Meta planned to cut some teams by 60% and replace the work with AI

Meta spent the first half of this year planning to replace much of the work its employees do with AI agents, then abandoned the most aggressive part of the plan when the technology did not deliver. Meta code-named the scheme Project OT, short for Organization Transformation. It explored cutting the size of many teams by as much as 60%.

Katie Paul reported the plan for Reuters, working from scores of internal documents, posts and recordings, and interviews with more than 20 people with knowledge of the company.

Mark Zuckerberg and his senior team hatched it at his Hawaii compound in January. AI would take over daily work performed by thousands of employees. Smaller cadres of what one planning document called “talent-dense” staff would supervise the virtual workers.

Two waves, one cancelled

The plan ran in two stages. A first purge would come in May, and a second shake-up in November. Layoffs would sit alongside the closing of open positions and the removal of people Meta rated as poor performers.

One human-resources executive projected a culling as big as or bigger than the roughly 25% Meta cut three years ago, according to an internal document.

On the night of 19 May, hours before the first wave, Zuckerberg called off the November planning. Meta went ahead the next day and cut about 8,000 jobs, roughly 10% of its staff. It moved 7,000 people into AI roles in the same week.

Reuters could not establish what changed Zuckerberg’s mind.

Meta confirms the plan

The company acknowledged Project OT. It described it as a year-long effort focused on cutting costs, redesigning team structures and moving staff into priority areas such as producing training data.

Meta also confirmed the two-wave structure and the 60% figure. It said the number covered scenarios for certain teams only. Several major units sat outside the exercise, it said, and it never intended to cut 60% of the whole company. Leaders cancelled the second wave before working out how many people would lose their jobs.

“This ultimately resulted in moving thousands of employees to do priority work on several newly-established teams, as has been publicly reported. Ultimately, we didn’t move forward with every scenario from the exercise, and it was never assumed we would,” Meta said in a statement.

Where the idea came from

Meta executives spent last year studying how AI startups organise themselves. Chief data officer Alex Schultz and head of product Naomi Gleit both visited Asia, and three people said they admired how startups there had built their org charts around AI.

Gleit told Reuters in June that she had spent “quite a bit of time” in Meta’s Singapore office, and that the practices there had “inspired some of the teams in California and New York”. She said many of the ideas were bottom-up.

The numbers that undercut it

Meta’s own data showed the agents were producing volume rather than value. Code changes to internal platforms and infrastructure rose 220% year on year, according to a June post by chief technology officer Andrew Bosworth. Changes that reached users as new or upgraded features rose 36%.

The gap did not stay harmless. An April post said unchecked AI agents were performing “large-scale, disruptive actions that humans are unlikely to execute”. Major technical and security incidents, including service disruptions and possible data leaks, climbed 40% on the previous year. Time spent firefighting them rose 70%.

Infrastructure teams had flagged reliability warning signs as early as March. In June, hackers exploited Meta’s new AI customer support bot to reach high-profile Instagram accounts, including the dormant Obama White House page.

Meta declined to comment on the internal data.

The staff worked it out

Employees became convinced they were training their replacements. Meta had reassigned engineers to write software puzzles used as training data, and many called the work rote in internal posts. It had also mandated tracking software on US employees’ devices to capture keystrokes and mouse clicks, so agents could learn to use a computer like a person.

Meta paused that programme in June. Twenty-six employees have since sued the company, saying its AI systems targeted workers on medical leave for the layoffs.

Staff replied to executive posts with pictures of elephants. Meta’s internal sentiment score fell from 74% favourable to 55% in its half-year Pulse survey. Labour organising gathered pace.

What replaced the org chart

An internal post titled the AI-Native Playbook set out the target structure. Product teams of 10 to 20 specialists would shrink to pods of three to five. Engineers, designers and product managers would share one title, “builder”. Middle management layers would go, and “agent-assisted analysis” would set daily priorities.

By June at least 11 units had implemented pods. Unit heads would oversee 30 to 50 people each, supported by human-resources staff and unspecified “AI systems”. Meta declined to explain that phrase, but said performance ratings and promotions “were and are made by people, not AI”.

Zuckerberg concedes the timing

At a July town hall the chief executive conceded he had misjudged the timing. The “trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected”, he told staff, in remarks reported by Engadget. He said he still expected benefits within three to six months. Bosworth had already told employees in August to stop asking for time off on the strength of AI gains.

The public message has changed shape. Meta now runs advertising saying it is betting on people. Zuckerberg’s recent 6,500-word essay predicts an abundance of jobs, while allowing that individual companies may employ fewer. “Company sizes may shrink, just as they did in the transition from industrial giants to tech companies,” he writes.

The spending has not changed. Meta plans to invest at least $130bn in AI chips and infrastructure this year. Its second-quarter filing listed 75,472 employees at the end of June, a figure that still included about 8,000 people leaving under the May cut, RuntimeWire reported from the document.

Zuckerberg has promised no further company-wide layoffs this year. That wording leaves team-level cuts and 2027 open, and employees have noticed. The number that will settle the argument is the one Meta stopped publishing: whether feature output ever catches up with the code its agents produce.