Why AI is making your team more exhausted, not less

There’s a pattern emerging in organisations that have gone all-in on AI.

The tools are faster. The output is higher. The velocity metrics are green. And yet — the teams are more drained than ever. People are burning out not despite the AI, but because of how it’s being deployed.

This isn’t a coincidence. It’s economics. And there’s a hidden tax nobody is talking about.

The Jevons Paradox of Cognition

In 1865, economist William Stanley Jevons observed something counterintuitive: as coal-burning engines became more efficient, coal consumption went up, not down. Efficiency didn’t reduce demand. It increased it. Because the more useful something becomes, the more uses people find for it.

We’re watching the AI version of this play out in real time.

AI makes generating content, code, or strategy cheap and fast, so organisations aren’t saying “great, the team can go home early.” They’re saying “great, now you can manage 50 projects instead of 5.” The efficiency gain gets immediately reinvested into higher output expectations. The machine gets faster, so we load it with more.

Here’s the hard truth: AI scales the output. It does not scale the human CPU.

We are hitting a biological ceiling. Pushing past it doesn’t produce more productivity — it produces system failure. Burnout isn’t a personal weakness or a failure of discipline. It’s what happens when you run any engine past its limits long enough, regardless of how much fuel you pour in.

We’ve spent decades talking about physical burnout — long hours, no breaks, unsustainable workloads. What we’re entering now is something different. Cognitive burnout. And the tools that were supposed to liberate us are, in many cases, accelerating it.

The Evaluation Tax

Here’s where it gets specific.

AI didn’t kill the work. It replaced it with what I’d call the Evaluation Tax.

This is why your team feels fried even though the AI is doing the heavy lifting.

The AI takes 10 seconds to generate a complex 12-month strategy. The human takes 15 to 20 minutes of deep, high-stakes focus to verify whether that strategy is brilliant or a complete hallucination. Force an employee to verify 50 AI outputs a day and they aren’t working less. They’re stuck in a high-fatigue loop of constant critical evaluation — and that loop is often more draining than original creation.

Here’s why. When you create something from scratch, you build the logic as you go. There’s a flow state in it. You’re constructing, not inspecting. When you evaluate someone else’s output — or an AI’s — you have to reverse-engineer the reasoning while simultaneously hunting for invisible errors. There’s no flow state in that. Just vigilance, all day.

We traded scrubbing for quality control and forgot to adjust the workload to account for the cognitive cost of the latter.

The Evaluation Tax is being paid right now, quietly, by teams everywhere. It doesn’t show up in a sprint velocity report. It doesn’t appear in a utilisation dashboard. It accumulates in the background — in slower decisions, in missed nuance, in the creeping sense that despite getting more done, nothing feels finished properly.

What This Actually Costs

Velocity metrics stay green while judgment quietly degrades.

That’s the dangerous part. The numbers look fine. The output is high. The AI is generating, the humans are reviewing, the machine keeps moving. But the quality of the decisions being made on top of all that output — the strategy calls, the risk assessments, the judgment calls that actually determine whether the company goes in the right direction — those are being made by people who are cognitively depleted.

You can’t automate judgment. You can only deplete the people who have it.

The Washing Machine Didn’t Work This Way

The washing machine is the standard metaphor for beneficial automation and for good reason. It didn’t teach anyone to scrub better. It made scrubbing disappear — and gave people back something more valuable in return. Time. Energy. Attention for things that actually mattered.

The key distinction: the washing machine runs in the background. It doesn’t require attention to operate. You set it, you forget it, and you come back to clean clothes. The cognitive cost of using a washing machine is essentially zero.

AI is not working this way in most organisations. It requires constant attention to operate well. It requires human judgment to validate, to direct, to catch errors, to determine what to do with the output. The more AI an organisation uses, the more cognitive overhead it creates — not less.

Some companies looked at the washing machine and said: great, now you can do ten loads a day. And make the employees monitor each cycle manually.

That’s not empowerment. That’s just a faster way to burn people out.

Before You Deploy, Ask One Question

Before reaching for AI as a solution, the only question worth asking honestly is: what happens to the human on the other side of this decision?

Not “how much faster will this be” or “how much can we save.” What actually happens to the person who now has to manage, verify, direct, and quality-check the output of this system — on top of everything else they’re already doing?

AI can feel very productive right up until there’s nobody left with enough cognitive bandwidth to check whether any of it was actually good.

The Evaluation Tax is real. It’s time to start budgeting for it.