Meta explored 60 per cent team cuts as AI agents reshape its workforce plans

MUMBAI: Meta is thinking smaller, at least when it comes to some of its teams. The technology giant explored organisational scenarios that could have reduced the size of certain teams by as much as 60 per cent, with AI agents taking on a larger share of the work, according to a Reuters report based on internal documents and interviews.

The exercise, known internally as Project OT, short for Organisation Transformation, formed part of a broader effort to rethink how work is structured across Meta, the parent company of Facebook, Instagram and WhatsApp.

The proposals reportedly emerged from discussions among senior leaders early in 2026. One scenario examined replacing conventional product teams of around 10 to 20 specialists with groups of just three to five employees, with AI systems handling a greater portion of development and operational tasks.

Employees in the smaller teams would primarily direct and oversee products, while AI agents would increasingly perform the work traditionally handled by larger groups.

The plans also looked at reducing management layers and identifying roles and employees considered particularly difficult to replace.

Meta confirmed to Reuters that Project OT existed, but said the scenarios were planning exercises rather than decisions that had already been made. Teams were asked to model the potential effects of redeployment, vacancies and workforce reductions, with not every scenario ultimately pursued.

The organisational planning came as Meta was already making significant changes to its workforce.

According to Reuters, the company carried out about 8,000 job cuts in May, equivalent to roughly 10 per cent of its workforce at the time. A second planned round of reductions was subsequently abandoned shortly before it was due to take place.

The developments underline the difficult balancing act facing Big Tech. Companies are pouring money into AI while simultaneously looking for ways to make their organisations leaner and more efficient.

At Meta, however, the push towards greater automation reportedly met resistance from employees.

One flashpoint was a workplace monitoring programme introduced for US employees. The initiative reportedly collected information including keyboard activity, mouse movements and computer screen data, with the stated aim of generating information that could help train AI systems to perform workplace tasks.

More than 1,000 Meta employees reportedly signed a petition opposing the monitoring measures.

Employee sentiment also weakened during the period. Data cited by Reuters showed favourable responses in Meta’s internal Pulse survey falling from 74 per cent to 55 per cent.

The debate around Project OT also exposed questions about whether AI was delivering the productivity gains that executives had expected.

Meta leaders had anticipated that AI tools could substantially accelerate software development and other technical work. Internal data cited by Reuters, however, reportedly showed a gap between the amount of code being produced and improvements visible to users.

Andrew Bosworth, Meta’s chief technology officer, reportedly said code changes to internal platforms had increased 220 per cent year on year, while changes that resulted in new or improved customer features rose 36 per cent.

In other words, the code was flowing much faster than the features were reaching users.

Internal posts reviewed by Reuters also reportedly pointed to increases in technical and security incidents, with engineering teams spending more time resolving problems.

That has complicated the argument that simply increasing AI-assisted output will automatically translate into greater productivity or business value.

Meta chief executive Mark Zuckerberg also appears to have adopted a more measured view of the technology’s immediate potential.

At a July town hall meeting cited by Reuters, Zuckerberg acknowledged that the development of AI agents was progressing more slowly than he had expected. He indicated that the technology could need several more months of development before reaching the capabilities Meta is targeting.

The comments suggest that Meta remains committed to AI agents, but recognises that turning the technology into a reliable replacement for large teams is more complicated than simply deploying new software.

The company’s spending plans show that the AI push itself is not slowing. Meta is expected to spend at least $130 billion on AI infrastructure this year while continuing to expand its AI engineering and research capabilities.

At Meta Superintelligence Labs, around 600 positions were reportedly cut even as the division continued hiring specialised AI researchers and engineers.

That split captures the current AI jobs story in a nutshell: fewer roles in some areas, fiercer competition for highly specialised talent in others.

Project OT therefore offers a glimpse into a possible future for corporate organisations, where a small number of employees could oversee fleets of AI agents instead of managing large human teams. But Meta’s experience also suggests that the maths of automation is not quite as simple as shrinking headcount and increasing code.

For now, employee resistance, reliability concerns and the still-evolving capabilities of AI agents are keeping some of the most aggressive restructuring scenarios on the drawing board. Meta’s bigger test will be whether its enormous AI investment can eventually produce the productivity gains needed to make that new organisational model work.