Automation projects are usually justified with a time saving. That framing undersells them, because it treats each process as independent when the real return comes from what connects.
Start where the data already is
The cheapest automation to build is the one whose inputs are already structured somewhere. Before automating a process, we ask what would have to become machine-readable — and often that step alone is the project.
Idempotency is not optional
Any job that cannot be safely re-run will eventually be re-run anyway, by someone under pressure, at the worst possible moment. Making replay safe is a small cost at build time and an enormous saving during an incident.
- Every job takes an explicit time window or record set
- Writes are keyed so a repeat is a no-op rather than a duplicate
- Partial failures leave the system in a valid state
- Replaying any window reproduces the same result
Instrument the gain
If you cannot show hours recovered and errors avoided on a dashboard, the automation programme will lose its budget to something more visible. Measure it from day one and the case for the next process makes itself.
The second automation is easier than the first. The tenth is easier than the second. That is the whole argument.