Inside Meta’s Project OT that could have cut some teams by 60% and why Mark Zuckerberg scrapped it

Meta CEO Mark Zuckerberg had a sweeping plan to transform the company around artificial intelligence — one that could have cut some teams by as much as 60%.

Codenamed Project OT, short for “Organisation Transformation,” the internal initiative envisioned an “AI-native” Meta where AI agents would handle a large share of everyday work, while smaller, highly skilled teams supervised them.

The plan was ambitious. And potentially brutal.


As per Reuters, which cited internal planning documents, there were two waves of restructuring, with one scheduled for May and another for November. Some teams could have been reduced by up to 60%, through a combination of layoffs, closed positions and employee redeployments.

But Zuckerberg ultimately abandoned the second wave.

Meta wanted smaller teams built around AI

Project OT was part of a broader push to rethink how Meta operated as AI rapidly changed the workplace.

The idea was to create smaller, “talent-dense” teams that could use AI agents to accomplish work previously handled by much larger groups of employees. Some staff would be moved into priority areas, including work related to AI training data, while others could lose their jobs.

The potential scale of the restructuring was significant enough to draw comparisons with Meta's earlier major workforce cuts.

Meta ultimately confirmed that it had considered scenarios involving reductions of up to 60% in some teams, while stressing that this was not a company-wide 60% cut and that not every scenario was implemented.

Zuckerberg changed course before the second wave

The first round of cuts went ahead. Meta reduced its workforce by about 10% on May 20, according to reports.

But hours before the cuts, Zuckerberg reportedly consulted senior executives and scrapped the planned November wave of Project OT restructuring.

The reversal came as resistance inside Meta grew.

Employee sentiment reportedly plunged from 74% favorable to 55% favorable in an internal Pulse survey, amid concerns that the initiative was fundamentally about replacing human workers with AI.

There was another problem: the AI agents weren't producing the productivity gains Meta had hoped for.

Employees were using AI more heavily, and code output increased sharply, but internal discussions reportedly suggested that higher AI-assisted output wasn't translating into an equivalent jump in overall productivity.

Investors were also beginning to question the enormous sums Meta was pouring into AI.

The strange twist: AI itself became part of the debate

Perhaps the most striking element of Project OT was that AI wasn't simply the technology being deployed — it was also being used to help evaluate the transformation.

Reports indicated that AI models were used to assess internal scenarios and potential employee feedback. Those systems reportedly highlighted the possibility of resistance and complications from such a sweeping transition.

Meanwhile, Meta was experimenting with ways to train AI agents to mimic human computer use. U.S. employees were reportedly asked to install tracking software capable of recording keystrokes and mouse activity for this purpose, a move that triggered anger and dark humor internally.

By late May, some engineering groups had already shrunk by as much as 30% through layoffs and reassignment.

Yet the larger November transformation never arrived.

Instead, Zuckerberg began publicly emphasizing a different message: Meta was “betting on people.”

Project OT showed just how far Meta was willing to go in imagining an AI-first company — and how quickly that vision could collide with employee resistance, uncertain productivity gains and investor pressure.

For Zuckerberg, the experiment didn't necessarily mean abandoning AI.

It meant that turning one of the world's largest technology companies into an AI-native organization might be considerably harder — and messier — than simply replacing people with agents.