I have spent a lot of time thinking about how generative AI is changing the labor market and the skills my students and children (still 6+ years away from college) will need to succeed. What should I be emphasizing when so many traditional academic tasks can now be performed, at least passably, by an algorithm (other than the obvious “Learn how to use generative AI well”)? I know many others are wondering the same thing.

In an age when generative AI can draft essays, write code, summarize articles, and solve textbook-style problems, I think the most valuable skills—in any field of work—will be those that determine when and how to do something. Three stand out above the rest: determining direction, good judgment, and problem-solving.

Determining direction is about defining the problem in the first place. A student who can ask the right question will outperform one who simply knows how to carry out a technique, especially when that technique can be automated. In the workplace, the most valuable employees will not be those who can produce outputs quickly, but those who can identify the right problem to solve and adapt when conditions change. It’s easier than ever to flood the world with output: essays, code, apps, and even academic papers. In a world where Generative AI has expanded what’s feasible, the valuable skills will increasingly lie not in the realm of “Can we?” but “Should we?”

Good judgment means recognizing when an answer is incomplete or wrong. Generative AI systems are confident, fluent, but often wrong (just like economists, as the joke goes!). Knowing when to trust an output, when to verify it, and when to discard it altogether requires domain knowledge, critical thinking, and experience. Of course, you can use one AI to check output in some cases, but ultimately human judgment is needed somewhere in there.

Problem-solving is pretty self-explanatory. As wonderful as generative AI is, it’s not going to eliminate all problems and will definitely create new ones. Learning how to recognize when a problem has emerged and how to work through it—which tools are appropriate, who should be involved in helping solve it, and, ultimately, what is the right solution—will continue to be valuable.

Can these skills be taught? I think so. Exactly how will vary by field, but the common thread is that education must focus less on producing answers and more on interrogating them. Students should practice framing ambiguous problems, evaluating competing explanations, and identifying the limits of a given analysis. In other words, critical thinking becomes not a vague aspiration but a core practice: asking what assumptions underlie a result, what evidence would change one’s mind, and what risks follow from acting on a particular recommendation.

None of this implies that generative AI will leave the structure of the labor market unchanged. Some fields will grow, others will contract, and the mix of tasks within many jobs will certainly evolve. But these shifts reinforce rather than weaken the case for emphasizing direction, judgment, and problem-solving. When tools become more powerful and outputs easier to generate, the scarce resource is not production but discernment: knowing what work is worth doing and how to evaluate the results.