My current framework for whether or not to use ai is the difference between incidental versus essential skills. 

An incidental skills is like writing the citations correctly. It’s grunt work. 

Essential skills are like understanding code or communicating well.

Incidental skills should be automated, but essential skills shouldn’t because expertise creates a distribution of outcomes from the prompt. 

i.e. the more expertise I have with the essential skills the better the output of the llm

So I want to do the essential skills myself to get better at it rather than automating it, and use technology to scale myself.