When a team decides to automate, the loudest idea usually wins — the one a senior person is annoyed by this week. That is a poor selection method. The processes that pay back quickest are often boring, and the ones that make good demos are often the hardest to get right. A short scoring exercise replaces opinion with a ranked list.
Score each candidate process on five factors
- Volume — how often does it run? Below a few dozen times a month, setup effort rarely pays back. Daily or hourly processes compound.
- Rule clarity — can an experienced person write the steps down, including the common exceptions? If two experts would do it differently, the process isn't ready.
- Input quality — does the data arrive in a consistent form (a form, an API, a standard document), or as free text in every shape? Messy inputs raise cost sharply.
- Reversibility — if the automation makes a mistake, can it be corrected in one step before it leaves the building? Internal drafts are safer than outbound messages and payments.
- Owner and metric — is there a named person who feels the pain, and a number (hours, errors, days) that will visibly move?
Read the pattern, not just the total
High volume, clear rules, tidy inputs and reversible outcomes is the sweet spot: automate fully, monitor lightly. High volume with messy inputs suits assistance — a system proposes, a person confirms. Low volume with high stakes is a human job with better tooling, no matter how appealing the technology.
If a process scores low on rule clarity, the first deliverable isn't software. It's a written procedure. Automating an undefined process just makes the confusion run faster.
Run the exercise in an hour
List ten candidate processes, score each factor from one to three with the people who actually do the work, and total them. Argue about the bottom and top, not the middle. Pick the highest scorer whose owner is willing to change how they work. That last condition decides more outcomes than any technical factor: an automation nobody owns is a demo with a maintenance bill.