Generally, I view the problem as one of managing potential tech debt. You’ll never get anything done if you don’t occasionally build temporary solutions to test things out, or bang out a quick one-off using some SQL+Excel pivot tables. The trick is to be aware of whether you’re generating future work for yourself or not.

As a rule of thumb, I’m happy to have people come request one-off reports and projects from me. If I find that they come back a 2nd or 3rd time (or they clearly need it regularly), then it’s worth the effort of putting in some effort to have some semi-clean auotmation around it. Otherwise a quick one-off is enough and helps them accomplish what they want.

If something is tied to a critical system or process, like it’ll be used in production, then there’s a much stronger case to work with engineering and build something much more robust.

Be pragmatic about things, if someone wants something but it doesn’t power any decisions and doesn’t potentially yield useful insight, spend less (or no) time on it.