NextFin News - Mark Zuckerberg's plan to remake Meta into an "AI-native" company, where virtual workers overseen by small cadres of humans would replace thousands of employees, collapsed within months after the technology failed to deliver the promised productivity gains and employees revolted. The CEO scrapped a second wave of layoffs on the eve of the first purge in May, leaving behind a demoralized workforce, a botched internal transfer program that a top executive later called "atrocious," and investors demanding to see a return on more than $130 billion in artificial-intelligence spending.
The unraveling of Project OT — short for Organization Transformation — is the clearest signal yet that the AI-driven workforce revolution Silicon Valley has been promising is arriving more slowly, and more painfully, than the hype suggested. For Meta, the episode raises a harder question: whether the company can keep funding an AI arms race at a pace few corporations in history have matched while the organizational machinery needed to profit from it is breaking down.
The Plan: An "AI-Native" Company Built on Scenario Planning
The blueprint was drawn up in January 2026, when Zuckerberg and his top lieutenants gathered for their annual leadership retreat at his Hawaii compound. There they hatched a radical plan to reimagine work at the owner of Facebook and Instagram for the age of artificial intelligence, according to an internal planning document and three people familiar with the project.
Code-named Project OT, the plan envisioned a future in which AI took over much of the daily work performed by thousands of human employees. Virtual workers would be overseen inside Meta by smaller, "talent-dense" cadres of human staffers. In scenario-planning exercises, executives explored slashing the size of many teams across the company by as much as 60 percent. Some employees would be offered roles in new units; others would be laid off as part of a culling that one human-resources executive projected would be as big as, or bigger than, the roughly 25 percent of the workforce Meta eliminated three years ago during its "Year of Efficiency."
The restructuring was to be carried out in two "waves," beginning with a first purge in May and followed by another shake-up in November, internal planning documents showed. Layoffs would be supplemented with the closing of open positions and the removal of employees Meta deemed poor performers.
But on the night of May 19, just hours before the first wave was set to begin, Zuckerberg blinked. Meta laid off 10 percent of its employees the next day — roughly 8,000 people out of a workforce of about 78,900 — but it called off planning for the November cuts, according to an internal document. The company confirmed the existence of Project OT, describing it as a year-long project focused on cost cutting, redesigning team structures and shifting staff into new priority areas, such as producing training data for its AI models. It acknowledged the plan was to be carried out in two waves and that the most drastic scenarios involved reducing some team sizes by up to 60 percent, but said it never intended to lay off 60 percent of its entire workforce and that several major units were not part of the exercise. Leaders cancelled the second wave before determining how many people overall would lose their jobs, the company said.
The reversal is striking for what it reveals about the gap between the AI transformation narrative sold to investors and the operational reality inside one of the world's most valuable companies. Meta asked teams to plan for cuts of up to 60 percent, moved roughly 7,000 employees into AI-focused units, closed about 6,000 open job postings, and then admitted the strategy could not be executed as designed. The company recorded $1.2 billion in severance expenses tied to the May headcount reduction in its second-quarter results.
Why the Machine Stalled
Three forces converged to sink Project OT, and each points to a different weakness in the AI-native thesis.
First, the workforce revolted. By May, Meta employees were in open revolt, convinced that the company's AI transformation initiatives were partly aimed at replacing them. The resentment ran deeper than the usual layoff anxiety: employees saw internal tools and agents being built to automate their own jobs, while being told the changes would make work "more rewarding." When a company's core product pitch — AI agents that do your work — becomes its internal management strategy, the conflict of interest is impossible to manage.
Second, the technology did not deliver. Internal data suggested that the autonomous AI "agent" technology at the heart of the strategy was failing to produce the hoped-for productivity gains. This is the crux of the matter. The entire restructuring math depended on AI agents absorbing enough routine work that far fewer humans were needed to produce the same output. If the agents cannot reliably execute multi-step tasks without human intervention, the "talent-dense cadres overseeing virtual workers" model collapses into its opposite: the same work, with an added layer of supervision and rework.
Third, investors began asking the question executives had hoped to defer: what does Meta have to show for its gargantuan AI spending? The company now expects 2026 capital expenditures of $130 billion to $145 billion, narrowed upward from a prior $125 billion to $145 billion range — more than double its 2025 outlay. That spending is funding AI chips, data centers, and the Superintelligence Labs organization under Chief AI Officer Alexandr Wang, including a $27 billion joint venture with Blue Owl Capital to build the Hyperion data-center campus in Louisiana — a project whose planned investment later grew to more than $50 billion — and a separate one-gigawatt data-center venture with BlackRock in El Paso, Texas.
The market's answer arrived on July 29, when Meta reported second-quarter results. Revenue rose 28 percent to $60.8 billion, beating expectations, but earnings per share of $6.18 missed the $7.14 consensus. Free cash flow — the number that separates an AI investment story from an AI spending story — plummeted to $784 million, down roughly 91 percent from a year earlier, as capital spending swelled to $31.08 billion in the quarter. The stock fell about 8 percent to 10 percent after hours.
The Applied AI "Gulag"
If Project OT was the strategic failure, the Applied AI division was the human one. Formed in March, the unit drafted roughly 6,500 engineers and product managers into work on Meta's generative AI models. Employees described the work as menial, and one worker characterized the assignment as "a gulag," according to a report on the unrest inside the unit.
In August, Meta's chief technology officer, Andrew Bosworth — long seen as a Zuckerberg loyalist — issued an internal memo acknowledging the damage. "We obviously did an atrocious job explaining the vision, giving people a clear picture of how we would support them and their careers in the shift, and painting a picture of how it would change over time," he wrote.
