The everyday problem
A pilot built only from clean examples may look successful while telling the team little about everyday operations. Hospital workflows encounter missing inputs, competing priorities, and dependencies on other systems. These conditions should be part of the evaluation rather than discovered after expansion.
A practical approach
Include representative exceptions in the test plan and define what should happen for each. Measure whether the workflow recognizes a blocker and routes it correctly, not merely whether it completes straightforward cases. Make the criteria for pausing or expanding the pilot explicit.
Try this with your team
Build three fictional cases: complete information, missing information, and conflicting information. Write the expected next action before reviewing the pilot’s output.