Mark Zuckerberg Is Building an AI Agent to Do His Job For Him

Mark Zuckerberg is building a personal AI agent to help him run Meta. The tool, still in development, retrieves information he would normally have to get through layers of people, effectively compressing the management chain between the CEO and the data he needs to make decisions. It’s the most literal version yet of what every tech company says it’s doing with AI: replacing human intermediaries with software.

Meta has made AI adoption a factor in employee performance reviews. The internal message board is now filled with employees sharing AI tools they’ve built, and the atmosphere has been compared to the early Facebook days when the motto was “move fast and break things.” Employees are using personal agent tools like “My Claw” that can access their chat logs and work files, talk to colleagues on their behalf, or even talk to their colleagues’ AI agents directly. There is literally an internal group where employees’ personal agents talk to each other.

Another tool gaining traction internally is called “Second Brain,” built by a Meta employee on top of Anthropic’s Claude. It indexes and queries project documents and is described as “an AI chief of staff.” Meta also recently acquired Manus, a Singapore-based startup that builds personal task-execution agents, and Moltbook, the social media platform for AI agents. Both are being integrated internally.

The structural changes are just as aggressive. Meta established a new applied AI engineering organization designed to be “AI native from day one,” with an ultraflat structure of up to 50 individual contributors reporting to a single manager. Zuckerberg said on the January earnings call that the company is “elevating individual contributors and flattening teams.” Meta’s CFO Susan Li said at a conference this month that her priority is making sure a 78,000-person company doesn’t “work any less efficiently than companies that are AI native from the start.”

That last line is the tell. Meta is a company with nearly 79,000 employees trying to compete with AI startups that have 50. The response is to make every employee an AI-augmented worker and then see how many of those employees you still need. One employee described the current era as “fun and empowering.” Others said the intense focus on AI use has fed anxiety about layoffs. Both descriptions are probably accurate. Meta cut 21,000 jobs between 2022 and 2023 during its “year of efficiency.” Head count has since climbed back to 78,865. Whether it stays there depends on how well the AI tools work, which is exactly what the performance reviews are now measuring.

The broader pattern is unmistakable. Block cut 40% of its workforce citing AI. Crypto.com cut 12%. OpenAI is doubling its headcount to 8,000. The companies building AI are hiring. The companies deploying AI are firing. Meta is trying to be both at once: building the models and using them to eliminate the need for the people who used to do the work the models now do. Zuckerberg building an AI agent to help him be CEO is the headline. The 78,000 employees whose performance reviews now depend on how well they use AI is the story.