Why does using AI leave you more exhausted, not less?

"AI was supposed to give me time back, so why do I finish the day more wrung out than before we had it?" You closed the laptop tonight having shipped three things in the time one used to take. And you feel flattened. Not the clean tiredness of a hard day's making, but a buzzing static behind the eyes, a fog you cannot quite think your way clear.

Here is the short answer, the one to hold while you read the rest. The tool is working. Your nervous system is the part that is overdrawn. This is what researchers have now named AI brain fry, and it is not a character flaw or a sign you are slow. It is what happens when the very cognitive system AI was meant to relieve becomes the system AI now leans on hardest, all day, in two-minute increments, until there is little capacity left to lead.

Why do I feel more fried after a productive AI day, not less?

Because the work changed shape, and the new shape is heavier on the part of you that gets tired. In a 2026 study by Boston Consulting Group and UC Riverside, 14% of 1,488 full-time workers reported AI brain fry, the mental fatigue that results from excessive use, interaction, and oversight of AI tools beyond one's cognitive capacity. Among marketing workers it reached 26%. Participants described a buzzing feeling, a mental fog, slower decisions, headaches. Some had to physically step back from the screen.

That is the contradiction in one line: the output arrives, and the capacity to judge it arrives a day late. The work feels done. The mind that should weigh it has nothing left in the tank.

What is actually loading my brain when AI is doing the work?

Oversight is the answer, and oversight is cognitively expensive. When you delegate to a person you trust, you hand off the thinking. When you delegate to a model, you keep the thinking and add a checking task on top. The same study found high-oversight AI work was associated with 14% more mental effort, 12% greater mental fatigue, and 19% greater information overload. The vigilance is rational, not neurotic: in Stack Overflow's 2025 Developer Survey (an annual poll of the working software industry, 49,009 responses), 84% of developers now use AI tools, yet only 33% trust their accuracy and 46% actively distrust it; 45% said debugging AI-generated code costs them more time, not less.

Then there is the juggling. Productivity in the data rose for people using three or fewer AI tools and fell once they were running four or more at once: more tools, more fragmentation, less shipped. This sits on top of a workplace that was already saturated. Microsoft's 2025 Work Trend Index, drawn from 31,000 knowledge workers, found the average person interrupted by a meeting, email, or message roughly every two minutes, about 275 times a day, while 80% of the global workforce say they lack the time or energy for the work in front of them. AI did not create that. It poured onto it.

So the picture resolves. You are not failing at AI. You are succeeding at it while paying a cognitive tax nobody quoted you, and the tax is levied on the one faculty a leader cannot afford to run short: clear judgement.

What kind of fix actually addresses this, when more tooling makes it worse?

The category of solution here is not another app. It is capacity. There is a quieter cost underneath the fatigue, too. A 2025 study from Microsoft Research and Carnegie Mellon University (a leading US computer-science institution), covering 319 knowledge workers and 936 real workplace uses of generative AI, found that the more a worker trusted the AI's output, the less critical thinking they applied to it, which is the opposite of the careful verification most AI policies assume. The faster the work feels, the less of you is actually in it. BCG's own research on scaling AI found that roughly 70% of the value comes from people and process change, 20% from technology and data, and only 10% from the algorithms themselves. The bottleneck is no longer the technology. It is the human operating the technology, and that operator runs on a body.

This is where the physiology earns its place. A 2019 systematic review of 20 studies covering 19,431 participants, published in Frontiers in Neuroscience, found that higher resting heart rate variability (HRV, the natural beat-to-beat variation in your pulse, a reliable marker of how well the nervous system regulates itself) tracks with better performance on cognitive tasks, with executive functions showing the strongest link, while lower HRV maps to weaker control from the prefrontal cortex (the front of the brain that handles planning and judgement). The neurovisceral integration model, set out by the psychophysiologist Julian Thayer and colleagues, frames HRV as a window onto the same brain networks that govern goal-directed behaviour and self-regulation. Read plainly: a regulated nervous system is the hardware that carries your judgement. Treat this as a working hypothesis worth testing in your own day rather than a settled prescription, and you have a lever the tool-buying conversation keeps missing.

Most leaders are trying to install new software on broken hardware. The recovery is sequential, and it is yours to run.

1. Cap the simultaneous tools. Hold to three at once. The data shows the fourth is where your return turns negative, so the discipline is addition by subtraction.
2. Batch the oversight. Group your checking into defined windows rather than reviewing in the gaps between everything else, so the vigilance has edges instead of bleeding across the whole day.
3. Protect one unbroken block. Reclaim a single stretch free of pings for the work that needs your full prefrontal engagement: the synthesis and the decisions only you can make.
4. Regulate first, then decide. Build a short practice that settles the nervous system, breath-led and physiological, before the high-stakes judgement, so you meet it with capacity rather than fumes.

The stakes are worth naming, because the drain does not stay private. In the same study, intent to quit rose from 25% among workers without brain-fry symptoms to 34% among those carrying them. An exhausted operator becomes an exhausted organisation. The leaders who win the next decade are the ones who upgrade themselves first, and that upgrade begins with the body that carries the mind. This is the move from one era of work to the next. Phase One, the Age of Effort: work hard, get a little more, linear growth. Phase Two, the Age of Scale: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration: output decoupled from human effort almost entirely, the phase AI unlocks. Phase One was muscle. Phase Two was machine. Phase Three is mind, and a fried mind leads nothing. Restore the capacity, and the tool finally keeps its promise.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.