Start with the work people repeat
Choose a workflow that happens often enough to matter and is stable enough to describe. It might be handling an enquiry, preparing a project brief, updating a record, or checking incoming information.
Describe the trigger, the information needed, the normal actions, the person responsible, and the final useful outcome. This gives the team a shared picture before tools are discussed.
Make exceptions visible
The normal path is not the whole workflow. Real work includes missing details, unusual requests, duplicate information, policy questions, and decisions that need human judgement.
Write down what should happen when the normal path breaks. An exception can request missing information, pause an action, or move work to a named person instead of being forced through silently.
Automate only the safe, clear part
The first improvement may be a clearer form, a structured handoff, an alert, or a review list—not a complex automated system. Start where the inputs, rule, and owner are clear.
After the route is visible, review the exceptions and corrections. They show whether the next improvement belongs in the process, the information, or the automation itself.