In January, Mark Zuckerberg and his top executives gathered at his Hawaii compound for their annual leadership retreat. What they planned there never made it to the public. Until now.
Reuters published a special report on Aug. 26, based on scores of internal documents, recordings, and more than 20 interviews, that lays out what Zuckerberg was actually trying to do and why it fell apart.
Meta Project OT plan to replace employees with AI agents
The initiative was code-named Project OT, short for Organization Transformation. The plan envisioned an “AI native” future for Meta, in which AI would take over much of the work performed by thousands of human employees, with smaller, “talent-dense” groups of human staff overseeing virtual workers, Reuters reported.
Some teams were being evaluated for cuts of as much as 60%. The restructuring was designed in two waves. The first would begin in May. The second would follow in November, targeting a broader set of roles across the company, according to Reuters.
More Mark Zuckerberg:
- Mark Zuckerberg, Microsoft CEO just made major AI decision
- Zuckerberg’s stock rebound made him world’s 5th richest at $222B
- Mark Zuckerberg backs Elon Musk Silicon Valley decision
Meta employed roughly 79,000 people at the time. A 20% reduction, which was among the scenarios being discussed, would have eliminated around 16,000 jobs, according to Barchart.
On the night of May 19, hours before the first wave of layoffs was scheduled to begin, Zuckerberg pulled back. Meta proceeded with roughly 8,000 job cuts, about 10% of its workforce, the following day. The November wave was canceled.
Zuckerberg subsequently told remaining employees he did “not expect other company-wide layoffs this year,” Reuters noted.
Why Meta’s AI replacement plan unraveled
The reversal came as internal data was raising serious questions about whether the technology was working.
Employees were generating far more code using AI tools. Code changes to Meta’s software platforms and infrastructure rose 220% year over year. But changes resulting in new or upgraded features that users could actually see rose just 36%, IBTimes reported.
The productivity gap was only part of the problem. Meta had also introduced tracking software on U.S. employees’ devices, capturing keystrokes and mouse clicks to train AI agents to reproduce human workflows.
The move contributed to what Reuters described as an open revolt among staff, who had become convinced that the company’s AI transformation initiatives were partly aimed at replacing them.
Meta’s Chief Technology Officer Andrew Bosworth clashed with some employees about how the company was handling the transition. Internally, Meta was simultaneously trying to identify what it called “Irreplaceable Talent,” a designation intended to protect key personnel from cuts, even as it was exploring broad reductions elsewhere, according to The Globe and Mail.
Meta confirmed the existence of Project OT after Reuters presented its findings, describing it as a yearlong project focused on cost-cutting, redesigning team structures, and shifting staff into new priority areas, including producing training data for its AI models.
Benjamin/Getty Images
Zuckerberg’s CEO agent and Meta’s AI strategy
While the company was pulling back from its most aggressive workforce reduction plans, Zuckerberg was building something for himself.
He has been using a personal AI agent, described internally as a “CEO agent,” that allows him to retrieve answers he would typically have to go through multiple layers of staff to get, Euronews reported.
That detail is not incidental. If the CEO’s own AI agent can compress decision chains that previously required layers of staff, the same logic applies everywhere in the organization. Zuckerberg told analysts on a January earnings call that he was already seeing projects that used to require big teams now accomplished by a single talented person, according to Barchart.
Meta’s internally developed AI-native groups have reporting structures with as many as 50 individual contributors per manager. Employee use of AI tools is now factored into performance reviews. The direction of travel is clear, even if the pace has slowed.
What Meta’s AI workforce shift means for employees and investors
For employees, the lesson from Project OT is that the risk of AI displacement at Meta is real and ongoing, even if the most aggressive version of the plan was walked back. The company is still restructuring around the premise that AI can absorb work that humans currently perform.
For investors, the more pressing question is the productivity problem the internal data revealed. Meta has spent aggressively on AI. Code generation is up dramatically. But the products users actually see have not kept pace.
That gap between AI investment and user-facing output is the number that matters most heading into Meta’s next earnings cycle.
The Oakland trial over alleged addictive design targeting children is already weighing on the stock, as TheStreet reported.
A workforce strategy that generated internal unrest and produced mixed productivity results adds another layer of complexity to a company that is simultaneously fighting a major legal battle and trying to remake itself around technology it has not yet fully proven.
Related: Mark Zuckerberg, Microsoft CEO just made major AI decision
