Meta’s AI Staffing Experiment Is Backfiring, and That Changes Everything

The Math That Didn’t Add Up

Meta launched a bold organizational experiment in early 2026. The plan, dubbed Project OT, was straightforward in theory: use AI to take over routine work, shrink teams by as much as 60%, and let a skeleton crew of “AI-native” employees manage virtual co-workers instead. The savings from layoffs would then fund aggressive hiring of top-tier AI engineers—creating a self-reinforcing cycle of investment and efficiency.

It didn’t work. Not because the idea was bad, but because the technology isn’t ready for the job it was asked to do.

Meta carried out its first round of cuts in May, shedding roughly 10 percent of its workforce. The second wave, scheduled for November, was cancelled. Internal surveys showed employee morale plummeting from 74 percent positive sentiment to 55 percent—a steep drop that coincided directly with the restructuring news.

The numbers tell a deeper story than headline layoffs. According to internal documents cited by Reuters, AI-assisted coding and agent deployment drove a 40 percent year-over-year increase in service disruptions and data leaks. The time employees spent cleaning up those mistakes rose 70 percent.

In other words, Meta’s own AI tools made its engineers busier, not freer.

Who Got Burned

The most immediate casualty is the productivity thesis itself—the idea that AI can systematically replace human labor at scale. Meta’s experience suggests otherwise. AI agents are not yet reliable enough to handle production work without generating errors that require more human attention to fix than the original work would have taken.

Employees are also footing a visibility tax. Meta’s plan to track workers’ mouse movements and computer activity to improve AI training sparked an online petition among U.S. staff. In the U.K., workers began exploring unionization specifically over AI-driven employment insecurity and surveillance concerns. This is not a hypothetical future—it’s happening now, and it’s shaping how tech workers organize.

The irony is sharp: Meta wanted to reduce headcount by replacing people with AI. Instead, it created conditions that demanded more human coordination, more oversight, and more defensive organizing.

The Money Illusion

Here’s what makes Meta’s situation especially significant for the broader industry. The company is still committing $130 billion to AI infrastructure this year. That’s an enormous sum—roughly equivalent to the GDP of a mid-sized European country—directed at a problem that hasn’t yet demonstrated the ROI being bet on.

Zuckerberg acknowledged publicly that AI agent technology is advancing slower than expected. His timeline for improvement? Three to six months. That’s a credible leader admitting a gap between vision and delivery, but it’s also a warning signal. If Meta’s own AI rollout is stalling despite unlimited capital, what happens when the rest of the market’s less-walled gardens try the same move?

The $130 billion spend also raises a second-order question: where is that money going if not toward replacing workers? A portion is flowing into recruiting star AI engineers—the very people Meta hoped layoffs would help fund. So the company is simultaneously cutting some roles and competing fiercely for others, creating a strange talent imbalance rather than the streamlined workforce it envisioned.

What Happens Next

A few things are likely to follow from this episode.

First, expect other tech companies to pull back on aggressive AI-driven layoff promises. Meta’s November cancellation signals that even the most committed executive leadership will rein in restructuring when the data doesn’t support it. The narrative that “AI will replace workers and we’ll hire fewer people” is already showing cracks.

Second, AI tooling will shift from replacement to augmentation—at least in public messaging. Companies will continue investing heavily in AI, but the framing will change from “replacing humans” to “augmenting humans.” That’s not just PR; it’s a reflection of the fact that pure replacement hasn’t worked yet, and the economics of error remediation are brutal.

Third, worker organizing around AI is going to intensify. The U.K. union efforts and U.S. petitions are early signals. As AI deployment expands into more functions—coding, design, customer support, data analysis—employees will increasingly treat AI adoption as a collective bargaining issue, not just a management decision.

Fourth, the next round of Meta layoffs remains possible. Zuckerberg ruled out another company-wide cut this year, but the structural pressure hasn’t disappeared. If AI performance improves on his three-to-six-month timeline, the cost math could shift enough to justify further restructuring—or it could expose another gap and deepen the credibility problem.

The Bigger Point

Meta’s experiment reveals a fundamental tension in how tech companies are approaching AI. The assumption has been that better tools automatically mean fewer people. But tools that break, leak data, and require constant human correction don’t reduce workload—they redistribute it in ways that often increase total effort.

This isn’t a failure of ambition. It’s a failure of timing. The technology will keep improving. But the current generation of AI agents isn’t ready to carry the organizational weight investors and executives are placing on them.

The lesson for the industry is sobering: you can spend billions, restructure aggressively, and still end up with more problems than you started with—if the underlying tooling doesn’t deliver. Meta learned that the hard way. Everyone else should be watching closely.