Think about a time you've done research. What does that process look like from the time you get an assignment to when you turn in a completed project?
There is an image and a link below that show how some people have mapped this process, but in reality it is probably a lot messier and less linear.
After you have an idea of what your "map" of research would look like, now think about where AI could be helpful to make this process more efficient (and maybe where you don't want to use AI as well).