Meta wanted to reduce teams by 60% because of AI

Meta didn’t just flirt with AI inside the company. According to Reuters reporting, Mark Zuckerberg and senior leaders went into January with a plan to remake the way work happened at Facebook’s parent, and that plan was aggressive enough to cut some teams by 60% through layoffs and shifting people elsewhere.

The internal project had a name: Project OT, short for Organization Transformation. The idea was to turn Meta into an “AI native” company, with machines doing much of the daily work and smaller human teams supervising them. Leadership apparently wanted one layoff and restructuring in May, then another in November. HR even expected the cuts would be larger than the 2022-2023 round, when Meta let go of 25% of staff.

That second wave never arrived. Reuters says Zuckerberg backed off on the night of May 19, just hours before the first wave. Meta still cut 10% of employees the next day, but it dropped the November plan. By then, employees were already rebelling internally, and they suspected the AI push was really a way to replace them.

The reporting makes that fear look justified. Meta had already moved 20% to 30% of engineers on infrastructure and product teams into AI training work, and earlier layoffs had hit about 10% of engineers. The result was predictable enough: lower morale, overloaded teams, and ugly failures like the Instagram password-reset bug that could let an account be taken over by asking the bot to swap an email address.

Meta’s leaders seem to have been chasing the image of tiny, fast AI startups. Reuters says executives visited Asia, studied how AI startups there were organized, and ran pilots of their own. One internal presentation reportedly showed the target: 3-to-5-person “AI-native” teams doing the work of groups that used to be 10 to 20 people.

That is a neat spreadsheet idea and a rough way to run a company. Smaller teams can move quickly, but they also lose domain knowledge, redundancy, oncall depth, and breathing room. Meta may have wanted the speed without accepting the cost, which is usually how companies end up paying twice.