Four criteria
- Happens often — daily or weekly, so the saving is visible quickly
- Well understood — someone can describe every step
- Low consequence if it gets something wrong
- Wanted by whoever currently does it
The fourth matters more than people expect. Automation imposed on someone who did not ask for it gets worked around; automation someone wanted gets defended.
Good first candidates
- The report assembled from several systems every week
- The file that arrives daily and gets reformatted
- The data re-keyed between two systems
- The check somebody performs manually
- The routine notification somebody sends
What to avoid first
- Anything customer-facing without review
- Anything financial without a validation layer
- A process nobody can fully describe
- Something the business is about to change
- The exception cases rather than the routine ones
Measure before you start
| Record | Why |
|---|---|
| Time per occurrence | The baseline |
| Frequency | The annual value |
| Error rate today | What improvement means |
| Who does it | Whose time is freed |
| What they would do instead | Whether the saving is real |
Two weeks of measurement before building costs almost nothing and it is the only baseline you will get.
Then the second one is easier
The first automation builds the monitoring pattern, the deployment approach and the organisational confidence. The second costs less and can be more ambitious for the same reason.
That compounding is the real argument for starting with something small and finishing it properly.