One of the first things I learned in high school was that my math textbook had an answer key at the back. 

Which was weird. My impulse was I could shortcut my homework simply by writing the answer key down on a piece of paper.

Of course, the teacher required us to show our work in our tests and exams. She would give us part marks for the correct thought process, even if the final answer was wrong.

She set the incentives so that the answer key mattered a lot less. The answer key really was just there to help you verify that you got the right answer.

One way to see generative AI is like this: it’s like an answer key (that often gets things wrong, by the way!). It’s not sufficient to show up with it and expect full marks.

While ninth-grade math curriculum doesn’t change as often, business situations and products and behavior can change extremely quickly. A static answer key isn’t necessarily as useful. 

That means the answer will change often. You’ll need to understand how to prompt AI, get a sense of what the AI is doing correctly (and how to correct the incorrect parts, and train it for the future), and also to know when to take instruction from the AI.

At the end of the day though, your thought process is still what matters most. 

Your answer will change as the situation changes. You need to understand what the AI is doing, so you know what to do with it. Your superpower is your ability to be held accountable, and to hold others accountable.

P.S., People used to see AI as a slightly high intern. I think that’s changing quickly.