When people think about automation they think about the complicated parts of the business. That is almost always the wrong place to start.
The right filter
Three questions, in this order:
- How often does it happen? Fifty times a week beats twice a month, regardless of how long each one takes.
- How much judgment does it need? If it is the same decision every time, it is a candidate. If it depends on context a person is reading, it is not.
- What does it cost when it is skipped? Some tasks are dropped under pressure and the cost only shows up later. Those are worth more than their time.
Rank by the product of those. The winners are usually unglamorous.
What tends to come out on top
- Confirmations and reminders.
- Follow-up on anything unanswered.
- Moving the same information from one place to another.
- Routine status updates.
- Requests for reviews or feedback.
- Recurring reports.
- Recurring reminders for anything with an interval.
None of those are interesting. All of them happen constantly and all of them get skipped when a day gets busy.
What to leave alone
- Anything requiring genuine judgment about a specific person's situation.
- Anything where being wrong is expensive.
- Anything a customer would be annoyed to discover was automated.
- Anything that happens rarely: the build cost is never recovered.
The mistake in the middle
Half-automating something so that a person still has to check every output is often worse than not automating it, because you now have both the system and the manual work. Either automate it properly with escalation for exceptions, or leave it manual.
Sequencing
Do one, get it working, measure it, then do the next. The temptation is to design the whole system before building anything, and it consistently produces something that does not match how the business actually operates.
The test afterwards
If the team stops using the manual process voluntarily, it worked. If they keep a shadow spreadsheet alongside it, it did not, and the reason they give will tell you exactly what to fix.