Mark Zuckerberg admitted that Meta’s bet on AI Agents isn’t working, here’s why
Follow Us
In an internal town hall meeting yesterday, Zuckerberg admitted to Meta employees something the company had been trying hard not to say out loud this whole year. The AI agents Meta restructured its operations around, laid off 8,000 employees for, and tasked 7,000 more employees to develop, are not going to arrive on time after all.
SurveyAlso read: WhatsApp username row in India isn’t about privacy vs security, but authentication
Zuckerberg was reported by Reuters to have said in an internal meeting transcript that “the pace of development of these agents has not really accelerated the way that we expected” and that the company’s restructuring plan “hasn’t really paid off yet.”
Why is this revelation especially painful? Because Meta was not playing around at all in 2023. It fired about 10% of its workforce worldwide and allocated over 7,000 workers to AI projects, including an entire division devoted to what is known within the company as “Agent Transformation.” This is not some experiment. This is a major company that was basing its structure on the assumption that autonomous software agents – programs capable of carrying out operations like digital employees – would be ready for implementation.
They are not. And the way Zuckerberg himself justified why the company decided to bet on an agent-first strategy is even more telling than the fact itself. As Zuckerberg explained, in the months of January and February, when executives were planning a restructuring, they felt that Meta was moving too slow adapting to the industry’s future direction. Back then, he continued, leaders were “super optimistic” about products like Claude Code from Anthropic, and essentially just extrapolated their momentum to the roadmap of the company.
Also read: 5 best laptops with 64GB RAM in 2026: Alienware 16 Area-51, Lenovo Legion Pro 7 and more
And this is the smoking gun right here. Meta didn’t base its agentic strategy on any internal indicators. It saw what a competing lab was delivering, assumed the curve would continue bending in the same direction and reshaped an organization of tens of thousands people based on that assumption. As the curve flattened out, the organizational chart had nowhere to run.
Zuckerberg admitted that the layoffs themselves weren’t handled correctly as the process wasn’t as “clean” as it should have been. The timing of the process was misjudged as well. It’s the same executive who told his staff that he “will almost certainly” be making more mistakes in the future.
This doesn’t mean Meta abandons agents. The company is still expected to invest as much as $145 billion into AI infrastructure development this year, which is just a part of the Big Tech spending spree that already exceeded $700 billion. According to Zuckerberg, the company expects tangible results in 3 to 6 months from now. This is just the same optimistic period that led Meta here.
More important for the industry is the following. If the company that was the quickest to adapt to agentic AI technology openly admits that the technology hasn’t reached its peak yet, the roadmaps of all the other vendors should be subjected to the same analysis.
Also read: Why India is right to halt WhatsApp’s username feature
