Meta built a detailed plan this year to remake itself around AI, and it went further than anyone outside the company realized. According to Reuters, the project, internally called Project OT for Organization Transformation, imagined a company where AI agents handled much of the daily work now done by engineers, designers, and product managers, with small groups of human staff overseeing them.
The plan traces back to a leadership retreat in January at Mark Zuckerberg's home in Hawaii. Executives had been watching AI-native startups, including some in Asia, build lean teams around AI tools instead of large staffs. They wanted Meta to do the same: shrink middle management, cut some teams by as much as 60 percent, and push remaining employees into flexible "builder" roles. One HR executive reportedly expected the cuts to match or exceed Meta's 2023 restructuring, which eliminated about a quarter of the company's workforce over two rounds.
A second wave of layoffs, on top of the 10 percent cut Meta already made this year, was reportedly planned for November. Zuckerberg pulled back on that plan, and the reasons line up with two problems Meta was having at the same time.
The first problem was that the AI simply was not ready. Meta's internal metrics showed a big gap between effort and output: code changes to its AI systems rose 220 percent year over year, but new features that actually reached users grew only 36 percent. Technical and security problems rose 40 percent, and the time employees spent fixing them rose 70 percent. In July, Zuckerberg told staff at a town hall that agent development "hasn't really accelerated in the way that we expected."
The second problem was self-inflicted. Around the same period, Meta installed monitoring software on employee computers that recorded mouse movements, keystrokes, and screen activity, partly to generate training data for the AI meant to eventually take over those tasks. Employees found out, and reaction was harsh: sentiment scores dropped by 19 points and organizing efforts inside the company grew.
Put those two things together, and the pause makes sense. Meta was asking employees to help build the tool that might replace them, while the tool was not yet good enough to justify the trust the plan required.
None of this means the idea is gone. Reuters reports Zuckerberg only promised employees that company-wide cuts would not happen this year, which still leaves room for smaller reductions now or a larger push in 2027. Meta continues to run AI-assisted teams in parts of the business and is on pace to spend more than 70 billion dollars on AI infrastructure this year alone.
The bigger lesson extends past Meta. Amazon, Salesforce, and other large employers have already cited AI as a factor in recent job cuts, and Wall Street is watching every company's AI spending for signs it produces real savings. Meta's experience shows what happens when that pressure moves faster than the technology: leaders overestimate how soon AI can replace real work, and the way they collect data to build that AI can damage trust faster than any layoff announcement. Any company weighing similar plans should treat both risks as connected, not separate.