Meta CEO Mark Zuckerberg admitted that the company’s artificial intelligence agents have not progressed as fast as expected, signaling a recalibration of ambitions amid sweeping internal changes. This rare acknowledgment came during an internal town hall where Zuckerberg reflected on Meta’s substantial investments and restructuring efforts focused on AI technology.
Earlier this year, Meta undertook a significant reorganization, cutting roughly 10% of its workforce and reallocating around 7,000 employees to AI-centered roles. The move aimed to support heavy infrastructure spending estimated at $145 billion and to drive efficiency gains through AI-assisted operations. However, according to Zuckerberg, the anticipated acceleration in agentic AI developments has yet to materialize, and the company’s reorganization results “haven’t come to fruition yet.”
Zuckerberg pointed out that executives were overly optimistic about the capabilities of AI tools, such as Anthropic’s Claude Code, when planning began at the start of the year. Despite the slower-than-expected progress over the last few months, he remains hopeful that Meta will see more tangible benefits within the next three to six months. This cautious outlook contrasts with growing momentum in the broader payments and commerce sectors, where companies like Visa, Mastercard, and American Express are already integrating agent-driven transactions into their systems.
In the same town hall, Meta’s Chief Technology Officer addressed concerns surrounding a recent data security review linked to the company’s mouse-tracking software used for AI training purposes. The review concluded that no employee data was incorporated into AI models. Meta had paused this program last month but is considering reintroducing it on a voluntary opt-in basis, reversing its prior stance that disallowed opting out.
The evolving landscape illustrates ongoing challenges in scaling agentic AI technologies despite the high investments and corporate restructuring geared toward AI integration. Industry projections, such as those from Goldman Sachs forecasting a 24-fold growth in AI-driven token consumption by 2030, underscore both the potential and the complex groundwork required—especially in the underlying payment infrastructures—to realize widespread AI commerce adoption.