Meta chief executive Mark Zuckerberg told employees the company’s AI agent technology has progressed more slowly than expected, acknowledging at an internal town hall on Thursday that its sweeping restructuring has yet to deliver the results executives anticipated.

Reuters reported that Zuckerberg said the company had misjudged both the pace of AI development and the execution of a reorganisation that included cutting around 10 per cent of Meta's global workforce and reassigning roughly 7,000 employees to AI-focused teams in May. The changes were designed to fund major investments in AI infrastructure and improve efficiency, but prompted employee concerns over morale.

Reflecting on the strategy, Zuckerberg said: “The trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected,” adding that the company's bets on the new structure “haven't come to fruition yet.” He said executives had been “super optimistic” earlier this year about the potential of AI coding tools such as Anthropic's Claude Code, believing the company needed to move faster to adapt.

Meta is projected to spend as much as $145 billion on AI infrastructure this year, forming part of more than $700 billion in planned AI investment across major technology companies. Despite the slower-than-expected progress, Zuckerberg said he expects Meta to begin seeing more meaningful returns from those investments within the next three to six months. A Meta spokesperson declined to comment on the internal meeting.

Reuters reported that the town hall addressed a separate review of Meta's employee monitoring software following a recent data security incident. Chief technology officer Andrew Bosworth said the investigation found no employee data had been used to train AI systems after sensitive information was inadvertently exposed, prompting the company to suspend the programme last month.

Bosworth said Meta may restore the software on an opt-in basis once the review is complete, reversing its earlier approach when the programme was installed on US employees' computers in April without an opt-out option. “For people who are comfortable, that's great, they can contribute to this kind of great human survey. To people who are not, it is not an issue,” he told employees.

---