Inside Meta's internal message board, a group called "agent-to-agent" is active — this is not sci-fi, but a reality revealed by The Wall Street Journal on March 23. The tech giant with 78,000 employees is redefining what "work" means faster than outsiders realize.
Zuckerberg's CEO Agent: Bypassing Bureaucracy for Direct Answers
Mark Zuckerberg is building a "CEO Agent" to handle daily tasks. Still in development, its primary function is to fetch information faster — querying internal systems directly to get answers that previously required multiple layers of communication. In a company of nearly 80,000 people, even the CEO often had to navigate organizational layers to obtain specific numbers or decision context. AI agents essentially replace those intermediaries. Zuckerberg's broader goal is for every employee — and eventually everyone outside Meta — to have their own personal AI agent.
My Claw and Second Brain: Employee AI Assistants
Meta employees can now use a personal agent tool called My Claw. It accesses chat history and work files, and communicates on behalf of the employee — more precisely, it communicates with colleagues' AI agents. This gives rise to agent-to-agent interaction: when your agent talks to my agent, the human steps back, and information flows at a completely different speed and efficiency. Another tool, Second Brain, indexes internal documents, allowing employees to quickly locate knowledge scattered across the organization. Together, they form Meta's "AI-augmented employee" infrastructure.
AI Usage in Performance Reviews, Engineer Output Surges 80%
More significantly, Meta has incorporated AI tool usage into employee performance evaluations. This goes beyond encouragement — it's a structural push: those who use AI well directly impact their reviews. Meta CFO Susan Li said that since early 2025, overall engineer output has increased by 30%, driven primarily by AI coding agents; among heavy users, that figure reaches 80%. Zuckerberg once said: "Projects that used to need large teams can now be done by one excellent person." That's no longer a vision — he's delivering on it with data.
Acquisitions of Manus and Moltbook: External AI Agent Ecosystem
Meta's AI agent ambitions extend beyond internal tools. In December last year, Meta acquired Chinese AI agent startup Manus at a valuation of roughly $2 billion — Manus is known for autonomously completing complex multi-step tasks. Earlier, Meta also acquired Moltbook, a social platform focused on AI agent interaction. Founders of both companies have joined Meta Superintelligence Labs. From internal tools (My Claw, Second Brain, CEO Agent) to external acquisitions (Manus, Moltbook), Meta is building both the supply side and demand side of AI agents — creating infrastructure for agents to run, and exploring how agents form an ecosystem.
Organizational Flattening Becomes Real
Together, these moves reveal a structural shift deeper than "AI productivity tools": Meta is replacing organizational layers with AI agents. Historically, middle management partly existed for information relay, aggregation, and coordination. When a CEO's agent can directly retrieve the answer, and when an employee's agent can negotiate directly with another employee's agent, flattening the organization is no longer just a management philosophy — it's technologically achievable. For the crypto and Web3 industries, this trend is noteworthy: decentralized protocols have long pushed for disintermediation, and Meta is doing the same within a traditional enterprise using AI agents — just via a different path. As the "agent economy" moves from concept into the daily operations of the world's largest tech company, the paradigm shift is only beginning.