Mark Zuckerberg is building a personal AI agent to help him run Meta, in a move that says as much about the company's internal restructuring as it does about its chief executive's working habits.
The system is designed to help Zuckerberg pull in the information he needs without waiting for it to travel through the usual chain of briefings, intermediaries, and management layers. The tool is still in development but is already in use, and Meta is evaluating whether to roll out a similar approach more broadly across the business.
Flattening the organisation
The CEO agent fits a wider internal agenda. At a January earnings call, Zuckerberg said Meta was investing in AI-native tooling so individuals could get more done, and that the company was elevating individual contributors and flattening teams.
That is not just rhetoric. Meta has incorporated AI-driven impact into its performance reviews and is among the tech giants that have established internal leaderboards ranking employees by their consumption of tokens, a measure of AI use. Zuckerberg has also returned to coding to stay actively involved in Meta's AI transition.
The tools already in use
The CEO agent is the most visible part of a broader internal deployment. Employees are already using My Claw, which can access chat logs and work files and communicate on behalf of users, and Second Brain, an AI assistant that indexes and queries project documents. The culture around these tools has taken on an experimental quality, with staff participating in AI hackathons and deploying personal agents that handle work on their behalf.
The risks that come with speed
The push to embed AI into decision-making is not without complications. In one reported incident, an AI agent responded to an employee's request for help on an internal forum and the employee followed that guidance, exposing sensitive data internally for about two hours. Meta said no user data was ultimately mishandled, but the episode showed how quickly an internal tool can create problems once it starts influencing real decisions.
What it signals beyond Meta
The project raises broader questions about the future role of human leadership and how organisations will balance human judgment against AI accountability. Anthropic chief executive Dario Amodei has described AI as a general labour substitute. Google's Sundar Pichai has said AI could replace his own role within a year.
Related reading
- Microsoft introduces Critique in M365 Copilot
- Microsoft warns AI agents risk becoming "double agents" as it unveils security controls at RSAC
- Microsoft launches Copilot Cowork in Frontier
For Zuckerberg, the more immediate point is competitive. Meta faces pressure from smaller, AI-native startups that operate with leaner teams and faster decision cycles. With AI spending projected to exceed $100bn by 2026, the company is betting that deploying these tools at every level, starting at the top, is how a 78,000-person organisation keeps pace.
Building the agent himself is, among other things, a signal to the rest of the company about how seriously that bet is meant to be taken.
The recap
- Meta developing AI assistant to aid CEO Mark Zuckerberg.
- Company uses AI tools across 78,000 employees to flatten structures.
- Tool is still being developed and used to test integration.
