Executives Fire Workers for Machine's Promise, Not Its Work

A survey of over 1,000 executives shows companies cutting jobs based on what AI might eventually do, not what it has actually proven it can do. Goldman Sachs tells the newly unemployed to expect a long search and a pay cut when they land.

The Layoff Is Real. The AI Is Aspirational.

Here is a sentence that should not make sense, and yet — as if this were not enough — here we are. American corporations have begun eliminating human employees not because AI has replaced them, but because AI might replace them, eventually, probably, according to a slide deck.

The evidence comes from Harvard Business Review, by way of Thomas Davenport of Babson College and Laks Srinivasan of the Return on AI Institute, who surveyed 1,006 executives worldwide in December 2025 and found that companies are restructuring their headcount around AI's anticipated capabilities rather than its demonstrated ones. Not what the software does. What executives believe it will do, once it grows up.

Who's Doing It

Ford, Amazon, Salesforce, and JPMorgan Chase have each had a CEO stand up in public and say, with the confidence of a man who will not personally be affected, that white-collar jobs are about to disappear. Entry-level workers, customer service reps, and programmers made the list first — which is its own small irony, since the technology being credited with their obsolescence is, as of this writing, a technology that will still confidently misspell its own name if you ask it twice.

One presumes nobody at the board meeting asked the model to actually do the job first. That would have been the empirical approach. Empiricism was, sources confirm, not on the agenda.

The Incentive Isn't the Robot, It's the Story About the Robot

This is, at bottom, not really a story about artificial intelligence. It's a story about incentives, and about how a sufficiently confident narrative can do a layoff's job before the technology underwriting it has done anything at all. Announce that AI is coming for the org chart, cut the org chart, and book the savings this quarter — the efficiency gain is real on the spreadsheet even if the software generating it is, at this stage, mostly generating enthusiasm. Even now, with U.S. unemployment sitting at levels most economists would call healthy, reporting citing roughly 1.2 million jobs slashed last year suggests the spreadsheet is winning.

It's the corporate equivalent of firing your car mechanic because you read a brochure about self-driving cars. The brochure is very optimistic. The car still can't change its own oil.

Meanwhile, Back in Reality

For the humans on the receiving end, the promise doesn't come with a severance clause. Goldman Sachs advises displaced workers to expect a long search and real earnings losses upon reemployment — the standard finding for anyone laid off in a downturn, except this isn't a downturn. It's a bet, placed by other people, with your job as the stake. This, Goldman notes, is entirely fine, in the sense that words can be arranged into that shape and technically qualify as a sentence.

Fired today for what the robot might do tomorrow. Dear reader, at least when they fired people for what the robot did do, there was a robot involved somewhere in the transaction.

Sources: Harvard Business Review — Companies Are Laying Off Workers Because of AI's Potential, Not Its Performance