Mark Zuckerberg develops a personal AI to help him run Meta

At Meta, AI is no longer just used to launch products or boost advertising. It also begins to touch the heart of internal power, where decisions are made.

An AI plugged into the boss’s desk

Mark Zuckerberg is developing a personal AI agent to help him manage Meta faster and with fewer intermediaries. This tool is still under construction, but it is already used to speed up the search for information by bypassing several hierarchical levels. The idea is simple, almost brutal: get the right data without making the request travel throughout the entire company.

This detail says a lot more than it seems. For years, the promise of AI in business has mostly been about customer support, coding or office productivity. At Meta, it now goes up to the command post. AI is no longer just a work tool. It becomes a management layer.

This choice fits with an obsession that Zuckerberg has displayed for several months: making Meta faster, more direct and more competitive against much lighter AI startups. When a company has nearly 79,000 employees, each added level slows down the machine. The personal agent imagined for the CEO therefore resembles a technological shortcut, but also a very clear cultural signal.

Meta wants to flatten his own machine

At the end of January, Zuckerberg had already set the scene. When releasing quarterly results, he touted 2026 as the year AI would begin to fundamentally transform the way Meta works. This sentence, at the time, might have seemed abstract. Today it is becoming much more concrete.

The Wall Street Journal also describes wider use of agents within the group. Tools like MyClaw provide access to work files and chat logs. Another system, Second Brain, is billed internally as a sort of AI chief of staff and builds on Anthropic's Claude infrastructure. In other words, Meta is not experimenting with an isolated gadget for its founder. The company is testing a new way of circulating information.

This is where the subject becomes strategic. If employees consult agents to retrieve data, organize a project, or get context in seconds, the value of the layers in between changes. It does not disappear automatically. But it must now be justified other than by the simple transmission of information. In many large groups, this is precisely where AI begins to shake up the established order. This reading is an inference based on the way in which Meta deploys its internal tools and on the stated objective of making work more fluid.

The promised efficiency also opens up an area of ​​tension

This shift occurs in a sensitive context. Meta could prepare a new layoff plan likely to exceed 20% of the workforce, even if the final scale and timetable have not yet been decided. The group rejected this reading as a speculative and theoretical scenario, without confirming such a plan.

Why is this suspicion growing so much? Because Meta is spending massively on AI. Reuters is talking about up to $135 billion in investments in 2026, in a context where the markets are still accepting the effort, but are closely monitoring the real return on this race. If AI improves productivity, the question of the number of employees automatically returns to the table.

The bottom line is therefore less technological than organizational. Zuckerberg isn't just looking for more powerful AI. He's looking for a more high-energy business, with less friction, fewer deadlines and perhaps fewer human relays between idea and execution. This is an attractive promise for investors. For employees, it can also resemble an era of permanent sorting, where AI becomes at the same time co-pilot, filter and silent judge of the usefulness of each person. This is why Jamie Dimon warns about the real impact of AI on global employment.