Key Takeaways

• A confidential initiative dubbed “Project OT” envisioned eliminating up to 60% of certain departments across two phases to build an “AI native” organization.
• Hours before initiating the first round of terminations on May 20, Zuckerberg abandoned plans for the second wave following internal resistance.
• The company’s AI agent systems underperformed expectations, contributing to a 40% surge in critical incidents.
• The social media giant ultimately reduced its workforce by 10% in May, with leadership pledging no additional broad-scale reductions through year-end.
• Analysts project Meta’s planned $130 billion AI infrastructure investment for 2026 could absorb the company’s entire operational cash generation.

Meta Platforms (META) pursued an ambitious plan to reshape its workforce around artificial intelligence, but the results fell well short of expectations and exposed the gap between internal AI ambitions and day-to-day execution.

Meta Platforms, Inc., META

A Reuters investigation published this week revealed details of “Project OT” — short for Organization Transformation — a restructuring blueprint developed during meetings at Zuckerberg’s Hawaiian estate in January. The plan was designed to turn Meta into an “AI native” enterprise, with intelligent agents handling a large share of tasks traditionally performed by human employees.

During strategy sessions, company leaders examined scenarios that would cut staff by as much as 60% in certain departments. The proposed transformation followed a two-step approach: initial terminations planned for May and a second round scheduled for November. An internal human resources leader estimated that the combined cuts could match or exceed the 25% workforce reduction Meta carried out three years earlier.

The Strategy Unravels

On the evening of May 19, just hours before the first phase was set to begin, Zuckerberg reversed course on the planned November reductions. Meta still carried out the May 20 terminations, which eliminated 10% of staff positions, but the more aggressive second wave was later shelved. The abrupt shift underscored how quickly the company’s restructuring plans were being revised as employee pushback and operational concerns mounted.

Resistance inside the company was substantial. Employees flooded Meta’s internal communication channels with complaints. The company’s biannual Pulse survey showed internal approval falling from 74% to 55%. Reuters reported that some engineering divisions saw headcount declines of nearly 30% by late May.

Meta also deployed monitoring software on U.S. employees’ computers to record keystrokes and cursor movements for AI model training. Many workers concluded they were helping build the systems that could replace them, and labor organizing efforts gained momentum.

The technology itself also failed to meet expectations. While code generation rose 220% year over year, new functionality delivered to end users increased by only 36%. Critical technical and security issues rose 40%, while employee time spent resolving incidents climbed 70%. For a company trying to use AI to increase output and reduce human labor, those figures suggested the tools were not yet absorbing the workload management had hoped they would.

In early June, cybercriminals exploited vulnerabilities in Meta’s AI-driven customer service chatbot to compromise prominent Instagram accounts. Meta declined to answer questions about the internal performance metrics.

Zuckerberg’s Response

At an internal company meeting in early July, Zuckerberg acknowledged that the implementation timeline had slipped. He said the AI agent capabilities had not “accelerated” at the pace he had anticipated and told staff he expected meaningful progress in the next three to six months.

After the May workforce reduction, Zuckerberg sent an internal message saying he did “not expect other company-wide layoffs this year.” Leadership later suspended the keystroke-monitoring initiative, allowed some employees to return to prior roles, and promised improved workplace amenities.

Meta acknowledged Project OT in its response to Reuters, describing it as a year-long effort focused on cost reduction, organizational redesign, and reallocating resources toward strategic priorities. The company said it never considered eliminating 60% of total headcount and that many major divisions were excluded from the planning scenarios.

Meta’s AI spending continues to draw investor attention because the company is still committing heavily to the infrastructure needed to support these systems even as the rollout remains uneven. The company plans to invest at least $130 billion in AI chips and infrastructure in 2026. LSEG analysts estimate that this level of spending would consume all of Meta’s operating cash generation for that fiscal year.

META stock was trading near $714 at the time of reporting.