Earlier this year, Meta executives hatched a bold plan to transform the company into an “AI native” organization — by replacing as much as 60 percent of the staff on some teams with AI agents. The initiative, reportedly codenamed Project OT (short for organization transformation), envisioned two rounds of layoffs that would dramatically reshape the social media giant. But according to internal sources speaking to Reuters, the scheme imploded within months, as the same AI agents meant to replace workers instead caused “large-scale, disruptive actions.”

A Plan That Backfired

Reuters, citing two people familiar with Meta’s internal affairs, revealed that Project OT explored scenarios in which teams would be trimmed by up to 60 percent. The plan included two rounds of layoffs, though Meta declined to specify which teams were targeted. The news, first reported by Ars Technica, paints a picture of an organization struggling to calibrate the boundaries between algorithmic automation and human judgment.

The failure is a cautionary tale for the broader tech industry, which has rushed to integrate AI agents into workplaces. Zuckerberg himself reportedly admitted that Meta’s AI development “hasn’t accelerated” despite billions invested and over 8,000 job cuts. His concession, reported by Inkl, undercuts the narrative that AI can seamlessly replace experienced professionals.

“The reality is that these AI agents are not yet mature enough to handle the complexity and context of corporate work,” said a veteran AI researcher familiar with the project, who asked to remain anonymous. “They can automate tasks, but they can’t replicate judgment, accountability, or the tacit knowledge that comes from years on the job.”

Lawsuit and Internal Fallout

Adding to Meta’s troubles, a lawsuit filed in California accuses the company of using AI to target workers for layoffs who took medical and maternity leave. The Mercury News obtained the complaint, which alleges that Meta’s algorithms were trained on productivity data that unfairly penalized employees with legitimate absences. The case highlights the legal and ethical risks of algorithmic human-resources decisions.

Meanwhile, reports emerged that Meta is now “drafting 7,000 workers into its AI rebuild,” according to Startup Fortune. But this shift, from culling workers to retraining them, suggests a scramble to salvage a faltering strategy.

A Wider Industry Reckoning

Meta is not alone. Companies across sectors have embraced AI-driven workforce reductions, only to confront unintended consequences. Oracle admitted that artificial intelligence has cost 21,000 jobs, according to Forbes, while Walmart recently cut 1,000 corporate roles as part of an AI reorganization. Microsoft has deployed AI agents in sales and finance roles, and PwC has faced a backlash from former employees who warned that “AI factory” models are unsustainable, as ABC News reported.

Yet a surprising trend is emerging: firms are quietly rehiring the very workers they fired for AI. Emerald Book reported that companies are rehiring because the AI systems they deployed are underperforming, require costly supervision, and lack the flexibility of human employees. “Tokens cost more than employees,” one executive told Asia Financial. Indeed, Forbes ran a piece bluntly titled “AI Costs More Than The People It Replaced.”

The Economics of AI Replacement

Why does AI replacement so often fail? A 2026 analysis from Forbes notes that the total cost of ownership for an AI agent — including engineering, integration, prompt tuning, infrastructure, and oversight — frequently exceeds the fully loaded cost of a salaried employee. A single AI agent may perform a specific task 24/7, but it requires human monitoring to avoid costly errors. One miscalculation or security flaw can erase months of savings.

Bill Gates, in his “Year Ahead 2026” essay, offered a more measured outlook, suggesting that AI will augment rather than replace human work in the near term. He emphasized that “AI is a tool, not a destiny,” and advocated for policies that support workers through transitions. Though his piece was inaccessible in full, his pre-publication commentary aligns with the growing consensus that automation is a journey, not a switch.

The Human Factor

Mark Cuban has been blunt about the risks of replacing employees with AI agents. “Remember, employees know that they will be fired,” he warned on MSNBC. “If you think they’re going to train their replacements, you’re delusional. You’ll lose your best people before you even cut them, and the morale of those who stay will crater.” Cuban’s point is that AI replacement creates a perverse incentive: the very knowledge needed to make AI work is retained by the people being let go.

Empirical evidence supports this caution. A study from MIT Sloan found that workers with less experience gain the most from generative AI, as it helps them bridge knowledge gaps. But that doesn’t mean novices can replace veterans. An Upwork analysis showed that AI-human collaboration boosts task completion by 70%, but only when humans oversee the process with domain expertise. In other words, AI is a force multiplier, not a substitute.

“The big AI job swap is already happening,” wrote The Guardian in a recent feature. “But it’s not simply robots taking jobs; it’s a reskilling revolution where white-collar workers are ditching careers they’ve spent years building.”

What This Means for Workers and Leaders

The takeaway from Meta’s failed experiment is clear: AI deployment requires a human-centric strategy. Layoffs based on AI projections risk legal liability, as Meta’s lawsuit demonstrates. They also create institutional amnesia — losing the very people who hold critical knowledge about systems, customers, and culture.

For workers, the advice from experts like those at NBC News is to embrace AI as a skill, not fear it as a sentence. The roles most affected are those with repetitive components, but new roles are emerging for AI trainers, prompt engineers, and AI auditors. The World Economic Forum has consistently projected that AI will create more jobs than it destroys, but only if companies invest in retraining.

A Path Forward

Successful organizations will treat AI as a complement to human workers, not a wholesale replacement. That means involving employees in the design of AI systems, providing transparency, and building in safeguards against bias. The case of Meta’s Project OT is a stark reminder that the cheapest solution is rarely the best one. As one former Meta senior engineer told Reuters: “We spent months trying to make an AI agent that could do a fraction of what our best analysts do. We ended up with a system that needed constant human corrections. It was cheaper to keep the humans.”

The future of work is not a binary choice between humans and machines. It’s a partnership. And that partnership works best when it’s built on trust, not cost-cutting.