Where this work earns its keep.

The way work moves is specific enough that packaged software forces awkward compromises, and a small gain per job compounds across thousands of them. The case here is arithmetic rather than ambition.

The point where the workbook that ran the business becomes the thing holding it back. Too many hands in it, no history of changes, no controls, and a growing amount of judgement living in one person's head. The risk is concentration as much as inefficiency.

Software that still works but costs more each year to sustain, or that no remaining member of staff can safely amend. Replacement is never urgent until it is, which is the worst moment to begin. Planning early is most of the saving.

An idea with real demand behind it and no agreed line between what version one is and what version four might be. Most of the value created here comes from deciding what to leave out and holding that decision under pressure.

Two or three tools that each do their job while somebody enters the same information twice. The integration is usually smaller work than expected and pays back faster than anything else on this list. It is also the easiest thing to keep postponing.