AI Is Accelerating Your Decline. We Designed a Way to Slow It Down.

You’ve asked ChatGPT to help you with something. You debate if you should switch to Claude. You're twenty minutes into a conversation, keeping track of its answers while you finish a report, clear your inbox, and fire off a couple of emails. Then something shifts. The information is coming thick and fast, and suddenly you're not keeping up. You've typed the email to the wrong recipient, you've lost track of which thread you were replying to, and somewhere between the original question and the outcome, you're not even sure if you're on the right lines anymore.

AI has sped up almost everything we do. It helps with ideation, carries out tasks we used to spend hours on, solves problems we can't solve alone. But do you ever feel just as fatigued using AI as you did without it? That's because AI is adding to the decision load, not taking it away. Let me explain.

What AI is actually draining: It's not burnout

Earlier this year, Boston Consulting Group and the University of California published a study in Harvard Business Review identifying a new form of cognitive fatigue specific to AI use. They called it "AI brain fry" — defined as mental fog, slower decision-making, and an inability to tell whether your AI-assisted work even makes sense. Workers experiencing it reported 19% greater information overload, and needed significantly more mental effort just to do their jobs.

But here's what caught my attention as a neuroscientist. AI users actually reported less burnout than their peers who weren't using AI. That sounds contradictory until you think about what's happening beneath the surface. Burnout is emotional and physical exhaustion. What AI is doing is different — it's depleting something more specific: your attention, your working memory, your executive control. You're not emotionally or physically diminished. You're cognitively depleted. And it can be hard to tell the difference when all you're feeling is fatigued.

What happens when you can't keep up 

I worry about the intricacies of pairing humans and AI. It may reduce burnout, and it has undeniably increased the speed at which we can get things done. But I wonder if that speed is too fast for us. AI increases decision tempo.

It generates, suggests, summarises, and recommends at a pace set by the system — not by your brain. When your brain can't keep up, one of two things happens. The first is automation bias: you stop scrutinising AI's output and start accepting it uncritically, not because you've decided to trust it, but because evaluating it properly would cost cognitive resources you no longer have. The second is automation disuse: you start rejecting AI's recommendations reflexively, even when they're right. 

One is over-trust. The other is over-distrust. Both lead to worse decisions. And neither is visible to the person experiencing it.

That should sound familiar if you've been following Numin's research. In our clinical trial, participants experiencing decision fatigue couldn't detect their own decline — subjective ratings showed no difference between Numin and placebo, but objective performance deteriorated significantly. The mechanism is the same. The context is just new.

“AI adoption is inevitable, and human cognitive capacity is finite.”

What you can do about it

The answer isn't to stop using AI. It's to stop pretending your brain can keep pace with it indefinitely.

1. Batch your AI work. Using a prompt that requires oversight while you're answering emails and checking Slack is the fastest route to depletion. Use AI in focused blocks, then step away before you act on what it gave you.

2. Don't evaluate AI output when you're already depleted. If you've been making decisions all day, that's the worst possible moment to decide if something AI generated is right.

3. Notice when you've stopped scrutinising. The moment you realise you've been hitting "accept" without reading properly is the moment to stop.

Protect it or be replaced

I do fear that one way this could end is replacing human decision-making entirely with AI systems that don't suffer from the same neurobiological constraints we do. But what would our world look like if we did that?

The goal shouldn't be "more AI" to replace human deciders. The goal must be deliberate cognitive protection. If we don't proactively manage the human side of the equation, we risk losing our most valuable resource: the quality of our decisions.

This is where Numin sits for me, and why I don't think there's been a more important time for it to exist. The neural resources AI is depleting — attention, working memory, executive control — depend on neurotransmitter systems that are measurably affected by sustained cognitive load. Numin was designed for the specific biology that's now under more pressure than ever.

The question isn't whether to use AI. It's whether you're protecting the brain that has to work alongside it.