Every enterprise layoff announcement in 2025 followed the same script. The CEO stood at the podium and said something about “realigning resources toward AI.” The market rewarded the cut. The stock moved. The narrative held because the numbers were small enough to feel surgical. 56,000 AI-attributed job cuts across the entire year. Painful, but absorbable. A rounding error in a 160-million-person labor market.
That was the floor everyone planned around.
Then 750 CFOs told the truth to a survey they assumed would stay academic.
The NBER, Duke University, and the Federal Reserve jointly asked chief financial officers what they actually projected for 2026. Not what they would say on an earnings call. Not what the board deck showed. What they were modeling internally.
The number was 502,000. Nine times the 2025 rate. Not a gradual acceleration. A step function.
But the survey asked one more question. It asked whether the AI tools already deployed had delivered measurable productivity gains. Over 80% said no.
Read that again. The executives planning to cut half a million jobs in a single year, specifically because of AI, privately report that the technology justifying those cuts has not yet demonstrated it works in their own companies.
This is not a story about AI replacing workers. That story requires the AI to be doing the work. This is a story about a capital reallocation decision that has decoupled from its own evidence. The CFOs are not cutting because AI proved it could do the job. They are cutting because every other CFO is cutting, and the market punishes the one who does not.
The 2025 floor of 56,000 was set by companies that had tested AI and found specific roles it could absorb. That floor made sense. It was slow. It was defensible.
The 2026 floor of 502,000 is set by companies that have tested nothing and are cutting anyway.
The survey was conducted by three institutions whose entire purpose is to measure what is actually happening in the economy. They measured a consensus to fire people for a tool that 80% of the people doing the firing admit has not worked yet.
The floor did not move because the technology improved. It moved because the fear of being last became more expensive than the cost of being wrong.
👉 Read the agentic security news. Instantly analyze the threat vector, see if it applies to your setup, and find the gaps with our interactive playbook. All free.
👉 Follow me on LinkedIn | X | Substack for weekly analysis of real agent failures, control gaps, and what the frameworks are and are not catching.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.