Record the before state
Measure a representative period before making changes. Document task volume, time per task, error types, and the effort to repair them. Keep unusually busy or quiet periods visible.
Define a correct result
For lead routing, check correct ownership and duplicate records. For meeting notes, check confirmed actions and incorrect commitments. Choose measures that reflect the workflow’s purpose.
Track maintenance and exceptions
Log failed runs, manual interventions, incorrect outputs, and the time to investigate. A workflow that silently loses inputs can look efficient if you only count completed outputs.
Set a stop rule
Decide in advance what would make the team pause: a consequential error, a missing audit trail, an unstaffed queue, or consistently negative net time. Review results with the owner before increasing scope.
A good pilot produces a decision, even when that decision is to stop.