Companies are spending more on artificial intelligence than ever, and most of them cannot find the payoff. In one Atlanta Federal Reserve study, about 90 percent of executives said AI had not yet lifted productivity at their firms. A new line of research offers an uncomfortable explanation for the missing gains, and it points straight at a strategy many of those same executives have embraced: the layoffs meant to prove AI is working are quietly destroying the very thing that makes AI work.
The argument comes from researchers who analysed millions of employee satisfaction reviews, thousands of corporate financial reports, and hundreds of AI investment and layoff announcements from US public companies over the past five years. They found a tight, unlikely-to-be-coincidental pattern: as a company's AI investment announcements go up, so do its announcements of AI-related job cuts. In some cases the sequence even ran backwards, with firms cutting staff first to free up capital for the AI spending to come.
The logic is easy to follow. A manager under pressure to show a return on an expensive AI rollout reasons that if the software makes people more efficient, the company needs fewer people. Trimming headcount books the saving immediately. The trouble is that investors do not seem to reward it. When the researchers looked at how markets reacted to these AI-branded layoffs, the average share-price response was close to zero, and for more than half of the events it was negative or flat. A muted reaction to a cost cut usually means the market suspects a hidden cost somewhere else.
Here that hidden cost has a name: fear. Combing through Glassdoor reviews, the researchers found that comments about AI were markedly more negative than the overall tone of employee sentiment, and that concerns about job security were the single most critical theme. This matters because employee sentiment toward AI turns out to be one of the strongest predictors of firm productivity once AI is in use. Workers are simultaneously being told to adopt the tools enthusiastically and watching those same tools cited as the reason their colleagues were let go. Half of Americans, one Reuters and Ipsos poll found, fear AI could put someone in their own household out of work.
Management, meanwhile, sounds untroubled. Across roughly 10,000 earnings-call transcripts, executives discussed AI in consistently optimistic terms, yet that optimism bore no meaningful relationship to actual productivity. The people on the calls feel great; the people doing the work feel exposed. And it is the second group, the research suggests, whose mood actually moves the numbers. Optimism at the top is cheap. Trust on the floor is what converts a licence for a chatbot into output.
None of this argues that AI cannot raise productivity. It argues that treating AI adoption and workforce cuts as the same project is a category error. A firm that shares the gains, retrains people, and frames the technology as something that works alongside employees preserves the goodwill that makes the tools effective. A firm that uses AI mainly as a reason to shrink is left with demoralised staff and a disappointing return, having spent heavily to manufacture the fear that undercuts its own investment. The efficient move, it turns out, is also the humane one.