Big Tech

Zuckerberg Builds Personal AI Agent to Bypass Meta's Management Layers

Meta's CEO is testing an AI tool that retrieves info he'd normally get through layers of staff. The real target isn't him.

Liza Chan
Liza ChanAI & Emerging Tech Correspondent
March 23, 20264 min read
Share:
Abstract visualization of corporate hierarchy being compressed, with data streams flowing directly between top and bottom layers

Mark Zuckerberg is building a personal AI agent to help him run Meta, according to a Wall Street Journal report published Sunday. The tool, still in development, already lets the CEO pull information directly from across the 78,000-person company without routing requests through the usual chain of managers and team leads.

The internet, predictably, ran with this as "Zuckerberg building his own replacement." That framing misses the point entirely. The person making decisions isn't going anywhere. The people who used to gather information for that person? That's a different story.

What the agent actually does

The system works as an on-demand information tool. Instead of Zuckerberg asking a VP, who asks a director, who asks an engineer what the status of some project is, the agent retrieves the answer directly. Think less "robot CEO" and more "extremely fast search engine with access to everything inside Meta."

One source described it to the Journal as functioning like a combined chief-of-staff and analyst, surfacing context and decisions from across product teams. Zuckerberg apparently uses it to stitch together signals that would normally require multiple meetings to collect.

He's not the only one. Meta employees have been building their own agents too. An internal tool called MyClaw can access chat histories and work documents while communicating with colleagues or their AI agents. Another one, Second Brain, was built by a Meta employee using Anthropic's Claude and is described internally as an "AI chief of staff" for project work. Some of these agents now communicate with each other autonomously, which sounds impressive until you learn that one triggered a SEV1 security incident days before the WSJ story broke.

The flattening

This is where it gets uncomfortable for middle management. During Meta's January earnings call, Zuckerberg told investors the company is "elevating individual contributors and flattening teams." CFO Susan Li added that engineer output had risen 30% since early 2025, driven by AI coding agents, with heavy users seeing 80% gains year over year.

Those are Meta's own numbers, so take them accordingly. "Output" for engineers can mean a lot of things, and a 30% increase measured by the company deploying the tools isn't exactly an independent audit. But the direction is clear enough: Meta wants fewer layers between Zuckerberg and the work being done.

The company is already restructuring around this idea. A new applied AI engineering organization uses a 50-to-1 ratio of individual contributors to managers, according to an internal memo reported by the Journal's Meghan Bobrowsky. That's not a team structure. That's a deliberate experiment in whether you need managers at all.

So about those layoffs

Reuters reported on March 14 that Meta is considering layoffs affecting up to 20% of its workforce, roughly 15,000 people. Meta called it "speculative reporting about theoretical approaches," which is the kind of non-denial that makes reporters more confident, not less.

The timing here is hard to ignore. You don't build a CEO agent that bypasses management layers and simultaneously restructure teams to have 50 engineers per manager and then claim workforce reduction isn't on the table. Meta already shed about 21,000 jobs during 2022-2023's "year of efficiency" and cut another 1,000 Reality Labs roles earlier in 2026.

AI adoption is now factored into performance reviews at Meta. That's not a suggestion; it's a metric your job depends on.

The money behind the push

Meta expects to spend between $115 billion and $135 billion on capital expenditures in 2026, nearly double its 2025 outlay of around $72 billion. Most of it goes to data centers and AI infrastructure. The company also acquired AI agent startup Moltbook and Singapore-based Manas AI in recent months, and brought in Scale AI's former CEO Alexandr Wang to lead a new AI unit.

"We're starting to see projects that used to require big teams now be accomplished by a single very talented person," Zuckerberg said on the January call. That sentence landed differently for the 78,000 employees hearing it than it did for the analysts on the line.

Meta declined to comment on the CEO agent project. The company's next earnings report, where these restructuring plans could become more concrete, is expected in late April.

Tags:MetaMark ZuckerbergAI agentscorporate restructuringtech layoffsartificial intelligencemiddle managemententerprise AI
Liza Chan

Liza Chan

AI & Emerging Tech Correspondent

Liza covers the rapidly evolving world of artificial intelligence, from breakthroughs in research labs to real-world applications reshaping industries. With a background in computer science and journalism, she translates complex technical developments into accessible insights for curious readers.

Related Articles

Stay Ahead of the AI Curve

Get the latest AI news, reviews, and deals delivered straight to your inbox. Join 100,000+ AI enthusiasts.

By subscribing, you agree to our Privacy Policy. Unsubscribe anytime.

Zuckerberg Builds AI Agent to Run Meta, Bypass Managers | aiHola