Meta considered cutting some teams by 60 per cent as AI restructuring grows

MUMBAI: Meta may have wanted to make its teams smaller, but the AI experiment is proving bigger than expected. The Facebook, Instagram and WhatsApp parent considered scenarios that could have reduced the size of some teams by as much as 60%, while shifting parts of their work to AI agents, according to a Reuters report based on internal documents and people familiar with the company’s operations.

The plans formed part of Project OT, or Organization Transformation, an internal exercise launched amid senior leadership discussions in early 2026 to rethink how Meta organised its workforce. The company ultimately did not implement every scenario examined.

Meta cut around 8,000 jobs in May, equivalent to roughly 10 per cent of its workforce at the time. A second planned round of reductions was reportedly abandoned shortly before it was due to begin.

One proposal under Project OT would have reshaped conventional product teams of 10 to 20 specialists into groups of just three to five employees. These smaller teams would focus on building and supervising products, while layers of management were reduced.

Meta also examined which employees were particularly difficult to replace as it reassessed its staffing needs.

The company has confirmed that Project OT existed, but stressed that the scenarios were exercises rather than predetermined plans. Meta told Reuters that teams had been asked to model the potential impact of redeployment, vacant positions and workforce reductions, with not every scenario eventually pursued.

The restructuring comes as Meta tries to turn AI agents into a genuine workplace force, rather than simply another tool in the software toolbox.

Executives had expected AI to significantly accelerate software development and other technical work. Internal figures, however, reportedly showed a sizeable gap between the amount of code being generated and the amount that actually translated into improvements for users.

Meta chief technology officer Andrew Bosworth reportedly said in June that code changes to internal platforms had jumped 220 per cent year on year, while changes resulting in new or improved user features rose by 36 per cent.

Internal posts also pointed to reliability and operational issues, with technical and security incidents reportedly increasing and engineering teams spending more time addressing problems.

At a July town hall, CEO Mark Zuckerberg acknowledged that AI-agent development had not advanced as quickly as he had expected and indicated that the technology needed several more months of development.

The AI push also ran into resistance from employees.

A workplace monitoring programme for US employees, which reportedly captured information including keyboard activity, mouse movements and computer screens to help train AI systems for workplace tasks, drew opposition from staff. More than 1,000 Meta employees reportedly signed a petition against the tracking.

Internal employee sentiment weakened during the period, with favourable responses in Meta’s Pulse survey falling from 74 per cent to 55 per cent.

Meta has since paused the employee tracking programme, according to the report.

Despite the organisational uncertainty, Meta is not putting the brakes on its AI ambitions.

The company is committing at least $130 billion to AI infrastructure this year while continuing to recruit specialised AI researchers and engineers.

Its Superintelligence Labs has also undergone changes, including a separate reduction of around 600 positions, even as the unit continues hiring specialist talent as Zuckerberg seeks to strengthen Meta’s position in advanced AI.

That leaves Meta walking a curious tightrope: fewer people in some parts of the organisation, more AI spending and more hiring in others.

Project OT ultimately underlines the difficulty of turning AI promises into workplace economics. Smaller teams and AI agents could reshape how Meta operates, but the reported retreat from some of the more aggressive plans suggests the technology still has to prove that it can deliver consistent productivity gains without adding new reliability, security and workforce-management headaches.