Zuckerberg admits Meta’s AI agents are lagging expectations

TL;DR:

• Mark Zuckerberg told a Meta town hall that AI agent development had not accelerated over the past four months in the way executives expected.
• He conceded a restructuring that cut roughly 10% of staff and moved 7,000 people onto AI teams was not as “clean” as it could have been and has yet to pay off.
• Meta is projected to spend up to $145bn on AI infrastructure this year.

Meta chief executive Mark Zuckerberg has acknowledged that the AI agents underpinning his sweeping reorganisation are progressing more slowly than he anticipated, according to a recording of an internal town hall heard by Reuters. It is a rare public dose of caution from a frontier lab leader, and a pointed counterweight to the “year of the agent” pitch UK enterprises are being sold.

Costly bets yet to pay off

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”, and that the company’s bets on the new structure “haven’t come to fruition yet”. The restructuring laid off about 10% of Meta’s workforce and reassigned some 7,000 employees to AI-focused teams in May, moves that dented morale. He now expects more significant benefits to arrive within three to six months.

The candour matters because it comes from a company spending on a scale few can match — up to $145bn on AI infrastructure this year, part of Big Tech’s $700bn-plus outlay. If agents are underdelivering even there, it reinforces a pattern British firms are already living: much-hyped autonomous systems are proving harder to operationalise than the marketing suggests. UK businesses are on course to waste an estimated £67bn a year on failed AI projects, and regulators are alert enough that the Bank of England has floated a “kill switch” for agentic AI in trading. Vendors, meanwhile, keep pushing agentic capability down in price — Anthropic’s Claude Sonnet 5 among them.

Looking forward

Zuckerberg’s three-to-six-month timeline will be watched closely; frontier bosses rarely set a clock on their own promises. For UK decision-makers, the takeaway is less “agents don’t work” than “agents are early”. Budgeting for pilots that may not compound this year — rather than betting an operating model on imminent autonomy — looks like the more defensible stance, whatever the roadmap slides claim.