Glonce

Meta scraps AI layoff plan as agents underperform, staff revolt

businessThe Glonce editorsThe Decoder🔥76Glonce rating76Impact60Novelty70Relevance100Surprise70Players85Credibility80
Meta scraps AI layoff plan as agents underperform, staff revolt

According to Reuters, citing internal documents, Meta had planned to replace far more of its workforce with AI than previously known. Under the internal codename "Project OT," the company planned to shrink many teams by up to 60 percent, leaving the remaining work to small, "talent-dense" groups of human staff supervising virtual AI workers.

CEO Mark Zuckerberg halted the plan's second wave, which had been scheduled for November, on the evening of May 19, just hours before the first round of layoffs was due to begin.

Several factors drove the reversal. The agent technology never delivered the productivity gains Meta expected. Investors criticized the size of the company's AI budget. And the workforce openly revolted: employees came to believe that tracking software logging their mouse clicks and keystrokes was training the AI systems meant to replace them, and they flooded Meta's internal network with angry posts, Reuters reports.

Internal employee sentiment dropped from 74 to 55 percent over the episode. In July, Zuckerberg himself admitted that the agent technology had not sped up as fast as he had expected.

Key facts

  • Meta's internal plan, codenamed "Project OT," aimed to shrink many teams by up to 60 percent and shift the remaining work to small human groups supervising AI agents.
  • CEO Mark Zuckerberg halted the plan's second wave, set for November, on the evening of May 19, just hours before the first round of layoffs was due to start.
  • The AI agents failed to deliver the productivity gains Meta expected, and investors criticized the size of the company's AI budget.
  • Employees revolted after concluding that mouse- and keystroke-tracking software was training their own AI replacements, flooding Meta's internal network with angry posts.
  • Internal sentiment dropped from 74 to 55 percent; in July Zuckerberg admitted the agent technology had not progressed as fast as he had expected.

Why it matters

Meta's about-face is a rare public data point on the gap between AI-substitution ambitions and what agent technology can actually deliver inside a large organization today. "Project OT" was not a modest pilot: shrinking many teams by up to 60 percent and replacing the cut roles with AI agents supervised by small human groups was, by the report, a more aggressive plan than Meta had previously disclosed. That the plan's second wave was halted hours before the first round of layoffs was due to begin, after real productivity numbers and real employee pushback, undercuts the narrative that agentic AI is already a drop-in replacement for headcount at scale.

Who it affects

Meta employees on the teams slated for cuts, whose roles were reprieved when the second wave was halted. Meta investors, who had already pushed back on the size of the AI budget behind the plan. And, more broadly, other large employers weighing similar AI-driven restructuring, who now have a concrete example of such a plan collapsing under underperforming technology and internal resistance rather than being scaled up as intended.

How to use it

For executives sizing AI-driven headcount plans, the episode is a caution against locking in reduction targets ahead of proven agent performance. Tying a restructuring timeline to technology that has not yet demonstrated the expected productivity gains, as Meta did, creates exposure to exactly this kind of reversal. It is also a reminder that internal monitoring tools deployed alongside an AI rollout will be read by staff in light of that rollout's stated goals, whatever the tools were nominally introduced for.

How solid is it

The account rests on Reuters reporting drawing on internal Meta documents, relayed here without independent verification by this outlet. The specific figures given, the up-to-60-percent team reduction, the 74-to-55-percent sentiment drop, and the May 19 and July dates, are stated plainly and are internally consistent, and Zuckerberg's July admission that the agent technology underperformed corroborates the broader claim from Meta's own side. No Meta spokesperson or official statement is quoted in the source.

Risks and caveats

No year is given for the May 19 halt, the planned November second wave, or the July admission. The source does not say how many employees or teams were actually affected, only "many teams" and "up to 60 percent," nor does it detail the scope of the first round of layoffs. The tracking software is not named, only described as logging mouse clicks and keystrokes, and it is not confirmed that the software was in fact used to train Meta's AI agents, only that employees believed this. Whether Project OT is permanently abandoned or merely paused is not stated.