Daron Acemoglu won the 2024 Nobel Prize in economics, but a few months before that honor, he published a paper that made him few friends in Silicon Valley. Contrary to CEO promises of wholesale white-collar work transformation, Acemoglu estimated AI would provide only a modest productivity boost and wouldn't eliminate the need for human labor. His cautious take hasn't caught on—talk of an AI jobs apocalypse now surfaces everywhere from political rallies to grocery store conversations.

I spoke with Acemoglu to see if recent developments changed his thinking. He's skeptical about AI agents—tools that operate independently rather than just answering questions. Companies pitch these as one-to-many worker replacements, but Acemoglu calls that "a losing proposition." He points to jobs like x-ray technicians, who juggle 30 different tasks requiring constant switching between formats and databases. Agents struggle to handle the orchestration humans perform naturally.

Acemoglu has noticed something else: AI companies are aggressively recruiting economists. OpenAI hired Ronnie Chatterji from Duke in 2024, Anthropic assembled a team of 10 leading economists, and Google DeepMind just brought on Alex Imas from the University of Chicago. These companies want to shape the economic narrative around their technology—and that's precisely what worries him.

Acemoglu identifies usability as a critical missing piece. While anyone can chat with AI, it lacks the intuitive simplicity of transformative software like PowerPoint or Word. That helps explain why AI hasn't yet disrupted the job market or boosted productivity in measurable ways. He's now watching for genuinely user-friendly AI apps—but acknowledges we'll see conflicting evidence for years.