CEO's typically make decisions when there isn't enough data to support the obvious direction. In this way, they are anti-pattern finders that rely on gut or some anecdotal experience. They are most often wrong, but when they are right and it's a success they are considered genius. I'm not sure an AI can make a non-obvious decision based on feeling.

> I'm not sure an AI can make a non-obvious decision based on feeling.

You just described that CEOs are like broken clocks: they are mostly wrong, but sometimes they are right by chance.

How do you conclude that AIs can't do that? If it's about eloquently phrasing a random idea, AIs are perfect.

Did you find any way to generate good ideas with AI? Anecdotally, I found it incredibly unsatisfactory in this sense. It normally re-hashes old ideas without any internal coherence, like those novelty websites combining two existing startups to create a random mission (e.g. “AirBnB for motorbikes”).

When you consider that LLMs are trained on what can be scrapped from the web, the so called "creativity" comes from mashing together less commonly co-associated ideas.