Core Event and Key Facts
Meta initiated Project OT—an internal restructuring plan—to大幅replace human roles with AI agents this year but ultimately abandoned the second round of layoffs, implementing only partial changes.
According to Reuters’ investigation based on internal documents and over 20 sources, the key facts are:
- Launch timeline: January 2026, directly directed by CEO Mark Zuckerberg
- Original target: Reduce certain team headcounts by up to 60% to achieve “AI-native” status
- AI deployment model: Small oversight teams supervise AI agents handling most daily work of thousands of employees
- Execution status: First layoff round occurred in May; second round was cancelled
- Actual outcome: Thousands of employees were reassigned to newly formed priority teams; not all planned scenarios were adopted
- Corporate stance: Meta declined to disclose which teams were affected, confirming only that scenario-planning exercises took place
The key reversal: The much-discussed 60% cut figure was never fully implemented. Meta’s public statement explicitly stated they “never assumed we would” execute all scenarios—highlighting the gap between theoretical potential and practical execution.
Plan Details and Information Source
Project OT’s evolution reveals common AI-transformation challenges faced by tech companies: the gap between ambitious goals and operational reality.
Reuters reported:
- Internal documents showed AI agents were imagined to perform “much of the daily work performed by thousands of human employees”
- The plan featured two sequential layoff rounds overseen directly by Zuckerberg
- At least three sources and one internal document confirmed AI would be supervised by minimal human teams
- The report drew from dozens of internal documents, posts, recordings, and 20+ informed sources
Meta’s official response adopted strategic ambiguity:
“As part of our company restructuring, we asked teams to conduct scenario planning… Ultimately, we didn’t move forward with every scenario.”
This language acknowledged the exercise without committing to specific裁员 numbers—a tactic reflecting public-relations pressure during early-stage AI adoption.
Industry Context: AI Replacement Realities
An AI Agent is autonomous software that perceive environments, plan multi-step tasks, and execute actions without continuous human conditioning—distinct from traditional scheduled automation.
Meta’s experiment represents one of the earliest high-profile attempts to deploy AI agents for large-scale职能取代. Three structural barriers commonly hinder such efforts:
- Task fragmentation difficulty: Most roles contain unscripted, context-dependent duties resistant to complete automation
- Oversight overhead: Even simple agents require human supervision for compliance, error handling, and continuity
- Organizational inertia: Culture, workflows, and interpersonal networks cannot be instantly reconfigured via code
Meta’s pivot—reassigning rather than terminating staff—demonstrates how organizations balance AI ambition with human capital preservation.
Practical Recommendations for Readers
- Enterprise AI teams: Replicate Meta’s lesson by defining clear “AI-suitable vs. AI-resistant” task boundaries and building 12-18 month hybrid-operation buffers into deployment plans
- Tech professionals: Prioritize AI supervision roles (e.g., AI behavior auditing, task-chain validation)—demand for these oversight positions is rising alongside AI adoption
- Skeptical adopters: Start with “AI augmentation” over “AI replacement”—retain human decision authority while deploying AI for routine, repetitive execution tasks
Final Note
Meta’s mid-course correction underscores a frequently underestimated truth: the organizational integration cost of AI often exceeds its technical development cost. When technical capability and organizational maturity diverge, temporary retreat is not failure but a necessary calibration mechanism of mature organizations.