Look for predictable repetition
Copying enquiry details between systems or sending a routine reminder can be a useful starting point. Choose work with a clear trigger and a stable result. A process that changes every week may need clarification before automation. Ask the people doing the work where they repeat information and where they rely on judgement; these are different kinds of problems.
Map the unhappy path
Write down what should happen when a record is missing, a customer uses a different email address, or an external service is unavailable. Decide whether the workflow should retry, pause, or ask someone to review it. An automation that silently skips failures can create more work than it saves. Name the person or team who will receive and resolve exceptions.
Limit access to what is needed
List the systems and fields the automation needs to read or change. Avoid broad access simply because it is easier to configure. Keep credentials outside shared spreadsheets and establish how access will be removed if the workflow is retired. For actions that send messages or modify important records, include a review step until the behaviour is understood.
Measure a small pilot
Run the workflow with a limited set of records and compare its output with the current process. Track failures, duplicate actions, and manual corrections, as well as time spent. Agree on a way to pause the automation without losing incoming work. Expand only after the team understands both the normal path and the recovery steps.
