Zuckerberg Tells Meta Staff ‘We’ve Made Mistakes’ in AI Overhaul

Mark Zuckerberg has told Meta employees that the company’s push to rebuild its workforce around artificial intelligence has produced mistakes, in an internal memo seen by Reuters and dated to mid-June. The admission lands five weeks after Meta laid off roughly 10% of its global workforce and reassigned 7,000 more employees into AI-focused roles, a sweeping restructuring that is now touching about one in five of the company’s nearly 78,000 workers.

“Given the complexity of these changes, we’ve made mistakes and will almost certainly make more,” Zuckerberg wrote, according to the memo reported by Reuters on June 12. He said he is “focused on providing as much stability as possible” in the months ahead and reiterated that Meta does not expect more company-wide layoffs this year. Meta declined to comment on the memo when Reuters reached out.

The Memo and What It Admits

The internal memo is the most direct on-the-record concession from Zuckerberg that the AI workforce transformation, which the CEO has spent most of 2026 pitching as the spine of Meta’s future, is creating friction inside the company as well as opportunity. The same document that defends the strategic case for the restructuring also acknowledges the human cost of running it at speed.

Zuckerberg framed the cuts as a calculated trade. “By creating important new roles for people, this also allowed us to shrink the size of teams knowing that if we make mistakes in some places, then we could transfer some people back,” he wrote. He added: “I don’t want to overpromise because the world is changing in ways that are out of our control.” The memo was first reported on by Reuters and obtained independently by the internal memo in which Zuckerberg conceded mistakes.

Given the complexity of these changes, we’ve made mistakes and will almost certainly make more.

The May 20 Restructuring in Numbers

Meta kicked off the restructuring on May 20, when the company began notifying employees of the cuts. The layoffs affected roughly 8,000 people, about 10% of global workforce, Meta spokesperson Erica Sackin confirmed to NPR at the time. The company also reassigned another 7,000 employees into new AI-focused initiatives and closed roughly 6,000 open job postings, according to a memo from Chief People Officer Janelle Gale obtained by Reuters.

Taken together, the layoffs, transfers, and previous role eliminations are expected to ultimately affect about 20% of Meta’s workforce. Gale, who authored the May memo, told employees the company was applying “AI native design principles” to the new structure. “As org leaders worked on the changes, many of them incorporated AI native design principles into their new org structures,” she wrote. “We’re now at the stage where many orgs can operate with a flatter structure with smaller teams of pods/cohorts that can move faster and with more ownership.” Workers inside Meta have reportedly taken to calling the reassignment process being “drafted.”

The pattern is not unique to Meta. Across the largest U.S. tech companies, AI investment has run alongside deep workforce cuts, a dynamic captured in Big Tech’s $700 billion AI infrastructure spend and 142,000 jobs cut. Meta’s particular version is the size of the reassignment pool: 7,000 people, roughly 9% of the company, are now doing work they did not sign up for.

• ~8,000 employees laid off in May 2026 (10% of workforce)
• 7,000 employees reassigned to new AI initiatives
• ~20% of workforce ultimately affected by layoffs, transfers, and prior role eliminations
• ~6,000 open job postings closed during the restructuring
• ~78,000 total employees as of end of March 2026, per company filings

Where the 7,000 Reassigned Employees Went

The 7,000 reassigned workers are landing in teams that, in many cases, did not exist 12 months ago. According to the May 18 internal memo from Janelle Gale obtained by Reuters, employees are being shifted to four new groups building AI tools and apps. Two of those teams have been named publicly: Applied AI Engineering, the unit tasked with building AI systems for the company’s own operations, and Agent Transformation Accelerator, a group focused on AI agents capable of autonomously performing workplace functions.

Many of the new roles involve training AI models, a function that has become a destination for displaced engineers, content moderators, and operations staff. Meta has framed the transfers as a way to redeploy talent that the company values rather than shed it, a message Zuckerberg repeated in the latest memo. The reassignments are also the operational backbone of the superintelligence push that Alexandr Wang’s bet on health AI at Meta Superintelligence Labs depends on.

