Mark Zuckerberg told Meta AI employees at an all-hands meeting on July 2, 2026, that AI agents are not developing at the pace the company's leadership had expected. He also acknowledged that the layoffs carried out earlier in the year were not as "clean" as they should have been, and that the intended benefits of Meta's AI-focused organizational structure have not yet emerged.
Key takeaways
- Meta cut approximately 8,000 employees (around 10% of its corporate workforce) and moved another 7,000 into AI groups
- One of those groups, Agent Transformation, was created to replace human tasks with AI agents
- Zuckerberg said the layoffs were not executed cleanly and acknowledged the restructuring was driven by fear of falling behind the industry
- The expected gains from the new AI structure have not yet materialized
- CEO expects improvement within 3–6 months. Meta plans to spend up to $145 billion on AI infrastructure in 2026
Restructuring under the AI banner
In May 2026, Meta carried out widely covered layoffs. According to Bloomberg, around 8,000 people left the company, while 7,000 were moved into various AI groups. One of those groups — Agent Transformation — was set up specifically to automate tasks previously handled by humans.
At the July 2 all-hands, Zuckerberg assessed that agent deployment is not matching earlier projections. He described the job cuts as not as "clean as they should have been," saying the leadership acted out of concern that the company would not adapt quickly enough to changes in the tech industry.
Gulag or normal transformation?
Earlier reports painted a harsh picture of Meta's AI unit. In June, TechCrunch reported that engineers moved into AI groups were describing the environment as a "soul-crushing gulag" — citing unclear goals, process overload, and a sense of disconnection from Meta's core products.
Zuckerberg did not directly address those accounts, but confirmed that the anticipated value of the new structure "hasn't come to fruition yet." He nonetheless expressed confidence that AI investments would begin delivering measurable results within the next 3–6 months.
Meta has set an AI infrastructure budget of up to $145 billion for 2026 — one of the largest such commitments in the industry.
Why this matters?
Meta is one of the few companies that deployed AI resources at scale and publicly admitted the results are behind schedule. That kind of statement from the CEO of a multi-trillion-dollar company adds weight to the ongoing debate about whether AI agents are genuinely ready to replace office work at corporate scale.
AI agents capable of operating autonomously in complex enterprise environments — managing data flows, coordinating tasks across departments, making decisions without human oversight — remain an unsolved problem. Benchmarks like SWE-bench show progress on narrow engineering tasks, but translating those capabilities to generalist enterprise settings has proved harder than the optimistic forecasts of 2025 anticipated.
For the industry, the admission carries two implications: it cools some expectations about the pace of automation, and it signals that even a company with a $145 billion budget and access to its own models (Llama) does not have a quick answer to the problem of agentic autonomy.
What's next?
- Zuckerberg signaled measurable AI results within 3–6 months of the all-hands (approximately Q4 2026) — no specific products or metrics were disclosed
- Meta plans to monetize surplus compute through its own cloud service — a move Reuters described on July 1, 2026 as a response to rising infrastructure costs
Sources
- TechCrunch — Mark Zuckerberg tells staff that AI agents haven't progressed as quickly as he'd hoped
- Reuters via Yahoo Finance — Exclusive: Zuckerberg says AI agent development not accelerating as expected
- Bloomberg — Meta begins job cuts in efficiency push spurred on by AI
- Reuters — Meta to sell excess AI computing capacity via cloud business
