How do you decompose knowledge work for AI

I hacked a bit on an AI experiment: idea generator.

In general, working on these AI apps makes you think a lot about the work process and unique value add of Knowledge workers in general, and myself specifically. How predictable is it? Which parts are easy to automate; which are easy to assist? How much Context -- and what exactly -- does the work require?

And specifically for the idea generator: how do I generate and evaluate ideas? Sometimes I try to generate a whole list following a single prompt ("how could I solve this technical problem?"), and it works. But sometimes this takes me nowhere and then I take a step back and try going into diffuse mode. Or sometimes I look for Inspiration elsewhere. If I had to write a Python function that returns inspiring text for putting into GPT context... how would I do that?

These are very interesting product design questions for building any Copilot style product. And also interesting questions into the nature of how humans operate at work.