Meta’s ambitious attempt to replace a massive portion of its workforce with artificial intelligence has hit a significant roadblock. Internal documents reveal that a strategic initiative aimed at massive staff reductions was abandoned following a combination of technological failures and intense employee pushback.

The Ambitions of "Project OT"

According to internal reports, Meta operated a highly aggressive workforce restructuring strategy under the codename "Project OT." The goal of this initiative was to shrink various departments by as much as 60 percent. Under this model, Meta intended to transition from large human teams to small, "talent-dense" groups of specialists tasked with supervising a fleet of virtual AI workers.

This shift was not merely a theoretical exercise; it was a calculated move to maximize efficiency by leveraging agentic AI to handle core operational tasks. However, the plan faced immediate friction when the promised productivity gains failed to materialize. In July, CEO Mark Zuckerberg himself admitted that the development of agent technology had not progressed as rapidly as the company’s aggressive timelines had anticipated.

Technological Failures and Investor Pressure

The collapse of Project OT was driven heavily by the gap between AI hype and technical reality. Meta’s AI agents, designed to act as autonomous digital workers, failed to deliver the expected leap in output. This technical shortfall created a vacuum of confidence, especially as investors began to scrutinize the company's massive capital expenditure on AI infrastructure.

When the technology failed to justify the staggering AI budget, the economic rationale for the mass layoffs began to crumble. The realization that AI agents were not yet capable of replacing human workflows at scale forced leadership to reconsider the entire roadmap for workforce automation.

Employee Revolt and the Trust Deficit

Perhaps the most significant factor in the reversal was a widespread internal uprising. The tension reached a breaking point when employees discovered that the company was using tracking software to log mouse clicks and keystrokes. Employees realized this data was being utilized to train the very AI models intended to replace them.

The fallout was immediate and measurable:

  • Internal sentiment plummeted from 74 percent to 55 percent.
  • Organized protests manifested through angry posts on internal networks and satirical memes posted on flyers throughout Meta's US offices.
  • Direct intervention by Mark Zuckerberg occurred on May 19, just hours before a planned second wave of layoffs scheduled for November was halted.

The Broader Impact on the AI Landscape

This development serves as a critical reality check for the tech industry. It highlights the "implementation gap"—the distance between the theoretical capabilities of LLM-based agents and their practical utility in complex corporate environments. For founders and developers, the Meta case proves that workforce transformation cannot be achieved through technology alone; it requires both technical maturity and a sustainable social contract with the human workforce.

Key Takeaways

  • Technical Limitations: Meta’s AI agents failed to meet the productivity benchmarks required to justify a 60% reduction in specific team sizes.
  • Workforce Backlash: Employee morale dropped significantly following the discovery of surveillance tools used to train replacement AI, leading to an internal revolt.
  • Strategic Reversal: CEO Mark Zuckerberg halted planned layoffs in November after realizing the agent technology had not scaled as expected.

Meta halted the AI-driven layoff scheme it called “Project OT” after the technology fell short and employees pushed back, marking a rare retreat from a plan that would have cut up to 60 percent of staff in several divisions. The decision came just hours before a second wave of cuts slated for November, and it underscores how far the promise of autonomous AI workers remains from everyday reality.

What “Project OT” was supposed to achieve

Des documents internes décrivent le Projet OT comme un effort de restructuration visant à remplacer de grandes équipes humaines par de petits groupes « à forte densité de talents » chargés de superviser des flottes de travailleurs virtuels dotés d'IA. L'objectif était une réduction allant jusqu'à 60 % dans certains départements, une mesure destinée à réduire les coûts tout en augmentant l'efficacité grâce à ce que Meta appelle l'« IA agentique » — un logiciel capable d'agir de manière autonome sur des tâches de routine.

Le plan n'était pas une idée vague ; il était intégré aux prévisions budgétaires et d'effectifs de Meta. Le PDG Mark Zuckerberg a publiquement reconnu en juillet que le développement des agents d'IA requis accusait un retard par rapport au calendrier agressif de l'entreprise, suggérant que les fondations techniques étaient déjà fragiles lorsque le déploiement a été annoncé.

Pourquoi l'IA n'a pas été à la hauteur

Les agents de Meta ont été conçus pour gérer le travail opérationnel de base sans supervision humaine. En pratique, ils n'ont pas réussi à produire les gains de productivité nécessaires pour justifier les coupes massives dans les effectifs. Les investisseurs, déjà méfiants face aux dépenses colossales de l'entreprise en IA, ont commencé à se demander si le budget « vertigineux » alloué à l'IA était justifié par les résultats. Le décalage entre le battage médiatique et la performance a érodé la confiance dans la pertinence économique des licenciements.

Réaction des employés et déficit de confiance

Le manque de performance technologique n'était que la moitié de l'histoire. Les employés ont découvert que Meta utilisait des logiciels de suivi pour enregistrer les clics de souris et les frappes au clavier, des données qui alimentaient directement les modèles d'IA destinés à les remplacer. Cette révélation a provoqué une chute rapide du moral interne — passant de 74 % de sentiments positifs à 55 % — et a déclenché une vague de protestations sur les plateformes de messagerie internes, des mèmes satiriques placardés sur des prospectus de bureau et des pétitions organisées exigeant l'arrêt du plan.

Zuckerberg est intervenu le 19 mai, juste avant que la vague de licenciements de novembre ne commence, mettant ainsi de fait le second tour en suspens.