Key facts
- Mollick said the standard advice to find AI-proof work such as plumbing is not useful because robotics experts have changed their view of what robots can do.
- Mollick named four advantages: deep knowledge, wide knowledge, taste and agency.
- Mollick said prompting has started to look more like managing, with users specifying the product, how to test it and what to show partway.
- Mollick said the apprenticeship model broke because junior employees' best bet is to use AI for everything and managers see less reason to use interns.
- Mollick said either more formal trade-school-style education or serious corporate mentorship is needed to train junior workers.
Ethan Mollick, a Wharton management professor, said in an interview with Peter Kafka on Kafka's Channels podcast that the common advice to find work that cannot be automated, such as plumbing or nursing, is not useful. "I don't think it's very useful," Mollick said, adding that robotics experts he knows have moved from saying robots are unrelated to large language models to saying there is something to the technology.
Mollick's book, "Co-Existence," argues that AI models are improving faster than most people understand and that the only response is to learn to live with them. He pointed to coding as the field most changed by AI, saying it is not clear whether there will be more or fewer coding jobs, but the work will keep changing. Writing error-free code matters less than conceiving what to build, he said.
He named four attributes that give workers an advantage: deep knowledge of one area, wide knowledge across many, taste in selecting among AI output, and agency. On agency, Mollick said prompting now resembles managing, with users specifying the product, how to test it and what to show partway, rather than relying on magic words.
Mollick said the harder problem is institutional. Apprenticeships at banks, consulting companies, audit firms and car dealerships trained junior workers through poorly paid grunt work in exchange for skills and contacts, but the incentives no longer align. Junior employees' best bet is to use AI for everything, he said, while managers see less reason to use interns. He said the fix is either more formal, trade-school-style education for white-collar work or companies treating mentorship as a real activity that costs time and productivity.