Corporations are slashing tens of thousands of jobs in the name of artificial intelligence. But a growing body of research suggests the promised productivity windfall has yet to arrive — and may not come the way executives think.
By Jordan Mehta · May 19, 2026 · 9 min read
The memo was polished and inevitable-sounding. “As we lean into the next era of intelligent automation,” it began, “we are making the difficult but necessary decision to reduce our workforce.” Replace the company name, adjust the headcount figure, and you have the template for dozens of similar announcements made over the past eighteen months. The script has become so familiar it practically writes itself — and that, perhaps, is part of the problem.
Across sectors ranging from cloud computing to financial services to consumer retail, executives are invoking artificial intelligence as both the cause of job cuts and the cure for whatever ailed their balance sheets. The promise is seductive: replace expensive, unpredictable human labor with software that never sleeps, never complains, and never asks for a raise. Yet a wave of hard data is now complicating that narrative in ways that boardrooms may not yet fully appreciate.
The central question being raised by economists, management consultants, and the laid-off workers themselves is deceptively simple: if AI is this transformative, where are the returns?
THE SCALE OF THE CUTS
To understand the stakes, it helps to absorb the sheer volume of what has happened. According to Crunchbase’s ongoing tracker, at least 127,000 U.S. tech workers were laid off in 2025 alone — and already, in just the first weeks of May 2026, more than 24,000 additional tech sector employees have been shown the door. These are not the pandemic-era over-hiring corrections of 2022 and 2023. They are deliberate, AI-branded workforce reductions by companies that are simultaneously reporting record profits.
The names attached to these cuts are not obscure startups stumbling on rocky ground. They are the anchors of the modern economy. Microsoft eliminated roughly 15,000 positions, framing AI as central to its productivity transformation. Amazon cut 14,000 corporate roles, stating that AI enables leaner structures and faster innovation. Salesforce reduced its customer support workforce by 4,000, with CEO Marc Benioff declaring that AI now handles up to half of the company’s work. IBM confirmed that AI agents had already replaced hundreds of back-office roles. Goldman Sachs, HP, Dell, Cisco — the list reads like a Fortune 100 roll call.
Some of the more striking cases go beyond big tech. Chegg, the education technology company, laid off 45% of its entire workforce after students began turning to generative AI tools instead of its traditional homework-help platform. CrowdStrike — the cybersecurity firm that made headlines in 2024 for triggering a global IT outage — cut 5% of its staff, explicitly citing “AI efficiency.” Even Workday, a company that sells HR software used to manage other companies’ workforces, eliminated roughly 1,750 of its own employees to redirect resources toward AI.
Key figures to know:
- 127,000+ U.S. tech workers laid off in 2025
- 55,000 layoffs directly attributed to AI in 2025
- 80% of AI-piloting firms reported workforce cuts, per Gartner
- AI-related layoffs projected to be 9× higher in 2026
THE PRODUCTIVITY PARADOX, REVISITED
There is a term economists have reached for to describe this moment, and it is one with a painful history. In 1987, Nobel Prize-winning economist Robert Solow observed something puzzling about the computerization of the American economy: “You can see the computer age everywhere except in the productivity statistics.” That observation became known as the productivity paradox — and it took roughly a decade before computers truly began to reshape output metrics in measurable ways.
Many economists believe we are living through a second version of this paradox today. Goldman Sachs senior economist Ronnie Walker recently noted that “we still do not find a meaningful relationship between productivity and AI adoption at the economy-wide level.” This is not a fringe view. It is increasingly the consensus among those who study the numbers rather than the press releases.
One CFO survey respondent captured the mood precisely: “Companies see the potential of AI without the financial results to match. They’ve invested and they’re realizing all these cool things they hope to do in the near future — but it’s not really showing up yet in revenue.”
Workers, meanwhile, are not exactly reporting a productivity utopia either. Some studies have found that AI tools are making certain workers less productive — placing greater strain on workflows and increasing time spent on some responsibilities by as much as 346%. Anyone who has spent meaningful time editing AI-generated content, correcting hallucinated citations, or re-prompting a model that has misunderstood context will recognize this phenomenon intuitively. The overhead of managing AI, it turns out, is not zero.
WHAT THE RESEARCH ACTUALLY SHOWS
Perhaps the most striking finding of this period comes from a Gartner survey of 350 global business executives whose companies each generate at least one billion dollars in annual revenue. The results should give pause to anyone who has accepted the AI-layoffs-equal-returns equation at face value.
Eighty percent of executives who had piloted AI or autonomous technology reported workforce reductions. So far, the headline writes itself. But here is where the story turns: the businesses cut jobs regardless of whether the technology was actually generating returns. The decision to shed workers was not contingent on AI delivering its promised value. It happened whether the ROI materialized or not.
