A Reuters investigation published today reveals the full scope of Meta’s internal plan to replace human workers with AI — and how it collapsed under the weight of employee rebellion, underperforming AI agents, and a productivity paradox that exposed the gap between AI hype and AI reality.
The investigation, based on scores of internal documents, posts, and recordings reviewed by Reuters plus conversations with more than 20 people familiar with Meta’s operations, details Project OT — short for Organization Transformation. The plan, hatched at Zuckerberg’s Hawaii retreat in January, envisioned an “AI native” future where AI would take over much of the daily work performed by thousands of employees, supervised by smaller “talent-dense” teams of humans.
The plan: 60% cuts in two waves
In scenario-planning exercises, executives explored slashing the size of many teams by as much as 60%, according to three people familiar with the project. The restructuring would come in two waves — a first purge in May and a second in November. One HR executive projected the culling would be as big as or bigger than Meta’s 25% cuts three years ago.
Meta confirmed the existence of Project OT to Reuters, describing it as a year-long project focused on cost cutting, redesigning team structures, and shifting staff into new priority areas. The company acknowledged the plan was to be carried out in two waves and that the most drastic scenarios involved reducing some teams by up to 60%. But it said several major units were not part of it and that the company never intended to lay off 60% of its entire workforce.
The internal “AI-Native Playbook” document laid out a radical restructure: traditional product teams of 10-20 people would shrink to pods of 3-5. Specialised roles — product managers, designers, data scientists — would be replaced by generic “builders” sharing pooled specialists. Middle management layers would be eliminated. Performance ratings and promotions would be decided by unit heads supported by unspecified “AI systems.”
The night Zuckerberg blinked
On the night of May 19, hours before the first layoff wave, Zuckerberg changed course. Meta proceeded with a 10% cut the next day but cancelled planning for the November reductions. The reversal came amid mounting pressure from three directions.
First, employees were in open revolt. Staff flooded Meta’s internal communication platform with angry posts and gallows humour, posting elephant images to signal that layoffs were the elephant in the room. Morale, measured by Meta’s internal Pulse survey, dropped from 74% favourable to 55%. Labour organising efforts gained momentum after the company mandated tracking software on employee devices to capture keystrokes and mouse clicks for AI training.
Second, the AI agents at the heart of the strategy were not delivering. Internal data showed that while AI-assisted code changes to internal platforms surged 220% year-over-year, changes that led to new or upgraded features reaching users rose only 36%. Infrastructure teams warned of “reliability warning signs.” Unchecked AI agents performed “large-scale, disruptive actions that humans are unlikely to execute.” Major technical and security incidents spiked 40%, with firefighting time up 70%.
The public saw a glimpse of the problem in June when hackers exploited Meta’s AI-powered customer support bot to access high-profile Instagram accounts, including the dormant Obama White House page.
Third, investors were questioning what Meta had to show for its gargantuan AI spending. The company plans to invest at least $130 billion in AI infrastructure this year, which analysts expect will consume its entire 2026 operating cash.
The productivity paradox
The Reuters findings align with a broader pattern documented in academic research. A study by the Atlanta Federal Reserve found that approximately 90% of executives believe AI has not yet boosted productivity at their companies, as reported by Fortune on August 22.
Research from the University of Pittsburgh, published in the same Fortune piece, analysed millions of Glassdoor reviews and found that AI-related layoffs actively destroy the conditions needed for AI to make workers more efficient. Employee sentiment toward AI is one of the strongest predictors of firm productivity when AI is used — and layoffs cause a sharp decline in that sentiment.
At Meta, this dynamic played out in real time. The company asked employees to adopt AI tools while simultaneously citing those same tools as grounds for job cuts. The result was a demoralised workforce and underwhelming returns on AI investment — precisely the outcome the Pittsburgh researchers predicted.
What Meta said publicly versus privately
After cancelling the second wave, Zuckerberg told remaining employees he did “not expect other company-wide layoffs this year” and expressed a desire to give them more “stability.” In July, he made a surprise appearance at an internal town hall and conceded that AI agent technology had not “accelerated” as quickly as anticipated, as we reported at the time.
Meta then launched a public-relations blitz positioning the company as people-centric, including a video advertisement proclaiming Meta is “betting on people” and a 6,500-word essay by Zuckerberg titled “The Future is for Everyone.” In the essay, Zuckerberg predicts “an abundance of jobs in the future” but acknowledges that “company sizes may shrink” — framing this as a transition to “a larger number of companies with fewer people each.”
Some employees remain sceptical. Zuckerberg’s careful use of “company-wide” and “this year” has prompted speculation that team-specific cuts or performance-based dismissals may continue, or that broader reductions could return next year.
The career lesson: AI replacement is not automatic
For workers worried about AI displacement, the Meta story offers an instructive counter-narrative. One of the world’s most AI-ambitious companies, with effectively unlimited resources, found that replacing humans with AI agents was far harder than the PowerPoint decks suggested. The agents generated volume but not quality. They created new problems faster than they solved old ones. And the human cost — plummeting morale, organised resistance, security incidents — eroded whatever gains might have materialised.
This does not mean AI poses no threat to jobs. Challenger, Gray & Christmas reports that AI has been the leading reason for US job cuts for five consecutive months, with 112,713 cuts attributed to AI so far in 2026. But it does mean the threat is more nuanced than “AI replaces worker, company profits.” The companies most aggressive about AI-driven layoffs may be the ones least equipped to manage the fallout — a pattern worth remembering for anyone navigating an AI-disrupted career.
As we noted in our coverage of the AI layoff productivity paradox, companies that invest in skills and expand opportunity rather than simply cutting headcount are the ones most likely to profit from AI. The Meta investigation is the most detailed case study yet of what happens when a company tries the shortcut.
❓ FAQ
What was Meta’s Project OT? An internal plan to restructure the company around AI, code-named Organization Transformation. It envisioned replacing traditional teams with small AI-equipped pods and explored cutting some teams by up to 60% in two waves of layoffs.
Did Meta go through with the plan? Partially. Meta executed a 10% workforce reduction on May 20 but cancelled the planned second wave in November. CEO Mark Zuckerberg said AI agent technology had not progressed as quickly as expected.
Were AI agents actually replacing workers at Meta? Internal data suggests AI increased code volume dramatically but did not translate into proportional feature output. AI agents also caused a 40% spike in major technical incidents, raising questions about their readiness for autonomous work.
What does this mean for workers? The Meta case shows that AI-driven workforce replacement is not straightforward even for well-resourced companies. However, AI remains the leading cited reason for US job cuts. The lesson is to focus on roles where AI augments human judgement rather than replacing routine tasks.
🔍 THE BOTTOM LINE
The Reuters investigation into Project OT is the most detailed account yet of a major corporation attempting to go “AI native” — and stumbling. Meta explored cuts of up to 60% in some teams, installed keystroke-tracking software to train AI replacements, and watched its AI agents generate a 220% surge in code changes that produced only a 36% increase in shipped features while spiking security incidents by 40%. The company pulled back, but the strategic pressure remains: $130 billion in AI spending demands results. For workers, the takeaway is that AI replacement is neither inevitable nor clean. The companies most aggressive about cutting humans often discover, as Meta did, that the technology is not ready to do what the org chart assumed it could.
