IA · 27 August 2026 · 5 min read
Meta Pulls the Plug on 'AI-Native' Restructuring: Project OT Scrapped After Autonomous Agent Disruptions
In brief: An investigative report reveals details of Project OT, an internal Meta initiative aimed at replacing up to 60% of staff across select teams with autonomous AI agents. However, pilot programs uncovered severe operational hurdles, with autonomous agents performing disruptive actions across internal systems, prompting CEO Mark Zuckerberg's management to cancel subsequent layoffs and reallocate personnel.
by Team Mocchi's
The ambitious vision of transforming a tech conglomerate into a fully automated organization has hit a significant structural roadblock in Menlo Park. Earlier this year, Meta quietly initiated an internal program dubbed Project OT (Organization Transformation), aimed at reshaping the company's operating model around generative AI and slashing headcounts in select teams by up to 60%.
The initiative, championed by CEO Mark Zuckerberg to accelerate Meta's transition into an "AI-native" company, envisioned replacing thousands of knowledge workers with autonomous agents supervised by minimal human staff. However, as reported by Ars Technica, the initiative was abruptly abandoned after internal pilot programs triggered severe operational disruptions and unpredictable behavior from the deployed agents.
The Promise of the 'AI-Native' Enterprise
Internal documents reviewed during investigations described an "AI-native" company as one where AI tools and autonomous agents interact autonomously, workflows are self-orchestrated, and new builds are AI-first. Under these projections, the substantial cost savings generated from cutting team headcounts—estimated to represent an overall company reduction of 25% or more—were meant to fund top compensation packages for specialized AI engineers.
Following an initial restructuring round in May, Meta launched pilot programs restructuring engineering, research, and at least eight other divisions into smaller clusters. The goal was to test whether human teams could be largely replaced by autonomous bots operating under high-level human oversight.
Scale-Breaking Anomalies and Agent Drift
Integrating autonomous agents into complex daily operations revealed severe architectural shortcomings. According to internal accounts, when autonomous systems were tasked with interdependent assignments, they executed "large-scale, disruptive actions," altering critical configurations, clogging approval pipelines, and failing to handle edge cases that required contextual human judgment.
Rather than accelerating operations, managing the autonomous bots placed an excessive burden on the remaining staff, who spent substantial time auditing, correcting, and rolling back unintended system changes. Faced with declining workflow reliability and escalating infrastructure risks, executive management decided to scrap the second wave of planned layoffs.
Meta's Pivot and Organizational Reality
In an official statement, Meta downplayed the initiative as a scenario planning exercise carried out during company restructuring, noting that not every explored model was intended for execution. The company confirmed that thousands of employees were redeployed to priority projects rather than eliminated.
The cancellation of Project OT serves as an important reality check for the enterprise tech ecosystem. While AI agents excel at deterministic, bounded tasks, attempting to delegate end-to-end operational workflows without deep contextual reasoning removes the institutional adaptability required to keep complex systems stable.
Mocchi's take
Meta's experience with Project OT provides a crucial takeaway for companies evaluating AI adoption in their core business operations. Attempting to replace entire functional teams with autonomous agents misjudges both the messy reality of enterprise workflows and the irreplaceable value of human oversight. The viable path for modern businesses is not total autonomous displacement, but cognitive augmentation through tightly governed, human-in-the-loop systems. Designing software architectures that keep human intent and accountability at the center remains the single most reliable strategy for achieving sustainable efficiency.