Mark Zuckerberg is building an AI agent to help him be CEO.
According to a Wall Street Journal report, the system already functions as an on-demand information tool, finding answers for Zuckerberg “that would normally require going through layers of people.” It’s not a chatbot for scheduling meetings. It’s a tool designed to bypass traditional management hierarchy.
The agent remains in development, but Zuckerberg’s personal experiment reflects a broader transformation at Meta. The company is deploying AI tools across its 78,000-person workforce with explicit performance incentives attached.
The Tools Meta Employees Are Using
Zuckerberg’s CEO agent isn’t an isolated project. Meta has rolled out two major internal AI systems that are reshaping daily work:
MyClaw grants employees access to internal files and chat logs through a conversational interface. Workers can query colleagues—or their AI agent counterparts—without navigating bureaucratic handoff points. Need a document from another team? Ask MyClaw instead of emailing a project manager.
Second Brain builds on Anthropic’s Claude infrastructure and functions as a “personal chief of staff.” It organizes tasks, surfaces relevant information, and streamlines retrieval of institutional knowledge. Think of it as a memory layer for work context that would otherwise be scattered across Slack threads, documents, and meeting notes.
Both systems share a common design philosophy: reduce the coordination overhead that slows large organizations.
The Productivity Numbers
Meta claims the approach is working. CFO Susan Li reported that output per engineer has risen 30% since early 2025, primarily driven by AI coding agents. The gains aren’t evenly distributed—“power users” who’ve fully adopted the new tools are seeing 80% annual output increases.
Content moderation offers another data point. Meta has been transitioning enforcement from third-party vendors to proprietary AI systems, with AI catching 5,000 scam attempts per day that previously slipped past human reviewers.
During Meta’s January earnings call, Zuckerberg declared 2026 would be when “AI starts to dramatically change the way” Meta operates. He suggested “projects that used to require big teams can now be accomplished by a single, very talented person.”
What This Means for Middle Management
The implications are explicit, not subtle. Meta’s stated goal is to elevate individual contributors while flattening teams.
When information flows directly from systems to decision-makers, the traditional role of managers as information brokers becomes redundant. If Zuckerberg can query an AI agent to understand what’s happening across the company, why does he need multiple layers of executives to aggregate and relay that information?
The company is signaling this shift in how it evaluates performance. AI tool adoption has become a factor in employee reviews. Embrace the tools and your output numbers improve. Resist them and you’re left explaining why your metrics lag behind colleagues who’ve adapted.
This creates a selection pressure. Over time, the workforce composition shifts toward people comfortable working with—and through—AI systems.
The Infrastructure Bet
Meta is backing this vision with capital. The company forecasts $115-135 billion in 2026 capital expenditure, nearly double its 2025 spending of $72 billion. Much of that goes toward AI infrastructure.
The organizational changes complement this investment. Meta recently established Meta Compute as a new top-level organization led by Santosh Janardhan and Daniel Gross—the latter recruited from Safe Superintelligence. The move signals that AI infrastructure is now a core corporate priority, not an R&D experiment.
The Dreamer Acquisition
The same week Zuckerberg’s CEO agent made headlines, Meta announced it was acqui-hiring the entire team from AI startup Dreamer. The founders—including former Meta VP Hugo Barra—are joining Meta’s Superintelligence Labs under Chief AI Officer Alexandr Wang.
Dreamer had raised $56 million at a $500 million valuation to build tools for creating personal AI agents. Now that capability lives inside Meta.
The timing suggests a coordinated strategy: internal AI agents for Meta’s workforce, and the talent to build the next generation of those systems.
What This Means for Everyone Else
Meta isn’t the only company deploying internal AI agents, but it may be the largest to do so with explicit organizational restructuring attached.
The model is straightforward: give employees AI tools that increase individual output, then adjust headcount and management layers accordingly. The “flattening” Zuckerberg describes isn’t metaphorical—it’s a structural change enabled by technology.
For other companies watching, the question isn’t whether this approach works at Meta. The question is whether similar tools will create similar pressures across industries. If AI agents can absorb information-coordination tasks currently performed by managers, those roles face the same displacement dynamics as any other automation target.
The 30% productivity gains Meta reports aren’t magic. They come from tools that handle retrieval, synthesis, and coordination that previously required human labor. As those tools spread, the labor they replace becomes visible.
Zuckerberg is building his own AI assistant. But the more significant story is that he’s building 78,000 of them—one for every Meta employee. And the company is already restructuring around that reality.