One in four marketers using AI tools at work is mentally fried from the experience. Not from the work itself, but from babysitting the AI.
That's the central finding from a study published in Harvard Business Review examining what researchers call "AI brain fry" - the mental exhaustion that comes from constantly reviewing, correcting, and guiding AI-generated outputs. Across all professions surveyed, about 14% of employees using AI reported this kind of cognitive strain. But the numbers spike hard in certain fields.
Marketing and HR Take the Biggest Hit
Marketing leads the pack at 26%, followed by HR at 23%, operations at 21%, and engineering at 20%. Finance and IT land at 19% and 18% respectively. Legal sits at the bottom with just 6%.
The pattern makes sense when you think about what these roles actually do with AI. Marketers and HR professionals generate massive volumes of text - job descriptions, campaign copy, social posts, candidate communications - and every single piece needs a human to read it, check the tone, fix the hallucinations, and make sure nothing embarrassing gets published. That's not automation. That's a proofreading treadmill.
Legal's low number is telling too. Lawyers tend to use AI more cautiously and in narrower ways, which apparently means less of the constant back-and-forth correction cycle that burns people out.
The Symptoms Are Real
Workers in the study described mental fog, difficulty concentrating, slower decision-making, and headaches after extended AI sessions. Many reported needing to physically step away from their computers to reset. These aren't people complaining about new software being confusing. These are people describing genuine cognitive fatigue from a specific type of work: supervising machines that are almost-but-not-quite good enough to trust.
This lands differently than the usual "AI will take your job" narrative. The emerging reality for a lot of knowledge workers is less "replaced by AI" and more "stuck editing AI's homework all day." The productivity gains from generating a first draft in seconds evaporate when you spend 20 minutes fixing it, then lose another 30 minutes to mental fog afterward.
What This Means for Tool Selection
The study's implication for anyone choosing AI tools is straightforward: the quality of the output matters more than the speed of generation. A tool that produces 80%-correct text you need to heavily edit creates more total work - including the invisible cognitive cost - than one that produces 95%-correct text you can approve with light changes.
This is also a strong argument for building better prompts and templates rather than just throwing queries at a chatbot and hoping for the best. Every minute spent setting up a reliable workflow is a minute you're not spending in the brain-fry zone of reading AI slop and trying to figure out what's wrong with it.
For teams running heavy AI workflows in marketing or HR, the practical takeaway is simple: build in breaks, rotate people off AI-editing tasks, and stop measuring productivity purely by output volume. The 26% figure for marketing isn't a curiosity. It's a workplace health signal.