Zuckerberg: AI Agent Progress Slower Than Meta Expected

Meta's AI agent ambitions hit reality check

Mark Zuckerberg acknowledged to Meta employees Thursday that the company's AI agent development has not progressed as quickly as anticipated, raising questions about the timing and scope of a restructuring that eliminated thousands of jobs earlier this year.

Speaking at an internal town hall, the Meta CEO said that "the trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," according to a recording obtained by Reuters. He added that the company's strategic bets tied to its new organizational structure "haven't come to fruition yet."

Why it matters

Meta is projected to spend up to $145 billion on AI infrastructure in 2024 alone — part of more than $700 billion Big Tech companies are collectively investing in artificial intelligence. Zuckerberg's admission that AI agent capabilities are lagging internal expectations suggests the return on that massive capital outlay may take longer to materialize than investors and employees were led to believe. The comments also provide rare insight into how optimism about tools like Anthropic's Claude Code influenced major corporate decisions that affected thousands of workers.

Restructuring driven by AI optimism

Zuckerberg revealed that when Meta began planning its reorganization in January and February, executives were "super optimistic" about emerging AI tools and worried the company wouldn't "move fast enough to adapt." That optimism appears to have influenced both the timing and severity of changes that proved messier than intended.

The CEO conceded the reorganization that included significant job cuts was not as "clean" as it could have been, and that leadership had miscalculated on timing. Despite the slower-than-expected progress, Zuckerberg told employees he expects Meta will begin experiencing more substantial benefits from its AI investments within three to six months.

Mouse-tracking program under review

Separately, Meta CTO Andrew Bosworth addressed a recent security incident involving the company's controversial employee monitoring software. The program, which tracks mouse movements and digital activity for AI training purposes, was paused last month after sensitive data was potentially exposed.

Bosworth said a review found no employee data was included in AI training. If Meta reactivates the program after completing its investigation, participation will be opt-in — a reversal from April when the software was installed on U.S. employee computers without an opt-out option.

A Meta spokesperson declined to comment on the town hall remarks.

These details were first reported by Reuters journalists Katie Paul and Courtney Rozen.