A simple framework for separating useful automation from expensive technical theatre. The useful answer depends on the company’s workflow, current systems, team and cost of staying as it is.
Look for volume and repetition
The best candidates happen often, follow clear rules and consume enough time to matter.
Include exceptions from the beginning
A process is not automated when the happy path works but every unusual case disappears into a technical queue.
Keep human judgement where it adds value
Automation should prepare information and move routine work. People should still handle ambiguity, relationships and material risk.
Measure the whole process
Saving two minutes in one step is irrelevant if the automation creates ten minutes of correction later. Measure completion time, errors and handoffs end to end.
Make the next step proportionate.
A short discovery, audit or prototype can often answer the highest-risk question before a company commits to a large build. The purpose is to reduce uncertainty, not create a long consulting phase.