When you ask AI to debug a backend issue, what output is most useful?

When you ask AI to debug a backend issue, what output is most useful?

We've seen a few different expectations:

- probable root cause

- likely affected files

- suggested fix

- prevention advice so it doesn't happen again

Which one actually saves you the most time?