The excessive use and supervision of AI in the workplace makes workers “AI-scarious,” which contradicts the technology’s assertions that it will relieve work stress.

Workers using artificial intelligence tools report that the technology is “condensing rather than simplifying work,” researchers from Boston Consulting Group and the University of California wrote in the Harvard Business Review on Friday.

A study of nearly 1,500 full-time U.S. workers found that 14% said they experienced “mental fatigue resulting from excessive use, interaction with, and/or supervision of AI tools beyond one’s cognitive capacity,” or what researchers called “AI brain fry.”

Participants described having a “mental hangover” with “fog” or “buzzing” and an inability to think clearly, along with headaches, slow decision-making, and difficulty concentrating.

AI companies have pushed their products as a way to boost productivity, allowing workers to offload some or part of their workloads, a message that some companies have internalized and begun measuring the use of AI as a measure of performance.

Brian Armstrong, CEO of cryptocurrency exchange Coinbase, said he fired engineers who didn’t want to use AI, and set a goal late last year to have AI create half of the platform’s code.

“As companies use more multi-agent systems, employees find themselves switching between more tools,” the researchers wrote. “In contrast to the promise of more time to focus on meaningful work, maneuverability and multitasking could become the ultimate hallmarks of working with AI.”

AI involves “significant costs,” but can improve fatigue

This AI-induced mental stress “carries significant costs in the form of increased employee errors, decision fatigue, and intention to quit,” the researchers said.

Study participants who said they experienced brain fluctuations experienced 33% less decision fatigue than those who did not, which researchers said could cost large companies millions of dollars annually. Those with AI were approximately 40% more likely to have an active intention to quit smoking.

Those who reported AI fluctuations also reported making major errors about 40% more frequently than those who did not, with a major error defined as an error that had “serious consequences, such as those that could affect safety, results, or important decisions.”

However, researchers have found that using AI to replace repetitive and routine tasks reduces burnout, a state of chronic workplace stress that leads to negative feelings about the job and decreased effectiveness.

Related to: Anthropics reopens Pentagon talks as tech groups pressure Trump to drop risk tag

Participants who used AI to reduce time spent on routine and repetitive tasks reported that their burnout levels were 15% lower than those who did not use AI in this way.

Business leaders looking to reduce AI volatility should “clearly define the purpose of AI in the organization” and explain how workloads will change with the tool, the researchers said.