Meta Abandons AI Agent Layoffs After Incidents Spike 40%

Meta recently halted an ambitious internal initiative codenamed Project OT, which aimed to replace up to 60 percent of employees on certain teams with AI agents. Initiated in January by CEO Mark Zuckerberg, the organization transformation plan sought to make Meta an AI native company. Under the proposed scenarios, which included a pilot restructuring engineering, research, and at least eight other teams, overall headcount would have been slashed by 25 percent or more. While a first round of layoffs occurred in May, Zuckerberg canceled a second wave scheduled for November.

The retreat followed significant technical friction and disappointing productivity metrics. Internal documents revealed that AI agents executed large-scale, disruptive actions that caused a 40 percent year-over-year spike in major technical and security incidents. Consequently, the time human employees spent resolving these system issues jumped by up to 70 percent. Furthermore, while code changes to internal software platforms rose 220 percent year-over-year, according to Meta CTO Andrew Bosworth, actual feature upgrades reaching users only increased by 36 percent.

These setbacks forced Meta to reconsider its AI-Native Playbook, which had outlined plans to eliminate middle management and use agent-assisted analysis for daily tasks. Zuckerberg acknowledged in July that the trajectory of agentic development had not accelerated as expected. For AI practitioners and enterprise leaders, Meta's retreat serves as a cautionary tale. It demonstrates that rushing to replace human workflows with autonomous agents can degrade system stability and increase operational overhead, proving that current agentic architectures are not yet reliable enough to operate without extensive human oversight.