Meta spent much of 2026 preparing for a radical transformation of how its employees work, with artificial intelligence at the center of the plan.
The company called the effort Project OT, short for Organization Transformation. Internal planning documents reviewed by Reuters show that Meta wanted AI agents to take over much of the routine work handled by thousands of employees.
The plan went further than simply giving workers better AI tools.
Meta envisioned smaller teams of highly skilled employees supervising virtual AI workers. In scenario-planning exercises, executives considered reducing the size of some teams by as much as 60%.
The company planned to carry out the restructuring in two waves. The first would come in May, followed by another round in November.
That second wave never happened.
Meta ultimately carried out a 10% workforce reduction in May. But hours before that first round, CEO Mark Zuckerberg called off planning for the November cuts, according to an internal document reviewed by Reuters.
The reversal came as employees pushed back against the transformation and internal data raised concerns about whether AI agents were delivering the productivity gains Meta expected.
Meta has now confirmed the existence of Project OT. The company said the year-long effort focused on cutting costs, redesigning teams and moving employees toward priority areas, including work that supports its AI models.
Meta also confirmed that some of the most aggressive scenarios involved reducing individual teams by up to 60%. However, the company said those scenarios included both layoffs and redeployments and never represented a plan to eliminate 60% of its entire workforce.
The distinction matters.
Meta did not abandon its broader AI strategy. Instead, the company pulled back from some of its most aggressive assumptions about how quickly AI could reshape its workforce.
The technology itself appears to have played a major role in that decision.
AI agents are supposed to do more than generate text or answer questions. They can take actions, complete tasks and operate across workflows with less human intervention.
Meta wanted to build its workforce around that model.
Internal experiments, however, exposed problems. Reuters reported that employees encountered workflow disruptions and technical issues, while internal data suggested the autonomous agents were not producing the expected productivity improvements.
The reaction inside Meta also became a problem.
Employees increasingly viewed the AI transformation as a threat to their jobs. Some workers were concerned that Meta was asking them to help build systems that could eventually replace them.
That created an awkward situation for a company trying to present itself as one of the world’s leading AI developers.
Meta was asking employees to embrace AI while simultaneously considering a major reduction in the number of people needed to perform their work.
Zuckerberg has since shifted the company’s public messaging.
Rather than emphasizing AI as a tool for eliminating human work, Meta has increasingly presented the technology as a way to empower employees and smaller teams.
The company recently launched a public campaign built around the message that it is “betting on people,” while Zuckerberg has promoted a vision in which AI gives individuals more power to build products and companies.
That does not mean Meta has lost confidence in AI.
The company continues to spend enormous amounts of money building AI infrastructure and recruiting top researchers and engineers. It also continues developing autonomous agents and other systems designed to perform increasingly complex tasks.
The problem is the gap between what those systems might eventually do and what they can reliably do today.
That gap has become increasingly important as technology companies try to justify massive AI investments.
Meta is betting that AI will transform everything from product development to advertising and workplace productivity. But Project OT shows that transforming an organization around AI can create problems that are difficult to solve with better models or more computing power.
Reducing headcount is relatively straightforward.
Replacing the work those employees perform is not.
Meta’s experience also highlights a broader challenge facing companies racing toward AI-native operations. A system can look impressive in demonstrations and still struggle when it has to operate inside complicated real-world workflows.
Those workflows involve security, accountability, institutional knowledge and decisions that cannot always follow a predictable sequence.
Meta encountered some of those problems firsthand.
The company therefore finds itself in an unusual position.
It remains one of the biggest corporate investors in artificial intelligence, yet its own attempt to reorganize around AI appears to have moved faster than the technology could comfortably support.
Project OT has not disappeared entirely. Meta continues to experiment with smaller teams, AI-assisted workflows and new organizational structures.
But the most aggressive version of the plan has already lost momentum.
For Meta, that may prove less like an abandonment of its AI strategy and more like an adjustment to reality.
The company still believes AI can reshape the workplace.
It just appears to have discovered that reshaping a workforce around AI is considerably harder than building the AI itself.
