Does using more AI tools always boost worker productivity?

Boston Consulting Group surveyed 1,488 full-time US-based workers and found that "the number of AI tools used did not always correlate with increased productivity." Respondents "reported increased productivity when using three or fewer AI tools," but "when they said they used four or more, self-reported productivity plummeted." BCG's researchers and other commentators label the overload point "AI brain fry": workers who reported it showed "34% active intention to leave the company," compared with 25% among those who did not.

The mechanism BCG describes runs through oversight burden rather than tool count alone. Work that required workers to read and interpret LLM output, rather than let an agent complete administrative tasks outright, cost "14% more mental effort," and was linked to "12% greater mental fatigue" and "19% greater information overload." Study author Julie Bedard told Fortune that workers were "getting a lot more done, but also feeling like they were reaching the limits of their brain power, like there were too many decisions to make." BCG found the fry eased when "managers provided training and support," and Bedard frames the fix as redesigning roles and teaching planning and prioritization, not removing AI — companies err by "dumping it on top of an employee's already-established set of responsibilities."

This gives a workplace-survey complement to the mechanism-level findings already in the library. Does AI assistance weaken our brain's ability to think independently? measures a neural cost of offloading; BCG's brain fry is the self-reported, felt-experience counterpart at a much larger and more heterogeneous sample, with turnover risk attached rather than recall deficits. It also sits next to Does AI assistance erode the skills needed to oversee it?: both describe oversight, not delegation, as the costly activity, though BCG's workers are a broad cross-section rather than engineers already skilled at supervising AI. And it extends Does AI really save time, or just change how we spend it? by giving that reallocated time a cognitive price tag — mental effort, fatigue, and information overload — rather than just a time-use shift.

The Fortune article folds in other studies (a Federal Reserve Bank of St. Louis estimate of a 1.1% aggregate productivity gain, a Goldman Sachs analysis finding no economy-wide productivity relationship outside customer service and software development, and a UC Berkeley field study linking AI-driven workload increases to burnout) as surrounding debate, not as replications of BCG's specific finding, and this note does not extend to them. BCG's own numbers are entirely self-reported — productivity, mental effort, fatigue, and overload are all what workers said, not measured output or behavior — and the tool-count and brain-fry comparisons are correlational, with no claim of causal direction given (brain fry could as easily follow from role design or workload as from tool count itself). BCG is a consultancy that sells AI implementation advisory services, and its "redesign roles, don't remove AI" conclusion is also its product pitch, which does not invalidate the survey finding but is a reason to weigh it as an interested party's framing. The honest implication is narrower than "AI causes brain fry": heavy-oversight AI use correlates, in this one sample, with felt cognitive strain and elevated quit intent, enough to warrant attention to how oversight work is designed, not proof that tool count itself is the cause.