room full of fast, error-prone interns. Researchers have started calling the result “AI brain fry”: mental fatigue caused by using or overseeing AI tools beyond a worker’s cognitive limits.
The pattern emerged in a survey of about 1,500 workers and was published in Harvard Business Review. Workers who had to monitor several AI systems, interpret output, and keep switching contexts reported more mental effort, more information overload, and more mistakes. One finding stood out: employees using four or more AI tools did not report steadily rising productivity. The gains flattened, then fell.
This is not the same problem as simple burnout. Researchers described brain fry as acute rather than chronic: a buzzing, foggy state that can ease after a break. That distinction matters, because it helps explain why AI can feel both useful and draining in the same workday. When the software actually removes repetitive tasks, stress can drop. When it multiplies decisions, handoffs, and review work, strain rises quickly.
The most revealing part of the research is not that AI creates fatigue. It is why. High-oversight AI work required 14% more mental effort, according to the study, and it also predicted more fatigue. The burden comes from fragmented attention: waiting for one tool, checking another, refining prompts, reviewing outputs, and deciding whether the machine is confidently wrong or merely incomplete. Francesco Bonacci, founder of Cua AI, described the sensation as “vibe-coding paralysis,” writing, “I end each day exhausted not from the work itself, but from the managing of the work.” The hidden cost is not only time. It is the constant demand to stay alert across too many semi-automated threads at once.
There is a deeper concern beneath the exhaustion. Some AI systems reduce the mental work that helps people retain information in the first place. An MIT-linked study described by Time found weaker memory integration among people who relied on ChatGPT for writing tasks, compared with participants working without it. In a workplace, that raises an uncomfortable possibility: tools marketed as cognitive support may sometimes offload the very effort that builds judgment and recall.
Not every team is equally exposed. One report on the survey found self-reported productivity dropped when workers moved beyond a small number of tools, and marketing workers showed especially high rates of AI-related fatigue. That makes sense. Jobs built around rapid iteration, content review, and constant switching are especially vulnerable to overload disguised as speed.
The practical lesson is less about rejecting AI than redesigning work around it. Researchers found that training, clearer boundaries, and more deliberate managerial support reduced brain fry. The technology can compress tasks, but it also expands what workers feel responsible for. That is the tradeoff many organizations are only now beginning to see clearly.
