The Self-Defeating Layoff: New Research Shows AI Job Cuts Destroy the Productivity Gains AI Was Supposed to Deliver

The evidence is getting harder to spin. As companies pour record sums into artificial intelligence, a growing body of research is documenting an uncomfortable gap between what the technology was sold on and what it has actually delivered inside most firms. The latest entry — a five-year analysis from University of Pittsburgh business professor Mark Ma and colleagues at the Federal Reserve Bank of Atlanta — doesn’t just confirm that gap. It offers a mechanism for it, and the mechanism is the layoff slip itself.

The numbers nobody in the boardroom wants to read

Start with the survey data. An Atlanta Fed study found that about 90% of executives believe AI has not yet boosted productivity at their companies. Research from the National Bureau of Economic Research pushes the picture further into the red: more than 90% of executives who responded said there was “no impact of AI on own-firm employment over the past three years,” and 89% reported “no impact on labor productivity” at all.

To be clear about what this does and doesn’t mean. It is not a claim that AI is useless — in specific, well-scoped workflows, from code generation to document review, controlled studies have shown meaningful gains for particular tasks. What it says is that at the level where it matters for the economy, the firm, the aggregate effect after three years of adoption is close to nil for nine out of ten companies. The productivity statistics tell a similar story: the broader gains the U.S. has enjoyed since 2021 are more plausibly attributed to remote work and post-pandemic restructuring than to AI, as other economists have argued.

Against that backdrop, the behavior of corporate America looks almost perverse. AI-cited layoffs keep accelerating even as the evidence for AI-driven productivity stays thin. Fortune’s January analysis found AI was cited as the reason for nearly 55,000 U.S. job cuts in the first 11 months of 2025. The Ma-led research adds a disturbing detail: some companies started laying off employees before pouring money into AI, as a way to free up capital for the investment itself.

The mechanism: fear is the productivity killer

Here is where the new research earns its keep. Ma’s team analyzed millions of Glassdoor reviews, thousands of corporate financial reports, and hundreds of AI investment and layoff announcements from U.S. public companies over five years. Two patterns emerged.

First, AI-related comments in employee reviews are “much more negative” than the overall tone of those reviews. Workers aren’t neutral about the technology being deployed around them — they are scared of it. Digging into what drives that hostility, the researchers found job security concerns were “the most critical by far,” ahead of lack of training, few chances to upgrade skills, poor corporate AI leadership, and doubts about whether AI actually improves anything.

Second — and this is the paper’s punchline — there is a strong association between employee sentiment toward AI and firm productivity, measured against the employer’s financial information. Anti-AI sentiment among workers lowers productivity and offsets the potential efficiency gains the AI would otherwise deliver. When the researchers tested how AI sentiment changes after companies announce AI-driven layoffs, they found a sharp decline.

The causal chain is not complicated. A company buys AI tools expecting efficiency. It then cuts headcount, citing the AI, to show a return on the investment. The remaining workforce watches colleagues lose jobs to the technology they are simultaneously being told to embrace. Their sentiment collapses. Their productivity drops. The AI’s benefits are partially or wholly cancelled out by the fear it now travels with.

“In fact, these job cuts damage employee sentiment toward AI — which is one of the strongest predictors of firm productivity when AI is used,” Ma wrote in The Conversation. “Using AI to justify cutting jobs is, in our view, a strategic miscalculation that cuts against the benefits of AI.”

Managers are optimistic. The market isn’t listening.

The study’s third finding might be the most bruising for executives. When Ma’s team analyzed the tone of management’s AI-related discussion across roughly 10,000 earnings-call transcripts, they found it consistently, glowingly optimistic — and bearing no significant relationship to productivity outcomes. Managerial enthusiasm, in other words, is noise.

The stock market, for its part, has stopped pretending. Examining market reactions to AI-cited layoff announcements, the researchers found the average return was close to zero, and negative-or-flat for more than half of the events. Investors, it seems, have learned to price in the possibility that an “AI restructuring” is a company confessing it overpaid for tools that didn’t move the needle. There are exceptions — Block saw its stock jump when it announced AI-driven staff trims — but they are exceptions.

The overall picture is one of a self-defeating loop: companies spend heavily, cut deeply to justify the spending, demoralize the workforce whose adoption determines whether the tools work at all, and then discover the market doesn’t reward any of it.

What companies should do instead

Ma’s recommendation is blunt: stop using AI to justify layoffs. Beyond that, the research points to sharing AI gains with employees — investing in skills and expanding opportunity rather than cutting headcount — as the path that actually captures the productivity upside. The firms that profit from AI, the argument runs, will be the ones that make workers feel the technology is working with them, not against them.

There is a broader reckoning underway too. Even OpenAI CEO Sam Altman conceded over the weekend that “we have not had the iPhone moment of like completely changing how someone interfaces with technology.” When the industry’s chief evangelist is hedging, and 90% of executives are reporting no productivity movement, and half of Americans tell Reuters/Ipsos pollsters they fear AI could put someone in their household out of work, the “trillions in funding on the strength of lofty promises” era of AI discourse is visibly losing altitude.

None of this means AI adoption will stop, or should. It means the intermediate period — the one we are living through — is governed by an unforgiving arithmetic: the technology’s value inside a firm is mediated by the humans asked to use it, and treating those humans as the cost line to be cut is the fastest way to ensure the value never shows up in the first place.