Anthropic CEO Dario Amodei says AI will kill more than 50% of jobs right now; but inside, one of his own top executives says the AI company can’t…
Antrophic CEO Dario Amodei wants the world to know that AI is coming for white-collar jobs—and fast. Half of all entry-level office roles could disappear within five years, he warned in a May 2025 interview with Axios. Unemployment could hit 10–20%. Governments and AI companies alike are “sugarcoating” the threat, he said.Dario Amodei wants the world to know that AI is coming for white-collar jobs—and fast. Half of all entry-level office roles could disappear within five years, he warned in a May 2025 interview with Axios. Unemployment could hit 10–20%. Governments and AI companies alike are “sugarcoating” the threat, he said. At the World Economic Forum in Davos earlier this year, Amodei made similar warnings, telling world leaders that AI’s impact on jobs would be far larger than most governments were prepared for. Fields like technology, finance, law, and consulting are first in line. And most people, he said, have no idea this is coming.But walk inside Anthropic’s own walls and the story gets more complicated. Amol Avasare, the company’s Head of Growth, recently said Anthropic is desperate for product managers. Claude Code—the company’s AI coding tool—has effectively tripled engineering output, but the humans directing all that work are overwhelmed.“PM and design is just squeezed. It’s absolutely squeezed,” Avasare said. “We need to actually hire a ton more PMs.”So the company warning the world about mass job elimination can’t fill one of the most traditional roles in tech. That contradiction says more about where AI disruption actually stands than any forecast.
Anthropic’s own research shows AI is nowhere near its full displacement potential—yet
In a March 2026 study, Anthropic researchers Maxim Massenkoff and Peter McCrory mapped the gap between what AI can theoretically do and what it’s actually doing in workplaces today. Computer programmers topped the list with a 74.5% “observed exposure” score—meaning AI is already performing nearly three-quarters of their tracked tasks. Customer service reps and data entry workers followed close behind.But zoom out and the picture changes. In the computer and math category overall, AI could theoretically handle 94% of tasks. Claude is currently covering just 33%. Legal, office admin, and finance roles show similar gulfs between capability and adoption. The researchers found no systematic rise in unemployment among highly exposed workers, though hiring of 22–25-year-olds into those roles has slipped about 14% since ChatGPT’s launch.
The most vulnerable workers aren’t who you’d guess—and the “software engineer” title may be next
The profile of the most AI-exposed worker is surprising. Anthropic’s data shows they earn 47% more than the least exposed group, are 16 percentage points more likely to be female, and nearly four times as likely to hold a graduate degree. Think lawyers, financial analysts, developers—not warehouse staff.Boris Cherny, the engineer who built Claude Code, predicted in February 2026 that “software engineer” as a job title will start fading this year. But even he sees the role morphing rather than vanishing—into spec writing, user research, and product thinking.Which loops back to Anthropic’s own hiring crunch. The company is telling the world AI will wipe out millions of jobs. Internally, it’s scrambling to fill the one role that all this AI output keeps making more essential.
