Where’s my AI Job Apocalypse? I had it a second ago.

Every generation has a technology that's supposed to end work as we know it.

When I was in middle school, it was personal computers. After college, it was the world wide web. Each time, the experts rolled out charts showing jobs falling off a cliff. Each time, the cliff never came.

Might AI be different this time?

Look at the Federal Reserve Bank of St. Louis employment data. The knowledge worker line (management, professional, and related occupations) has been flat since March 2025. Not collapsing. Not growing. Flat. If generative AI were the primary driver of structural displacement among knowledge workers, you would expect something more decisive in that line by now. The same goes for Sales and Office Occupations which are arguably more exposed to AI task displacement.

So what’s going on here? It’s hard to read Substack or LinkedIn these days without an expert claiming widespread job loss.

ATMs are a great case in point. In 1985 there were roughly 60,000 ATMs in the United States. By 2000 there were over 300,000. Many experts predicted a precipitous decline in jobs. Machines would replace tellers, bank branches would close, and banking employment would fall. Instead, ATMs reduced routine cash-handling work and banks shifted teller work toward customer service, sales, and relationship management.

Markets hate routine work. Given agentic AI's uncanny ability to do routine work, the nature of work is surely changing. But enabling technologies simply don't produce the broad structural job collapse that each generation's warnings predict. If AI is displacing jobs at scale, then commensurate growth in other job categories are masking this at the aggregate level.

That doesn't mean that AI won't be different from other technologies. But the burden of proof is high — and we are not there yet.