Key facts
- Meta explored reducing some teams by up to 60% to become 'AI native'.
- The plan, codenamed Project OT, aimed to use AI agents for daily tasks.
- A first round of layoffs occurred in May, but a second round was canceled.
- Internal documents described 'AI native' as a company where AI-first builds and AI agents are central.
- Meta CTO Andrew Bosworth noted a large increase in code changes but a smaller increase in new features.
- Internal reports indicated AI agents led to a 40% rise in technical and security incidents.
Meta explored a significant restructuring plan, codenamed Project OT, earlier this year with the goal of becoming 'AI native.' This initiative involved scenario planning that could have led to slashing some teams by as much as 60 percent, according to a Reuters report citing internal sources. The plan aimed to leverage AI agents to perform many daily tasks currently handled by human employees, potentially reassigning some staff and laying off others.
Meta confirmed that the scenario planning exercise explored potential impacts of redeployments, open role closures, and cuts. However, the company stated that not all scenarios were pursued and that the plan was never assumed to be fully implemented. The first round of layoffs related to Project OT occurred in May, but Meta subsequently canceled a second planned wave of job cuts.
Internal documents reviewed by Reuters described an 'AI native' company as one where AI-ready tools and agents interact, workflows are automated, and new builds are AI-first. This also included the potential to sell AI agents to third parties. Meta had reportedly restructured some engineering and research teams into smaller groups as part of a pilot program.
Despite initial plans, Meta reportedly scaled back Project OT due to factors including employee morale concerns, uncertainty about AI's productivity boost, and Zuckerberg's acknowledgment that AI agent development had not accelerated as expected. Internal data indicated a substantial increase in code changes and technical incidents following AI agent implementation, with less corresponding improvement in user-facing features.