Test Support Triage Before Scaling
Design a small PDSA cycle that tests routing value and urgent-case safety before live automation.
18 minutes
Scenario
CloudDesk wants an AI classifier to route incoming support tickets. Leaders want a full launch after a vendor demo.
- Your role
- Service-improvement lead
- Method
- PDCA/PDSA Cycle
Evidence pack
Baseline accuracy
Agents correctly route 81% of tickets on first assignment.
Lead time
Median queue-to-agent time is 42 minutes.
Urgent cases
6% of urgent tickets are initially misrouted.
Volume
The service receives 3,000 tickets weekly across twelve categories.
Concentration
Three common categories represent 64% of volume.
Vendor claim
A demo dataset produced 94% overall accuracy; urgent-case sensitivity was not reported.
Review effort
A supervisor can review about 120 predicted labels per hour.
Rollback
The routing engine supports shadow mode and instant return to manual routing.
Constraints
- The first cycle must not change live routing.
- Sensitive ticket text must remain in the approved environment.
- Any later live test must stop if urgent-case performance worsens from baseline.
Case steps
Work through each prompt using the evidence pack. Answers and rubric weights stay protected in the interactive flow.
Write a measurable aim, baseline and change theory for the first cycle.
Write a prediction and define outcome, process and balancing measures.
Design the smallest informative Do phase with data, ownership, deviations and safeguards.
Assume overall accuracy is 92% but urgent-case sensitivity misses the threshold. Study the result and choose the next action.