When a CEO Blames AI for Layoffs, Check Three Things

Salesforce's support organization went from about 9,000 people to about 5,000 in 2025, as the company's own AI agents took over the conversations. That's an AI layoff in the plainest sense.

Plenty of what gets called an AI layoff is something else. We tag events in our tracker as AI-related conservatively, and even with a strict label we find three different mechanisms hiding under it. They mean very different things for the job market.

Some cuts are straight substitution

In the first kind, AI demonstrably does the work and the people who did it are let go. Duolingo's translation contractors in 2024. Salesforce's support cuts. IBM leaving back-office roles unfilled because AI absorbs them. Chegg and Stack Overflow are the market-side version: AI never joined those companies, it took their customers.

This is still a minority of announced cuts. It's also the fastest-growing category, and it clusters in a few functions: support, content, translation, routine QA, data entry and tier-1 IT.

Most AI layoffs are swaps

The dominant pattern is reallocation, where a company swaps people while its overall size holds roughly steady. Intuit cut 1,800 people and hired 1,800 different ones. SAP restructured 8,000 roles to fund "business AI." Microsoft cut around 15,000 across 2025 while spending record sums on AI infrastructure, converting payroll into compute. Meta cut people it called "low performers" and hired machine-learning engineers at premium pay.

It's why the market feels contradictory, with mass layoffs and talent shortages at once, sometimes at the same company. The demand didn't disappear. It moved to skills most of the laid-off people don't have yet, and we think that gap, between who's being let go and who's wanted, is the central tension of this whole transition.

Some are cost cuts with a better story

"AI" also works as cover, a flattering explanation for ordinary cost-cutting. Say a layoff positions you for the AI era and it sounds strategic. Say you over-hired or missed your numbers and it sounds like management failed. Investors currently reward the first version, so companies use it.

The tells are cuts that are broad instead of function-specific, no AI system actually taking on the work, and vague AI investments.

Cover was the most common type in 2023, and it's shrinking as real substitution and reallocation grow. Sorting an announcement into this bucket is often a judgment call, we admit, since no company says it's using AI as cover.

Reading the next announcement

When the next one lands, look first at which functions were cut. If they're the AI-exposed ones, take the AI claim seriously. If it's across the board, be suspicious. Then check the company's careers page that same week, because a company hiring while it fires is usually reallocating. And ask whether a named AI system actually does the work, or whether "AI" is only doing narrative work.

Substitution shows you which tasks are falling and reallocation shows you which skills are rising. Cover tells you a company had a bad quarter.

All three show up in our tracker. Only two of them should change what you do with your career.