Begin with a task your team can explain and observe. A process with a clear trigger, a known destination, and a person who owns the exceptions makes a better first trial than a broad goal like automating sales. Use recent examples to identify the steps the system can repeat and the decisions people should keep.
Find the repeated handoff
Ask the operator to walk through a normal case and a difficult one. Write down where the work starts, which information is copied, where it waits, and what closes the item. Separate a delay caused by missing information from time spent doing the task. A reminder may help the first problem, while prefilled data may help the second.
Rank frequency, friction, and consequences
Compare candidate workflows using the same period and the same unit of work. A recurring copy step with consistent inputs is easier to validate than a rare decision involving judgment. Note the consequence of a wrong action, the effort to recover, and whether a reviewer can catch an error before it affects a customer.
Choose one reversible trial
Write the start event, acceptable output, exception owner, and stop rule on one page. Start with a reminder, draft, or review queue when the final action is consequential. Run known normal and abnormal samples, including one with missing data. Keep a manual route available so the team can continue working while it learns.
Example
Compare three sample candidates: reminders, lead routing, and invoice field entry.
What to do next
Choose the first workflow when the owner can describe a successful run, a safe failure, and the condition that ends the trial.