Where does the term brain fry come from?
From “When Using AI Leads to ‘Brain Fry’” by Julie Bedard, Matthew Kropp and colleagues at BCG (Harvard Business Review, 5 March 2026, read 5 November 2026), based on a survey of 1,488 full-time US workers (Fortune, 10 March 2026, read 5 November 2026). By October, Deloitte Insights was using it as settled vocabulary, describing what happens “when workers are attempting to work beyond their own cognitive limits on an ongoing basis” (Deloitte Insights, 15 October 2026, read 5 November 2026). Brain fry is an oversight problem, not a usage problem: the strain comes from supervising AI, not from having it.
What causes AI brain fry?
Two things in the BCG data. The first is heavy oversight, such as reading and checking text a model generated: workers with high AI oversight reported 14% more mental effort, 12% greater mental fatigue and 19% greater information overload, in contrast to those who let an agent take over administrative tasks. The second is tool count: self-reported productivity was higher with three or fewer AI tools and dropped sharply at four or more (Fortune, read 5 November 2026). Every extra agent or tool is another stream of output someone has to judge. That is why agent sprawl is a people problem as much as a governance one.
Brain fry vs workslop vs deskilling: how do they differ?
| Brain fry | Workslop | Deskilling | |
|---|---|---|---|
| Who pays | The person reviewing AI output | The person receiving AI output | The person whose skill fades |
| Cause | Too much oversight, too many tools | Careless AI use by the sender | Too little practice of the core task |
| Shows up as | Fatigue, errors, intent to quit | Rework, lost trust | Weak judgment on hard cases |
| Time scale | Days and weeks | Each handoff | Months and years |
They compound. Workslop adds to the reviewing load, and a fried reviewer is the weakest point of any human in the loop control. In the study, 34% of workers reporting brain fry intended to leave, against 25% of those who did not, so the cost also lands in retention, while deskilling erodes the expertise review depends on.
How do managers reduce brain fry?
Start with support: the study found brain fry decreased where managers provided training and support on AI tools. Then design the load. Cap the number of AI tools per role, batch review into set blocks of the day, and spend human attention by risk, with confidence gating letting routine output pass and routing only uncertain or consequential items to a person. Deloitte also suggests making AI reasoning visible, with sources and flagged assumptions, so reviewers do not rebuild how an output was produced. Measure review hours, not just output produced, because an AI adoption plan that ignores the reviewer’s capacity just moves the bottleneck into someone’s head.