Meta's scrapped plan to cut teams 60% via AI

Reuters reported that Meta created an internal plan, codenamed Project OT for “organization transformation,” that explored cutting some team headcounts by as much as 60 percent by having AI agents take over much of the daily work currently done by thousands of employees, with small human teams left to oversee the automation. Meta confirmed the plan’s existence to Reuters and confirmed CEO Mark Zuckerberg set it in motion earlier this year, but said it ultimately did not move forward with every scenario the planning exercise considered.

Project OT reportedly called for two rounds of layoffs. The first happened in May; Zuckerberg canceled the second immediately afterward, and Reuters says it could not determine what prompted the reversal. One internal document described “AI native” as a company where “AI-ready tools and agents interact, workflows are automated, [and] new builds are AI-first,” including selling AI agents to third parties, something Meta started doing in June. The plan surfaced months after Zuckerberg had already acknowledged mistakes in Meta’s broader AI overhaul and described deep morale problems inside its restructured AI unit. An HR executive quoted in Reuters’ reporting said the scenarios under consideration could have cut headcount by roughly 25 percent or more.

What actually happened when Meta tried it

Internal posts cited by Reuters said AI agents took “large-scale, disruptive actions that humans are unlikely to execute,” contributing to a 40 percent year-over-year rise in major technical and security incidents, with the employee time spent resolving those incidents up by as much as 70 percent. Meta CTO Andrew Bosworth noted in an internal post that AI-assisted code changes to internal platforms were up 220 percent year over year, but changes that actually shipped new or upgraded features to users were up only 36 percent, a gap between activity and output that undercut the productivity case for the cuts. Zuckerberg reportedly told a company meeting in July that “the trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected.”

For any company weighing a similar bet, the lesson from Meta’s own retreat is concrete: more AI-generated code changes did not translate into proportionally more shipped value, and the agents given more autonomy produced incident rates humans rarely would have on their own. “AI native” org design, even inside a company with Meta’s resources and appetite for the bet, is still closer to an unproven pilot than a plan you can execute on schedule.