Fifty-four percent. That, according to the tracking site Layoffs.fyi, is the share of 2026's layoff events in which a company has explicitly blamed artificial intelligence or automation, up from under eight percent a year ago. More than 170,000 workers have been cut with AI named as a factor. It is the statistic anchoring every story about machines coming for white-collar work. The awkward part is that the most ambitious corporate attempt to actually replace workers with AI just fell apart from the inside.
Reuters this week detailed Project OT, short for Organization Transformation, a confidential Meta program conceived at a January leadership retreat to remake the company as "AI native." The ambition was sweeping. AI agents would absorb much of the routine work of thousands of employees, overseen by smaller, "talent-dense" teams of humans. In scenario planning, executives explored cutting some teams by as much as 60 percent. The overhaul was to arrive in two waves, one in May and a larger one in November.
The second wave never came. On the night of May 19, hours before the first round began, Mark Zuckerberg pulled the plug on the November phase. Meta still cut about 10 percent of its staff the next day, but the grander vision was quietly shelved. Two things had gone wrong at once. Employees were in open revolt, convinced the AI drive was aimed squarely at them. And the internal data was refusing to cooperate.
That data is the most revealing part of the story. Meta found that code changes across its platforms had risen 220 percent year on year, a figure that sounds like a productivity boom. Yet new or improved features actually reaching users rose just 36 percent. More code was not becoming more product. Worse, major technical and security incidents had climbed 40 percent. The AI agents at the heart of the strategy were generating enormous activity and comparatively little value, exactly the gap a company would want closed before wagering its headcount on them.
Set that against the wider evidence and a pattern emerges. Deutsche Bank analysts have coined the phrase "AI redundancy washing" for firms that pin ordinary cost-cutting on a fashionable cause. A Stanford study tracking payroll data finds the clearest AI effect is not mass layoffs but a hiring freeze at the bottom, with employment for workers aged 22 to 25 in exposed jobs now running about 19 percent below where the trend would put it. And a fresh YouGov survey of 1,250 US workers found only about 3 percent believed they had lost a job to AI since 2023. The sociologist who commissioned it, a self-described AI pessimist, called the finding surprising.
None of this means AI leaves work untouched. The young and the entry-level are plainly feeling something, and reduced hiring is its own kind of harm even when no one is escorted out. But the loud version of the story, that agents are ready to run the office, keeps colliding with the quiet version, in which the tools produce plenty of motion and uncertain results. Meta had every incentive and every resource to prove the loud version true. Its own dashboards talked it out of the attempt.