AI Brain Fry: Why Your AI Tools May Be Quietly Fuelling High-Functioning Burnout

A student described it to me last year (this was before everyone had an AI assistant in their workflow) as "the exhaustion of doing nothing." She'd had what looked like a light afternoon: emails, a bit of research, some writing. No meetings, no calls. And she was completely hollowed out. "I didn't do anything hard," she said. "I just made a lot of small decisions."

That description has come back to me more than once this year. Because what she was describing (before we had a name for it) is now arriving with a precise mechanism and a new face.

A 2026 Harvard Business Review study found that 88% of the most productive AI-enabled workers also reported burnout, and were twice as likely to be considering leaving. A UC Berkeley longitudinal study tracked workers over six months: productivity rose in the first quarter, and by the third, the same workers were reporting decision paralysis, cognitive fatigue, and quiet plans to exit.

The culprit wasn't hours. It was decisions.

The load that doesn't show up on the calendar

There's a term circulating in workplace wellbeing research right now: "AI brain fry." It captures something clinical language doesn't quite get at: a specific kind of wrung-out feeling that arrives not after a hard day, but after a day of managing AI-assisted work.

Reviewing outputs for errors. Deciding which suggestions to act on. Maintaining mental context across multiple threads. Holding the low-level anxiety of watching parts of your role quietly shift. None of that appears on a to-do list. All of it draws from the same reserve.

That reserve has a name: cognitive load. And the research is fairly clear that cognitive load doesn't care whether it came from a spreadsheet, a deadline, or a language model. The system keeps score the same way.

For people who adopted AI tools early and competently, the exhaustion has an extra layer. These tools were supposed to make you better. Admitting they're draining feels like admitting something went wrong. Add to that: the Fair Play Talks UK survey (May 2026) found seven in ten workers fear AI-related redundancy. That's not background noise. A workplace where job security feels contingent on how well you use the very tools that are draining you is a specific kind of pressure.

Why this is a high-functioning burnout problem

High-functioning burnout has always been the kind that sneaks past the warning lights. The output stays high. The standards stay high. Externally, nothing looks wrong. Which is exactly why it goes unaddressed until the system has been running on fumes for longer than anyone would like to admit.

AI-era cognitive load runs the same pattern, accelerated. You're still delivering. The work is technically getting done. The AI might even be making you measurably more productive. But productivity is not the same as capacity. They're measured differently. They recover differently. And when you've been treating productivity as the proxy for your own wellbeing, a dashboard that says you're fine becomes the thing that keeps you from noticing you're not.

The signals tend to be low-grade and easy to dismiss:

Finishing a day of AI-assisted work feeling hollowed out rather than tired in any satisfying way. Small decisions that once felt automatic now requiring actual effort. Reviewing AI outputs more anxiously than you used to — second-guessing corrections that would once have been instinct. Sleep that's technically adequate but leaves you unrefreshed and already behind. Weekends that don't clear the fog.

None of these are dramatic. None of them would make you call a doctor. That's the point.

What the nervous system is actually doing

The mechanism here is worth naming, because the standard advice misses it.

Cognitive load fatigue isn't primarily physical: it's about the brain's capacity to make decisions and hold context. Recovery that targets the wrong system (an early night, a screen-free weekend) doesn't reach what's actually depleted. The system that runs decision-making, error-monitoring, and context management is the prefrontal cortex, and it runs on a particular kind of fuel: absence of demand.

Not rest-as-sleep. Absence of demand. A walk without a podcast is different from a walk with one. A meal eaten without a screen is different. The system needs time where nothing requires a response — not even a low-stakes one.

A couple of things tend to move the needle alongside that.

Breathwork with an extended exhale has solid evidence for helping shift a dysregulated state: specifically by engaging vagal tone and supporting the parasympathetic nervous system, which governs the body's capacity to actually recover rather than just pause. The distinction between engaging vagal tone and "getting calm" matters: one is a physiological shift; the other is a description of how you feel. The first enables the second; the second alone doesn't produce the first.

The aparigraha question

In yoga philosophy, aparigraha — non-grasping — is the practice of honestly examining what you're reaching for, and whether the reaching is serving you.

I've been teaching this long enough to see it show up in some predictable places. It appears in people who push through injury because rest feels like giving up. It appears in people who can't take a holiday without staying accessible. And it appears, increasingly, in how people adopt AI tools: as an extension of the same drive that generated burnout in the first place: more, faster, better. The efficiency that was supposed to give you your time back somehow just raises the floor on how much you're expected to produce.

That's worth sitting with honestly. Not as a moral judgment about ambition or productivity — both are fine. But as a real question: is this tool serving the life I want, or is it serving the same machine I was already inside?

Mental Health UK's 2026 burnout report (YouGov, n=4,502) found that 91% of UK adults experienced high or extreme pressure in the last year, and one in five workers took sick leave for stress-related reasons. AI-era burnout is arriving into an already poorly-served landscape. Most organisations aren't measuring cognitive load. Most individuals aren't either.

Recovery starts with naming what's actually depleted — and designing for that, not for the dashboard that says you're still productive.

If this is landing, more in this vein — practical, honest, no fluff — arrives when you sign up at mindbodyzen.co.uk.

A few questions worth thinking through

The HBR study found the most productive AI users were also the most burned out. Why would being good at something make burnout worse?

Because in high-functioning burnout, the warning lights don't fire. Productivity stays high; the system just charges less and less between uses. Being good at managing AI tools means you produce more output with the same cognitive drain — so the drain compounds without any natural brake on it. The 88% figure in the study isn't a paradox; it's the pattern.

Is AI cognitive load really different from ordinary work stress?

The underlying mechanism is the same: chronically overloaded decision-making system, too little genuine recovery. What's different is the source: the load now comes from tasks that feel low-stakes (reviewing a summary, approving a draft, choosing which of three generated options to use). That invisibility is the problem. We're much better at noticing and recovering from obviously hard work than from a long stream of small choices that feel like they shouldn't cost anything.

If the load is cognitive, not physical, does the recovery look different?

Yes, and this is the part most advice misses. Cognitive recovery isn't sleep alone — it's genuine non-demand time. Walks without inputs. Meals without screens. Breathwork that genuinely shifts your physiological state rather than just making you feel briefly calmer. And honest examination of whether you're reaching for more efficiency as the solution, when the problem is already too much demand on the system that processes it.