When to Use GenAI: A Personal Decision Tree
Everyone is increasingly affected by generative AI (GenAI) in their everyday tasks these days. In this case I refer to any tool that can generate text, images, program code, etc. I have a lot of discussions revolving around when and when not to use these tools, especially also when it comes to learning. The following is my personal internal decision tree that I currently more or less employ. It will be subject to pruning, growth, and updates in the future, caused by further developments and changes in my own perspective. The idea of this site is that it actually behaves more or less like a living tree, so revisit later to see the changes if you like.
When to use GenAI?
- Does something need to be done quickly or in limited time?
- yes: *Do I want to learn something in the process?
- no: Does an approximate solution with some errors suffice?
- yes: Use it however I see fit.
- no: Can I make an informed decision on whether the output is correct?
- yes: Use it and check the result.
- no: Can I find somebody that can?
- yes: Use it and let them check the result.
- no: Don’t use it.
- yes: Use it sparingly and only for inspiration and search; try to solve every problem myself first.
- no: Does an approximate solution with some errors suffice?
- no: Do I enjoy the process of doing the thing?
- yes: Just do it myself and on my schedule.
- no: GOTO *
- yes: *Do I want to learn something in the process?
Some further statements
When familiarizing with a new subject, use it as a kickoff to get an overview, but definitely read some primary resources as well!
To test one’s knowledge on a subject, e.g., the relation between two concepts, query the AI and based on the result use manual search (e.g., Google and Wikipedia) to double check.
If something needs to be done quickly and could be done with it, but I cannot check the result, the thing cannot be done quicker than I am.

