Mark Zuckerberg acknowledged in an internal town hall on Thursday that Meta's AI agent development has not progressed as quickly as expected, and that the sweeping restructuring that included 8,000 layoffs 'hasn't come to fruition yet,' according to a recording heard by Reuters.

'The trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected,' Zuckerberg told employees. He added that executives were 'super optimistic' about tools like Anthropic's Claude Code when planning the restructuring in January and February, but the expected productivity gains have not materialized.

The rare admission from the CEO comes as Meta is projected to spend as much as $145 billion on AI infrastructure this year — a significant portion of Big Tech's more than $700 billion total AI outlay. The company laid off roughly 10% of its global workforce and reassigned about 7,000 employees to AI-focused teams in May, moves that prompted employee pushback and raised concerns about morale.

Zuckerberg said he expects more significant benefits from Meta's AI investments within the next three to six months. The timeline extension is notable: in earlier statements, Zuckerberg had predicted AI agents would transform Meta's operations much sooner.

In the same town hall, Meta's CTO Andrew Bosworth said a review of a recent data security incident involving the company's controversial mouse-tracking software — which tracked employee movements and digital activity for AI training — indicated that no employee data was included in AI training. If the program is turned back on, Bosworth said, it will be on an opt-in basis, reversing the previous policy that offered no opt-out.

The admission is the most concrete signal yet that the industry-wide race to deploy autonomous AI agents is hitting real-world friction. It mirrors broader challenges: while frontier AI models continue to improve rapidly at benchmarks, translating those capabilities into reliable, production-grade autonomous agents that can operate without human supervision remains an unsolved engineering problem.

Sources