Zuckerberg admits Meta made mistakes in AI workforce overhaul

Mark Zuckerberg told employees Meta had made mistakes in its AI transformation of the workforce, a rare admission from a company pouring hundreds of billions of dollars into artificial intelligence. The memo landed after a sweeping May restructuring that cut 10% of Meta’s global workforce and moved 7,000 employees into new initiatives tied to AI workflows, underscoring how hard the company’s pivot has already been to execute.

Zuckerberg said Meta will almost certainly make more mistakes as it keeps reorganizing around AI, but he stressed that his focus was on giving employees as much stability as possible. He also said he did not want to overpromise, arguing that the world is changing in ways outside the company’s control. The message was aimed at reassuring workers that the shift is strategic, even as he acknowledged that parts of it have already gone wrong.

He said Meta does not expect more company-wide layoffs this year and will try to find new roles for employees reassigned to train AI models. Zuckerberg also said that by creating important new roles for people, the company could later transfer some employees back if it discovers mistakes in how teams were resized or reorganized. That language points to a company still adjusting the basic mechanics of how work gets done, from staffing levels to team structure to the division between human labor and AI systems.

The stakes are unusually high for Meta because its social platforms still generate enormous ad revenue, but Zuckerberg is betting that the next competitive edge will come from AI tools, infrastructure and assistant products. His memo reflects a broader pattern across major U.S. tech companies, which are cutting, reshaping or redeploying staff to chase the AI boom while trying to protect product development, recruiting and employee confidence. Meta’s experience is an early warning for corporate America: AI transformation is not only about building new products, but about deciding which jobs, workflows and management choices can survive the transition intact.