The four receiving teams, as described in the May memo:

• Applied AI Engineering (building AI systems for the company’s own operations)
• Agent Transformation Accelerator (AI agents for workplace functions)
• Two additional teams building AI tools and apps, named in the Reuters memo and not yet identified publicly

The Manager Sprawl Problem

The same restructuring that put 7,000 people into new AI roles also stretched the people who manage them. Meta’s new Applied AI Engineering unit, the most prominent of the receiving teams, has a flat structure with up to a 50:1 ratio of individual contributors to managers, a figure first reported by the Wall Street Journal and detailed by Fortune. The 50:1 number is double the 25:1 that organizational researchers generally treat as the outer limit of healthy span of control.

The result, on paper, is a flatter, faster-moving engineering org. In practice, executives and academics have warned it is a recipe for burnout and neglect, the very conditions that the 7,000 reassignments are meant to address. André Spicer, executive dean of Bayes Business School in London, told Fortune the structure was bound to fail. “It’s going to end in tragedy is the bottom line,” Spicer said. He pointed to research suggesting the right-sized team is seven people per manager, give or take a few, and noted that flat organizations tend to recreate informal hierarchies once formal ones are stripped out.

Zuckerberg’s latest memo signals Meta has heard the warning. The CEO said the company has “taken note of concerns over the widening of manager oversight responsibilities and plans to scale back the practice.” The concession is notable because Applied AI Engineering, the unit with the 50:1 structure, was announced in March 2026 as a flagship of the new org design. Meta did not respond to Fortune’s request for comment on how the unit will function, and the latest memo does not specify how aggressively the company will roll back the ratio. A look at Meta’s 50 engineers per boss structure shows the gamble is now in retreat.

The Spending That Made It Necessary

The May restructuring is downstream of a financial commitment Meta made just weeks earlier. In April 2026, the company raised its 2026 capital expenditure forecast to between $125 billion and $145 billion, up from a prior range of $115 billion to $135 billion, per the company’s earnings update. CFO Susan Li told analysts the increase “reflects our expectations for higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity.” Li added that the company needed to spend big “to meet our infrastructure needs and ensure we maximize our strategic flexibility over the coming years.”

The spending case is tied directly to the workforce case. The same quarter, Meta recorded a $107 billion jump in contractual commitments tied to “multi-year cloud deals and our infrastructure purchase agreements.” The company is also designing custom silicon with Broadcom and buying “a significant amount of AMD chips to complement the new Nvidia systems that we’re rolling out,” Zuckerberg said on the earnings call. The infrastructure needs to be staffed, and the staff is being reorganized around it.

Investors did not uniformly buy it. Meta’s stock fell 7% the day of the capex update, while Alphabet’s shares rose 7% after it raised its own 2026 capex guidance to between $180 billion and $190 billion, per CNBC. The reason, analysts say, is structural: Meta is the only one of the four hyperscalers (Meta, Alphabet, Microsoft, Amazon) without a cloud business to monetize the AI spend. A look at AI stock market winners and losers in 2026 shows the divergence is now a pattern, not an event.

The new capital is also buying new products. On April 8, 2026, Meta launched Muse Spark, the first proprietary foundation model from Meta Superintelligence Labs, a sharp break from the open-weight Llama family the company had championed for years. The launch is the strategic justification the workforce cuts were organized around. A look at investors’ reaction to Meta’s $145 billion AI capex shows the model will be judged against cloud-driven peers before it can prove the case for the spend.

The July Hackathon and the Manager Rollback

The corrective moves Zuckerberg announced in the memo are small and specific. Meta is organizing a large-scale hackathon in July to foster cross-team collaboration and development on its latest models, the CEO wrote, and the company plans higher budgets for offsites and corporate events. The company also plans to scale back the widening of manager oversight responsibilities, the practice that produced the 50:1 ratio in Applied AI Engineering. The hackathon will be the first public test of whether the flatter structure can ship the models the $145 billion capex is buying.

Zuckerberg’s “stability” pledge is a commitment, not a forecast. He did not promise no further individual team changes, and the “almost certainly more” admission leaves room for more. What he did promise is that, for the rest of 2026, the company is done cutting at the company-wide scale. Whether that holds will depend on whether the July hackathon, the manager rollback, and the Muse Spark roadmap can deliver enough signal to justify the spending and absorb the cost of the restructuring the CEO has now publicly called a mistake.