Define completion before launch
A call that ends without a transfer may still leave the customer waiting. For each workflow, define what finished means: a booking confirmed in the scheduling system, a technical issue verified as resolved, a complete claim intake, or a collections promise recorded with the right details. The metric should match the customer's need and the owning team's work.
Track the quality of that completion
Measure whether the agent used the correct record, followed the permitted path, and created the right downstream state. Look at corrections, repeat contacts, complaints, and inappropriate escalations. A high completion count is weak evidence if employees spend their time repairing the result.
Learn from a clear baseline
Before changing the workflow, record how it works now: waiting time, handling effort, error types, and the number of requests that actually finish. Then evaluate the agent on ordinary requests and hard exceptions. Review examples with the people who own the process, and expand only when the evidence supports it.
The purpose of measurement is not to produce an impressive single percentage. It is to know which work improved, which risk increased, and what should change next.