Mark Zuckerberg has never been someone who just talks about things. When Meta bet on the metaverse, he attended meetings in virtual reality and his avatar surfed on a hydrofoil. When Meta bet on smart glasses, he wore them in public regardless of how they looked. And now, as Meta bets on artificial intelligence, Zuckerberg is building his own AI agent to help him run the entire company.
Zuckerberg is working on what is known as a CEO agent, a tool that is still in development but is already saving him time. Specifically, it helps him obtain information faster without having to go through layers of people and wait for answers. Tasks that used to take hours and involve several intermediaries can now be handled by the agent on its own.
78,000 people and their AI assistants
This is not just Zuckerberg's personal project. It is part of a much larger plan. Meta wants to rebuild its entire organizational structure so that it is "AI-native from day one," as Maher Saba, head of the new engineering organization for applied AI, put it.
Meta's internal message board is filling up with posts from employees sharing new ways of using AI tools they have built themselves. The atmosphere inside the company reportedly resembles Facebook's early days, when the motto was "move fast and break things." Employees attend AI tutorials several times a week, participate in creative competitions, and build their own tools.
Some have deployed personal agents such as My Claw, which has access to their chats and work files and can communicate with colleagues on their behalf. Or even with their colleagues' AI agents. Yes, you read that correctly: there is a group on the internal platform where employees' AI agents chat with one another.
Another tool, Second Brain, was built by a Meta employee using the Claude model. Its creator described it as an "AI chief of staff." It can index documents, search projects, and serve as a smart assistant for everyday work.
AI usage as part of employee evaluations
At a January meeting, Zuckerberg said that Meta is "investing in AI-native tools so individuals can do more." At the same time, he announced that teams would be flattened and individual contributors promoted. The new engineering organization will have a structure in which as many as 50 people report to a single manager.
But this enthusiasm also has a downside. AI usage has become part of employee performance evaluations. Meta is among the technology companies that rank employees based on token consumption, meaning how much they use AI tools. Some employees describe this as liberating. Others admit that the rapid changes and pressure to use AI are causing them stress and fueling fears of layoffs.
And those fears are not entirely unfounded. Meta laid off more than 11,000 people in 2022 and another 10,000 in 2023. At the time, Zuckerberg declared a "year of efficiency." The company now has roughly 78,000 employees, and CFO Susan Li openly says that Meta must not operate less efficiently than AI-native startups with much smaller teams.
Zuckerberg leading by example: Does it actually work?
A survey by Stanford economist Nicholas Bloom revealed something interesting. Nearly 70% of CEOs and senior executives use AI at work for less than one hour a week. More than a quarter do not use it at all. Meanwhile, Gallup research shows that employees whose manager actively supports AI and uses it personally are more than twice as likely to use it regularly. And eight times more likely to find that it actually helps them at work. Zuckerberg knows this. That is why he is building his own agent.
But not everything is going smoothly. Meta recently experienced a security incident in which an AI agent deployed by an employee published a response without their consent. Another employee acted on it, and sensitive company data was accessible to people without the appropriate authorization for nearly two hours. Rapidly deploying AI tools comes with risks, and Meta has experienced them firsthand.
Meanwhile, Meta is not just buying technology; it is buying entire companies. It recently acquired Manus, a Singaporean startup that makes personal agents, as well as Moltbook, a social network for AI agents. Both acquisitions indicate where Zuckerberg is heading: he wants everyone to eventually have their own AI agent.
Sources: futurism.com and wsj.com
