1. Start with the current trigger and outcome
Write down what starts the process and what successful completion looks like. A workflow such as enquiry handling may start with a form submission and end when the right person has accepted ownership and the customer knows what happens next.
Keep the scope narrow enough to observe. Mapping every process in the company at once usually creates diagrams that are impressive but hard to use.
2. Record the real steps
Follow a recent example and note where information is copied, decisions are made, approvals happen and people wait. Include spreadsheets, inboxes and informal messages if they are genuinely part of the workflow.
The purpose is not to make the current process look tidy. It is to expose where the business relies on memory, duplicated effort or unclear responsibility.
3. Define ownership and exceptions
Automation needs an owner just as manual work does. Decide who is responsible for the workflow, who handles exceptions and what should happen when required information is missing.
Design an exception route before launch. A process that works only when every input is perfect will fail in exactly the cases where a person most needs clarity.
4. Automate repetition before judgement
Good early candidates include copying known fields, sending routine confirmations, creating standard records and notifying the right person. Keep judgement-heavy decisions with a responsible person until the business has a clear rule and evidence that automation is appropriate.
Manual steps are not automatically bad. A deliberate human check can be the safest and most efficient part of the system.
5. Measure whether the workflow improved
Compare time, errors, handover delays and customer experience before and after the change. Novelty is not a benefit on its own.
Document what was built, where settings live and how to operate the fallback. An automation that only its creator understands is another dependency, not a resilient system.
