Meta is training an AI replica of Mark Zuckerberg to sit in meetings, advise staff, and help dismantle the management layers that currently separate the CEO from the people doing the actual work. Up to 16,000 jobs may disappear as a result.
This isn’t some glorified chatbot. According to the Wall Street Journal, the AI version of Zuckerberg is already pulling answers for him that “he would typically have to go through layers of people to get.” It’s trained on his tone, his public statements, his decision-making patterns, his understanding of company strategy. At some point soon, a meaningful chunk of Meta’s workforce could be receiving direction from software built to sound like their CEO, and that CEO won’t have the faintest idea who any of them are.
SFist covered the story when the Financial Times first reported it. The detail that won’t leave me alone: the clone isn’t just a time-saving convenience for Zuckerberg. It’s tied directly to his stated goal of changing “the day-to-day jobs of its employees to remain competitive with AI-native startups with much smaller staffs,” per the Journal. That’s not efficiency language. That’s headcount language dressed up in efficiency clothing.
The numbers already tell the story. Meta had roughly 79,000 employees as of December. In 2023, Zuckerberg declared it “the year of efficiency.” The company cut around 21,000 jobs that year. Now reports suggest another round is coming, potentially eliminating 16,000 more positions. Do the math.
I reached out to a friend who works at a mid-sized tech company in SoMa. Not Meta, but close enough to feel what’s coming. “The AI usage metrics,” she said, “started as optional. Now they’re in my review.”
That matches what the Journal found internally at Meta. AI usage by employees has been folded into performance evaluations. You’re not just expected to do your job well. You’re expected to demonstrate that AI is helping you do it, which implies, pretty directly, that you should need fewer human colleagues going forward.
The acquisitions point the same direction. In March, Meta bought Moltbook, a Singapore-based company that built a social network for AI agents, so that employees’ personal AI assistants can coordinate with each other. Meta also picked up Manus, a startup that builds agents designed to execute tasks on behalf of individual users. The Federal Trade Commission’s merger database is going to get crowded. Researchers at MIT’s Work of the Future initiative and economists at the Economic Policy Institute have both tracked how rapidly AI adoption is compressing hiring timelines across the tech sector, and Meta’s moves fit that pattern exactly.
Think about what this architecture actually looks like in practice. Individual workers get AI agents. Those agents communicate with each other through Moltbook. Above them, the AI Zuckerberg absorbs strategy and pushes guidance downward through the structure. The humans aren’t gone yet, but it’s hard to look at what’s being built and conclude they’re central to the design.
In 2026, a lot of companies are making similar noises about AI and productivity. Most of it doesn’t go much further than a slide deck. Meta’s different because it’s already spent the money, cut the staff, acquired the infrastructure, and built the tools. The 04 in the date on that SFist story isn’t a coincidence. This is happening now, not as a concept.
What nobody’s answered is the question that actually matters for the 79,000 people who currently work there: when the AI Zuckerberg is fully operational and every employee’s performance review tracks their AI usage, what’s the number they’re optimizing toward?
My SoMa source put it plainly enough. “Started as optional. Now they’re in my review.”