Zuckerberg Says Meta’s AI Agent Progress Is Slower Than Expected
The Meta CEO told staff that agentic development hasn’t accelerated the way executives predicted when they restructured the company’s AI division in January.
July 3, 2026 – 7:46 am
Mark Zuckerberg shared with Meta employees on Thursday that the company’s AI agents have not progressed as swiftly as he had hoped, four months after a restructuring aimed at speeding up development. He stated:
"The kind of trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected."
This admission comes amidst significant investment and structural changes at Meta. The company is projected to spend up to $145 billion on AI infrastructure this year, including cutting roughly 8,000 jobs in May while simultaneously shifting thousands of staff onto AI-focused teams.
Zuckerberg explained that executives were "super optimistic" about coding tools like Anthropic’s Claude Code during the planning phase in January and February, expecting this optimism to translate into faster progress with Meta’s own products. However, he acknowledged that these bets haven’t materialized yet and the reorganization wasn’t as clean as intended, with timing influenced by concerns over Meta not moving fast enough to adapt.
Despite the challenges, Zuckerberg expressed optimism about future gains from AI investments, stating that Meta expects "more significant benefits" within the next three to six months. However, he did not elaborate on specific products or teams expected to deliver these improvements.
During the same town hall meeting, Meta’s CTO Andrew Bosworth addressed a separate issue related to an internal review of data privacy practices. He informed employees that while no employee data from a paused mouse tracking and keystroke monitoring tool was used to train AI models, the program may resume on an opt-in basis rather than its original opt-out design.