Meta Was Hours Away From Another Massive AI Layoff Wave, Then Zuckerberg Pulled the Plug

Mark Zuckerberg was hours away from starting Meta's first major AI-driven workforce restructuring when he made an unexpected change. The first round of layoffs would still happen. But a second wave, planned for later in the year and potentially much larger, was abruptly cancelled the night before the cuts began.

The decision centred on Project OT, short for Organization Transformation, a confidential programme designed to make Meta an AI-native company. Internal planning documents reviewed by Reuters showed executives had considered reducing some teams by as much as 60%. The restructuring was planned in two phases. The first was scheduled for May, while the second was planned for November.

But on the night before the May layoffs, Zuckerberg cancelled the November phase. The reversal came as Meta faced questions about whether its AI systems were producing the productivity gains needed to support such an aggressive workforce overhaul.

Project OT Was Designed Around AI

Project OT emerged after a leadership retreat in Hawaii in January. The ambition was straightforward but far-reaching: AI agents would take over much of the routine work previously performed by thousands of employees, while smaller groups of workers would supervise and work alongside the systems. The plan also envisaged fewer management layers and smaller teams with broader responsibilities.

Executives considered cuts of up to 60% within some teams. One internal projection suggested the overall workforce contraction could have been comparable to, or greater than, Meta's roughly 25% reduction during its previous major restructuring. That figure did not represent a confirmed 60% company-wide cut. The plans involved a combination of potential layoffs, redeployments, and unfilled roles. Meta ultimately cut about 10% of its workforce in May. The planned November phase did not go ahead.

AI Productivity Raised Questions

Internal data later highlighted a gap between AI-assisted activity and the results reaching users. One measure showed code changes to Meta's software platforms and infrastructure had risen 220% year on year. Yet new or upgraded features reaching users increased by only 36%.

The difference mattered. More code did not necessarily mean more useful products. Internal posts also indicated that major technical and security incidents had increased by 40%, while employee time spent dealing with those problems rose by as much as 70%.

Meta declined to comment on the internal figures reported by Reuters. Zuckerberg later acknowledged that AI agent development had not accelerated as quickly as expected. At a July company meeting, he said the company's investments had not yet produced the anticipated results. The comments suggested that Meta's AI ambitions were moving faster than the technology's ability to deliver every expected benefit.

Employees Were Already Pushing Back

The restructuring was also unfolding amid growing employee concerns. Meta had introduced software that monitored keyboard and mouse activity on some US employees' computers as part of efforts connected to AI development. The move fuelled concerns that workers were being monitored while helping develop systems that could eventually reduce the need for human labour.

Employee sentiment also deteriorated. Reuters reported that Meta's internal Pulse survey recorded approval falling from 74% to 55%. Meta later paused the mouse-tracking programme. The reaction added another challenge to an already sensitive transformation.

The Cost of Betting on AI

Meta's AI ambitions come with an enormous price tag. The company expects to invest $130 billion or more in AI infrastructure and chips this year. The spending is intended to provide the computing capacity needed to develop and operate increasingly powerful AI systems. The business case partly depends on productivity. If AI allows smaller teams to accomplish more, Meta could potentially increase output while controlling costs.

But the internal figures show that greater AI-assisted activity does not automatically translate into greater productivity. That leaves Zuckerberg balancing two pressures: spending enough to remain competitive in AI while ensuring those investments deliver tangible results.

The Layoffs May Have Been Delayed, Not Forgotten

For Meta employees, cancelling the second wave offered immediate relief. Zuckerberg has also told staff that he does not expect other company-wide layoffs this year. But Project OT has not simply disappeared.

Meta continues to experiment with smaller teams, AI-assisted work, and organisational structures in which employees work alongside AI systems. The episode therefore does not show Meta abandoning artificial intelligence. Instead, it highlights the difficulty of converting ambitious AI predictions into immediate workforce changes.

Meta may still use AI to reshape thousands of jobs. But the abandoned November cuts suggest the company has discovered that replacing work with AI is more complicated than simply replacing workers. For Zuckerberg, the decision bought Meta more time to prove that its AI investment can deliver the productivity gains behind the strategy.