Mark Zuckerberg Wanted AI to Replace Meta Workers - Plan Collapsed Within Months
Meta launched "Project OT" in early 2026 to transform into an "AI-native" company, envisioning AI agents replacing traditional roles and potentially shrinking teams by up to 60% The initiative stalled after internal data revealed a massive productivity gap: code changes surged 220% year-over-year, but user-facing feature improvements rose only 36% AI adoption caused a 40% increase in technical and security incidents and a 70% jump in time spent resolving them, alongside deteriorating employee mo
Analysis
TL;DR
- Meta launched "Project OT" in early 2026 to transform into an "AI-native" company, envisioning AI agents replacing traditional roles and potentially shrinking teams by up to 60%
- The initiative stalled after internal data revealed a massive productivity gap: code changes surged 220% year-over-year, but user-facing feature improvements rose only 36%
- AI adoption caused a 40% increase in technical and security incidents and a 70% jump in time spent resolving them, alongside deteriorating employee morale (sentiment dropped from 74% to 55%)
- Zuckerberg abandoned the second wave of company-wide layoffs and pivoted public messaging toward AI as a worker empowerment tool rather than a replacement mechanism
- The core assumption that AI agent development would accelerate rapidly proved premature, forcing Meta to pause initiatives like employee keystroke-tracking for AI training
Why It Matters
This case study illustrates the growing gap between AI's code-generation capabilities and its ability to deliver real product value—a critical concern for any organization betting heavily on AI-driven transformation. It also highlights the human and cultural risks of aggressive AI adoption, showing how morale collapse and security degradation can undermine even well-funded strategic initiatives.
Technical Details
- Project OT envisioned replacing traditional product-development teams of 10–20 specialists with "pods" of 3–5 "builders," eliminating middle management layers and conventional job titles
- Meta deployed tracking software on U.S. employees' computers to record keystrokes and mouse movements for AI training purposes, later pausing the initiative after internal backlash
- Internal metrics showed a 220% YoY increase in code changes to Meta's platforms and infrastructure, but only a 36% increase in changes resulting in new or upgraded user-facing features
- Major technical and security incidents, including outages and possible data leaks, increased 40% YoY, while employee time spent responding to those incidents jumped 70%
- The company identified "Irreplaceable Talent" including hypothetical "10X Performers" who could produce the output of much larger teams with AI assistance
Industry Insight
- Organizations pursuing AI-native transformations should anticipate a significant "productivity illusion"—where output metrics (code volume, task speed) improve dramatically without proportional gains in end-user value, requiring new evaluation frameworks beyond raw throughput
- Employee trust is a fragile asset in AI-driven restructuring; surveillance-adjacent practices like keystroke tracking can trigger rapid morale collapse and internal resistance that undermines the very AI training data companies seek to collect
- The pacing mismatch between AI capability development and organizational readiness is a recurring risk; leaders should consider phased, team-specific pilots rather than company-wide overhauls to avoid the kind of strategic reversal Meta experienced
Disclaimer: The above content is generated by AI and is for reference only.