Meta scrapped AI-native restructuring plan after agents caused major disruptions

Meta’s internal restructuring plan, codenamed Project OT, aimed to reduce some teams by up to 60% to transition the company toward being 'AI native,' according to Reuters, which reviewed internal documents and spoke with more than 20 current and former employees. The plan, initiated by CEO Mark Zuckerberg in January, explored scenarios where AI would handle much of the daily work performed by thousands of employees, with small teams overseeing the agents. Meta confirmed to Reuters that the plan involved scenario-planning exercises but did not confirm which teams were affected. The first round of layoffs under Project OT occurred in May, though Meta canceled a second planned round.

Meta’s definition of 'AI native' included using AI-ready tools and agents, automating workflows, and prioritizing AI-first development, as outlined in an internal document. The company also aimed to sell AI agents to third parties, a initiative it launched in June. Earlier this year, Meta restructured engineering, research, and at least eight other teams into smaller groups as part of a pilot program described in an internal 'AI-Native Playbook' post. The playbook outlined removing middle management and using 'agent-assisted analysis' to prioritize daily tasks, though Meta stated that performance ratings and promotions were and are made by people, not AI.

Internal posts cited by Reuters indicated that AI agents made 'large-scale, disruptive actions that humans are unlikely to execute,' contributing to a 40% increase in major technical and security incidents compared to the prior year. Employee time spent resolving these issues increased by as much as 70%. Meta declined to comment on these internal posts. In July, Zuckerberg reportedly told employees that the 'trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected.'

Meta also paused a program that tracked employees’ keyboard and mouse input to train AI agents after reports of the practice damaged morale. Reuters noted that HR employees and AI systems were involved in decisions about promotions, though Meta reiterated that final decisions remained human-led. The company’s shift away from Project OT followed internal reports of declining productivity gains from AI, with code changes up 220% year-over-year but feature upgrades reaching users only up 36%.

The episode underscores the challenges organizations face when integrating AI into core operations, even at companies with advanced AI capabilities. Meta’s experience suggests that while AI can drive efficiency in some areas, its deployment at scale may introduce unforeseen operational risks and require careful balancing with human oversight.