Mark Zuckerberg spent years arguing that artificial intelligence would transform how people work. In early 2026, Meta reportedly tried to put that belief to a much bigger test. The company explored rebuilding parts of its workforce around AI agents, with fewer employees handling work that once required much larger teams.
However, the experiment did not go as planned. According to a detailed Reuters investigation, Meta’s internal Project OT considered shrinking some teams by as much as 60% through layoffs, redeployments, and closed positions. Meta went ahead with major cuts in May, but plans for another company-wide round were abandoned as the ambitious restructuring ran into resistance and questions about AI performance.
Project OT stood for Organization Transformation, and its goal went far beyond giving workers another AI assistant. Meta wanted to become what internal planning material described as an “AI native” company. AI agents would perform a large share of everyday tasks, while smaller groups of employees would supervise the systems and handle more complex work.
The idea took shape in January during a leadership gathering at Zuckerberg’s Hawaii property, Reuters reported. Meta examined several restructuring scenarios, including some that could reduce individual teams by up to 60%. The company later stressed that this figure did not mean it planned to eliminate 60% of its total workforce.
Meta’s AI Workforce Experiment Quickly Hit Reality
SiiLa / The theory behind Project OT sounded attractive from a business perspective. If AI agents could write code, complete routine assignments, and automate internal processes, Meta could operate with smaller teams.
The company could then redirect money and talent toward its biggest priorities, particularly its costly race to develop advanced AI.
Reality proved much messier. Internal figures cited by Reuters showed that code changes made to Meta’s internal software platforms and infrastructure jumped 220% from the previous year. Yet changes that actually produced new or improved features for Meta users increased by only 36%. More activity was not translating into comparable gains for the people using Meta’s products.
The problems became more serious when reliability entered the picture. Internal posts reviewed by Reuters described AI agents carrying out large-scale actions that human employees would be unlikely to perform. Major technical and security incidents reportedly increased by 40% compared with the previous year, while employee time spent dealing with those problems rose by as much as 70%.
That created an awkward form of AI productivity. The technology could complete tasks quickly, but humans still had to watch its work and repair problems when something went wrong. Instead of removing work completely, AI sometimes shifted employees toward checking, correcting, and containing automated actions.
Zuckerberg himself later acknowledged that progress had fallen short of expectations. Reuters reported that he told employees in July that the trajectory of agent development during the previous four months had not accelerated as expected.
Meta Employees Feared Training Their AI Replacements
Maria / Unsplash / Project OT caused deep anxiety among employees who saw AI moving from a workplace tool into something that could directly affect their jobs.
Some workers were shifted into assignments that were training material for Meta’s AI systems.
That created an uncomfortable situation for staff. Employees feared that improving the technology could eventually reduce the need for their own roles. Reuters reported widespread internal resistance as workers tried to understand which jobs might disappear and how deeply the company intended to restructure its teams.
Meta also introduced software to capture computer activity from some US employees as part of efforts to train AI systems to imitate how humans use computers. The monitoring added another source of tension inside a company already dealing with job insecurity and major organizational changes.
Some employees learned important details about the possible restructuring through media reports instead of hearing them first from company leaders. That fueled anger and uncertainty at a moment when workers were already questioning what an “AI native” Meta might mean for their careers.
Meta still carried out substantial layoffs in May. Around 10% of its workforce was cut, while thousands of other employees were moved into priority areas. However, Zuckerberg halted planning for another company-wide round that had been expected later in the year. He subsequently told employees that he did not expect further company-wide layoffs in 2026.
