Start with a shortlist
Common candidates include enquiry routing, appointment reminders, CRM updates, document filing, routine reports, internal notifications and follow-up reminders. Treat these as possibilities to assess, not processes that should always run without human involvement.
Map the trigger and the exception
For each workflow, record what starts it, where the information comes from and the intended outcome. Then ask what should happen with incomplete information, a duplicate request or an unavailable service.
Make failures visible
Name an owner, keep useful status information and create a practical recovery path. An automation that quietly stops can create more confusion than a manual process. Test retries and duplicate handling before depending on it.
Review the whole journey
Removing a single step is not always the same as improving the process. Check whether the result reaches the right person, reduces repeated work and remains understandable to the team who will operate it.