Take Francesco Bonacci, a software engineer and founder of Cua AI, who warned of “vibe coding paralysis” last month. In an X post, he described AI’s ability to complete incredible taskloads, leaving workers time to generate new ideas they can then give to bots to flesh out. But the result was not an empowered, productive employee. Rather, it was a mountain of half-finished projects and a human too overwhelmed to complete or make sense of any of it.

“The paradox: the more capability you have, the more you feel compelled to use it. The more you use it, the more fragmented your attention becomes. The more fragmented your attention, the less you actually ship,” Bonacci wrote.

Steve Yegge, a longtime blogger about computer programming, called this the “AI vampire” in a Medium essay, arguing that it was a “concerning new phenomenon.” He likened AI’s tendency to encourage human overwork to Colin Robinson, an “energy vampire” from the FX TV series What We Do In The Shadows, who thrived off the enervation of human beings. Boston Consulting Group has another phrase for this. 

Future-of-work experts call it “AI brain fry” and warn the excessive oversight of AI tools could overwhelm employees at the expense of workplace productivity. A study conducted by Boston Consulting Group found in a survey of 1,488 full-time U.S.-based workers, the number of AI tools used did not always correlate with increased productivity. While respondents reported increased productivity when using three or fewer AI tools, when they said they used four or more, self-reported productivity plummeted. 

The researchers indicated AI brain fry could lose companies valuable talent and cost them millions of dollars. They cited a 2018 report from Gartner, which found suboptimal decision-making at a $5 billion revenue firm cost it $150 million per year. Moreover, the study found that among workers who reported AI brain fry, 34% showed active intention to leave the company (ie, quit). That’s compared to 25% among those who did not report AI brain fry.

The AI productivity debate

The surveyed workers said that when their AI-related work required higher levels of oversight—reading through and interpreting text a large language model generated versus an AI agent completing administrative tasks, for example—they expended 14% more mental effort at work. High AI oversight was also associated with 12% greater mental fatigue and 19% greater information overload.

Many respondents reported a “fog” or “buzzing” associated with overuse of AI that required them to physically step away from their computers. Others said the number of small mistakes they made increased as a result of feeling this brain fry.

“People were using the tool and 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,” Julie Bedard, study author and managing director and partner at Boston Consulting Group, told Fortune. “Things were moving too fast, and they didn’t have the cognitive ability to process all the information and make all the decisions.”

Sweeping claims about AI’s productivity-increasing capabilities combined with inconclusive data has sparked widespread debate on the efficacy of using the technology in the workplace. Erik Meijer, a former senior engineering leader at Meta, recently marvelled that Anthropic’s Claude Code has “pushed the state of the art in software engineering further than 75 years of academic research” in just a matter of months.

A February 2025 report from the Federal Reserve Bank of St. Louis estimated a 1.1% increase in aggregate productivity as a result of using generative in the AI workplace—a boost that translated to workers becoming 33% more productive each hour they use the tool.

An analysis from Goldman Sachs this month, however, found no “meaningful relationship between productivity and AI adoption at the economy-wide level,” but rather was effective in just two specific use cases: customer service and software development tasks. That report follows a survey of 6,000 of C-suite executives, 90% of whom found no evidence of AI impacting productivity or employment in their workplaces in the past three years. Those executives forecast AI would increase productivity by 1.4% in the next three years.

Reported increases in workplace productivity as a result of AI also seem to come with a cost. An eight-month study of a 200-person U.S. tech firm led by researchers at the University of California at Berkeley found that AI tools were able to increase an employee’s workload, which subsequently led to more burnout and overall acted as a drag on workplace efficiency in the long run.

The researchers concluded that AI actually intensifying work, as opposed to freeing up more time and mental space. Employees are processing more information and have less of a boundary between work and non-work. AI really is something like a vampire or a fryer, in other words—it won’t do the work for you, but force you to use your brain a lot more than you’re used to.

“Those are all real costs,” Bedard said. “Companies [say], ‘We want fewer errors, we want better decisions, and we want our best people to stay.’”

Redesigning the workplace in the age of AI

Bedard noted the answer to resolving AI brain fry is not to eliminate AI at work, but to think critically about how it is being implemented. Too many companies are introducing the technology in the workplace by dumping it on top of an employee’s already-established set of responsibilities. Instead, workplace leaders should instead redesign roles and give employees training on planning and prioritization skills, she said.

The study found when managers provided training and support on using AI tools, brain fry decreased. The Berkeley researchers suggested the antidote to AI brain fry is to batch activities requiring AI tools to a specific block of the work day. They said employees should build in times to take a step back from their work, in particular ahead of a challenging decision or demanding task. In other words, AI is such a powerful tool, you need to step back from it to catch your breath. Vampires can fly, after all, but you need to clip their wings at times.

“One really hopeful message for leaders and managers is, You have a really important role to play here in rethinking what work looks like in a world of AI,” Bedard said.