The AI jobs apocalypse hasn't shown up yet

If you read the headlines this year, you'd think graduates were walking straight out of university into a wall of AI. A careful new study of the numbers says that wall hasn't turned up in the unemployment data.

Economists Robert Fairlie and Jane Wu looked at US Current Population Survey data for June, July and August 2026. Their NBER working paper set out to catch the moment AI started showing up in the jobless rate for recent college graduates, the group most often described as first in line for replacement. They found unemployment for that group didn't spike in summer 2026 compared with previous summers.

It also didn't rise in any significant way against two comparison groups: older college graduates and young workers without a degree. The authors even widened the definition of unemployment to include people who said they wanted a job. That added nearly two percentage points to the rate, and there was still no statistically significant increase this summer.

Why this matters for leaders

I'm not saying AI isn't changing work. It plainly is. But a lot of leadership teams are making big calls on the assumption that a jobs collapse is already here. Hiring freezes and cancelled graduate programs are on the table. Some of those decisions are being sold internally as "getting ahead of AI."

This paper suggests caution. At least in the data the authors studied, the dramatic displacement everyone's been bracing for hasn't arrived in the unemployment figures yet. If you're cutting because you think the market has already moved, check that it has.

What the paper doesn't say

It's worth being precise here, because this is exactly the kind of finding that gets overstated in the other direction. The paper looks at unemployment for recent graduates over one summer. It doesn't say AI will never affect jobs, and it doesn't measure how roles are changing inside companies.

Unemployment is also a blunt measure. People can stay employed while their job changes shape, or while they take work below their skill level. The authors are careful about what they claim, and I'd be careful too.

They did find one interesting signal. When they matched unemployment against occupation, they saw some evidence of a link with remote work availability. That's a thread worth watching, and a reminder that more than one force is moving the graduate market at once.

Read the evidence before you restructure

The broader lesson is about how leaders process this stuff. AI news moves fast and it's loud. A single viral prediction can end up in a board paper as if it were a fact. Meanwhile the slower, more careful studies get ignored because they're less exciting.

So build a habit. When someone in your executive team says "AI is wiping out jobs," ask what the evidence is and what it actually covers. Then ask whether your own data agrees. You may find the case for a restructure is weaker than the slide suggests, or you may find it's stronger. Either way you'll be deciding on something real.

That habit is a skill, and most teams don't practise it. A critical thinking session that works on your own board papers is one way to build it, before the next confident prediction lands on the agenda.