Meta’s plan to use AI agents in place of thousands of employees has been abandoned after the technology fell short and workers pushed back. The plan, developed under CEO Mark Zuckerberg, aimed to reshape the company around software agents that could take on human staffers’ roles.
The decision came after a difficult year for Meta employees. In May 2026, the technology company laid off roughly ten percent of its global workforce, affecting almost 8,000 people. Executives then considered another round of cuts, with teams facing reductions of as much as 60 percent.
That second wave of layoffs was called off. On August 26, 2026, Meta reported that it had abandoned the AI-focused restructuring plan known as Project Organization Transformation, or Project OT.
The ambitious plan behind Project OT
Meta executives developed Project OT after discussing Zuckerberg’s AI vision at an executive retreat at his Hawaii estate in January. The plan was far more than a push to add AI tools to existing jobs. It focused on having AI agents assume the roles of thousands of human staffers, changing how teams were organized and how much work people handled.
The proposed cuts showed the scale of the idea. Executives considered reducing some teams by as much as 60 percent, following the May layoffs that had already removed almost 8,000 positions. The approach tied Meta’s workforce plans directly to the company’s belief that AI agents would soon handle a wide range of tasks.
Meta later said the exercise did not lock the company into every option under discussion. “Ultimately, we didn’t move forward with every scenario from the exercise — and it was never assumed we would,” the company said.
That explanation came as the technology showed signs of strain. Meta’s internal AI software platforms and infrastructure saw code changes rise 220 percent year over year, but new or improved features reaching users rose only 36 percent. The gap pointed to a large amount of development work without a matching increase in finished features.
More code, more problems
The same systems also brought more trouble for employees. Technical and security incidents rose 40 percent, while the time employees spent dealing with those problems grew by 70 percent. Those figures made the AI transition harder to present as a simple way to reduce work or replace staff.
Zuckerberg acknowledged the slowdown in July 2026. “The trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected,” he said. His comment offered a clear reason the workforce plan could not move ahead at the pace executives had considered.
The problems were not limited to software performance. Meta installed tracking software on US employees’ computers to train AI, which led to complaints from employees and a 19-point drop in employee-sentiment scores. The move connected the company’s AI push to concerns about workplace monitoring, adding pressure from inside the workforce.
Meta keeps spending as the strategy changes
Abandoning Project OT does not mean Meta has walked away from AI. The company raised its spending forecast for 2026 as high as $145 billion, showing that AI remains a major priority even after the workforce plan failed to deliver what executives expected.
The spending increase also highlights the tension inside Meta’s strategy. The company is committing more money to AI platforms, infrastructure, and development while its AI efforts have not yet produced significant tangible gains. More code changes have not translated into the same level of new features, and the systems have created more technical and security problems for employees to manage.
For Meta staff, the shift carries two messages at once. The immediate threat of another layoff wave has been removed, but the company continues to invest in tools designed to change the work people do. Project OT is gone, yet the push to build AI agents remains part of Meta’s plans.
Meta’s experience shows the gap between planning an AI workforce and making one work in practice. Replacing thousands of staffers required more than ambitious targets and rising investment. The agents had to progress fast enough, produce useful features, avoid creating new security problems, and win the confidence of employees. Meta’s own results failed to meet those demands.
Based on