"We've undermined the trust you have that your specific expertise and contribution will be valued, that you will grow and advance your career, and that this will be a place where you can actually have an impact," Bosworth wrote. "We shook up the management structure that was providing you stability while rapid changes in strategy, including the boom/bust cycle of hiring, left entire teams in the lurch."
The memo promised concrete remedies: capping managers at about 20 direct reports, limiting how often employees switch managers during restructurings, refocusing managers on managing rather than independent contributor work, improving office microkitchens, and increasing travel budgets and spending on social events. "I hope we can rekindle the best of the culture we joined," Bosworth wrote.
Maher Saba, the vice president leading the Applied AI team, told employees who had been forced onto the unit that they would now be allowed to transfer to other roles within Meta if they could secure them. "Moving forward, we are returning to business as usual and giving people the agency to apply to roles that interest them," Saba wrote, recasting the company's once-infamous motto as "moving fast and fixing forward."
The sequence is a near-textbook case of how not to execute an AI transformation: announce a grand strategy, draft talent by force into a new unit, discover the work is neither fulfilling nor as productive as promised, watch morale collapse, then walk back the compulsion after the damage is done.
Cyclical Pain or Structural Shift?
The central analytical question is whether Meta's implosion reflects a cyclical stumble in an otherwise sound transformation or evidence that the "AI-native company" model itself is structurally flawed. The answer is both — and the distinction matters for investors.
The execution failures are cyclical and fixable. Communication broke down, managers were stretched thin, transfers were coerced, and the agent technology was immature. Each of these can improve: Bosworth's memo addresses the management-span and communication problems directly, and AI agent reliability is a moving target that will almost certainly improve over the next few product cycles. Meta's advertising business remains enormously profitable — revenue grew 28 percent in the second quarter — and the company retains the capital to keep funding the transition long after competitors have run out of runway.
But the deeper premise — that AI agents can soon absorb enough routine knowledge work to justify 60-percent team reductions — is a structural claim that the evidence does not yet support. The productivity gains required to validate Project OT's most aggressive scenarios depend on agents executing complex, multi-step workflows autonomously and correctly. Internal data suggesting those gains have not materialized is not a timing problem; it is a capability problem. Until agents can be trusted without extensive human oversight, the "talent-dense cadre supervising virtual workers" model adds cost rather than removing it.
This is why Zuckerberg's reversal is more significant than the reversal itself. Cancelling the November wave was an admission that the denominator of the equation — how much work AI can actually absorb — was unknowable at the scale and speed the plan required. A company can cut its way to a leaner cost structure; it cannot cut its way to a technology that does not yet exist.
The Counter-Thesis: Meta Is Simply Ahead of the Curve
The strongest case for Meta runs as follows: every technological transition produces a messy middle, and the companies that restructure first capture the productivity dividend before rivals do. From this perspective, the botched rollout is an execution problem, not a strategic one. Zuckerberg has been willing to absorb short-term morale and earnings pain to position Meta at the frontier of AI, just as he absorbed years of Reality Labs losses to build the metaverse bet. The advertising engine funds the transition, and when agent technology matures, Meta will have already built the organizational muscle to exploit it.
This argument has real force. Meta's scale in data, distribution, and compute is unmatched outside a handful of labs, and its willingness to spend $130 billion to $145 billion in a single year on AI infrastructure is a moat in itself. If agentic AI does deliver a step-change in software productivity, Meta's forced-march approach could look prescient in retrospect.
But the counter-thesis has one fatal weakness: it assumes the technology will arrive on the timeline the restructuring requires. If agent capabilities mature more slowly than capex commitments, Meta is left funding an arms race with money borrowed from its own operating efficiency — and the layoffs that were supposed to pay for the transition instead erode the institutional knowledge needed to build the products that would monetize it. The falsifying signal is specific: if Meta's agent-driven productivity metrics do not show sustained double-digit percentage gains in engineering output per employee over the next two quarters, and if AI-related revenue lines fail to appear as a discrete, growing segment in 2027 guidance, the "ahead of the curve" defense collapses into a capital-allocation error.
What Comes Next
In the short term, expect stabilization, not acceleration. Bosworth's memo is a pause button: cap manager spans, restore perks, let transferred employees move back. The second wave of layoffs has been cancelled, and the company has signaled it will not repeat the coercion that triggered the revolt. Morale will not recover on the strength of improved microkitchens alone, but the immediate turbulence should subside.
Over the medium term, the pressure shifts to the numbers. With free cash flow down to $784 million in the second quarter and capital expenditures running at $31.08 billion a quarter — an annualized pace above the top of the company's own $130 billion to $145 billion guidance — investors tolerated the earnings miss because AI spending was framed as an investment in future dominance. That tolerance has a limit, and it is measured in quarters, not years. The next two earnings reports will matter more than the next two product announcements.
In the long run, the verdict on Project OT depends on a question no internal memo can answer: can AI agents do enough real work to justify the organizational upheaval? If yes, Meta's painful restructuring will be studied as a case of painful but necessary transformation. If no, it will be remembered as the moment Meta cut deep into its own capabilities in pursuit of a productivity miracle that never arrived.
The base case is a slower, quieter AI integration: fewer headline-grabbing reorganizations, more incremental adoption of AI tools inside existing teams, and capex that stays elevated but becomes more disciplined as investors demand evidence of returns. The upside case is that agent technology inflects sooner than expected, validating the forced transfers and turning Applied AI into Meta's next growth engine. The downside case is a repeat of the pattern: another ambitious restructure, another morale collapse, and a growing cohort of investors who conclude that Meta's AI strategy is a spending program in search of a business model.
Meta's AI restructuring did not fail because the company spent too much. It failed because it tried to reorganize around a future that had not yet arrived — and discovered that you cannot lay off your way into a technology that still needs you.
Explore more exclusive insights at nextfin.ai.