Helen Poitevin, VP Analyst at Gartner and a key researcher of the study, was direct about the implications: “There seems to be no link between laying people off and getting ROI from AI investments. ROI is driven more by reinvestments in the workforce, rather than replacing employees with automation.”
What Gartner did find is instructive. The companies achieving the highest returns from their AI investments were not those who cut the most aggressively. They were those who invested in upskilling existing employees — training workers to build and manage their own AI agents and automations. The concept researchers call “people amplification” is the precise opposite of the replacement narrative that has dominated executive messaging.
As Poitevin put it bluntly: “The value isn’t in cutting people. That’s not where the productivity gains are going to come from.”
THE “AI WASHING” PROBLEM
There is another uncomfortable dimension to this story that receives insufficient attention: the possibility that many AI-attributed layoffs are not really about AI at all. Anthropic CEO Sam Altman has publicly raised this concern, describing a practice known as “AI washing” — attributing workforce reductions to artificial intelligence when the underlying motivations may be pandemic-era over-hiring corrections, macroeconomic pressure, or simply executive preference for leaner headcount.
The incentive structure for this kind of messaging is obvious. Blaming AI for job cuts sounds visionary and forward-thinking. Admitting that you hired too many people during a period of cheap capital and now need to right-size is considerably less flattering. “AI-driven restructuring” plays well with institutional investors in a way that “we made bad hiring decisions in 2021” simply does not.
This is not to say that AI is irrelevant to the employment picture. The outplacement firm Challenger, Gray and Christmas found that AI was the leading stated reason for layoffs in March and April 2026, with total AI-attributed cuts for the year on pace to be nine times higher than the 55,000 recorded in 2025. The trend is real. The question is whether the justification offered by corporate communications teams accurately reflects what is actually happening inside these organizations.
THE HUMAN COST OF A MISCALCULATION
Behind the statistics are people. Not “human capital.” Not “FTEs.” People with mortgages, children in school, and specialized skills they have spent years developing. And there is a dimension to mass AI-justified layoffs that rarely surfaces in shareholder letters: the destruction of institutional knowledge.
Professor Dilan Eren of Ivey Business School has been particularly direct, calling the elimination of junior and mid-level roles an “exponentially bad move” that hollows out the internal talent pipeline companies will need to develop the very AI systems they are betting on.
The argument is not sentimental. It is strategic. AI systems — even the best ones available today — require skilled humans to define problems, evaluate outputs, catch errors, and make the judgment calls that no model has yet demonstrated it can reliably make. The workforce you eliminate to pay for your AI infrastructure may be the same workforce you desperately need to actually run it.
As one industry analyst put it: “Talent and experience are among the largest assets an organization has. Without the insight to know what AI can and cannot do today, there is a real risk of losing the people who could shape future product innovations.”
WHAT RESPONSIBLE AI ADOPTION LOOKS LIKE
The companies most often cited as models of successful AI integration share a few characteristics that receive far less media attention than their headcount announcements. They treat AI as an augmentation layer rather than a replacement system. They invest heavily in retraining. They measure success not by how many positions they have eliminated but by how much output their existing teams can generate.
Researchers recommend what they call an “automate-to-augment” strategy — using AI to handle routine, low-risk tasks so that human workers are freed to tackle higher-order problems that actually differentiate companies in competitive markets. Most organizations have extremely busy workforces. Automating the tedious parts of those jobs does not eliminate the need for the people; it potentially expands what those people can accomplish.
This is a harder story to tell to Wall Street. It does not produce a dramatic reduction in the salary line. It requires patience and sustained investment in training programs that will not show immediate returns. But according to the available data, it is the approach that actually works.
THE QUESTION THAT DEMANDS AN ANSWER
We are, by most accounts, in the early innings of a genuine technological transformation. Few serious technologists dispute that AI will reshape how work is done across most of the economy over the next decade. The debate is not whether AI matters — it clearly does — but whether the current playbook being executed by corporate America represents a rational response to that reality or a reflexive cost-cutting exercise dressed up in forward-looking language.
The evidence, at this moment, points more toward the latter. Layoffs are rising. AI spending is rising. Productivity, at the economy-wide level, has not measurably followed. The returns that were promised when the first severance packages were signed have not yet appeared in the way they were described.
That may change. The productivity paradox of the 1980s resolved itself, eventually. Computers did transform the economy — just on a longer timeline than the hype suggested, and through channels different from those initially anticipated. The same may well prove true of AI.
But the workers being asked to bear the cost of this transition deserve more than assurances that history will eventually vindicate the decision. They deserve an honest accounting of what companies actually know, what they merely hope, and what they are still very much guessing at.
The AI blindspot is not a failure of the technology. It is a failure of candor — and the bill for that particular inefficiency has not yet fully come due.
