Investigate When The Data, Users And Narrative Disagree
The Cross-Examination Desk opens one conflict topic at a time for SM-PILOT-03 — Supporting evidence on the left, Opposing evidence on the right, applicable conditions and open questions in the middle. Reviewers can limit scope, request more testing, or mark the conflict Unresolved instead of forcing a side.
“Pilot reduced handling time”
- Product Telemetry: median handling time 26 min vs 38 min Prior pilot baseline
- Workflow Result: agents complete Tier-2 paths without engineering handoff more often
- Applies only to Tier-2 billing escalations in scope
- Does not apply during documented surge weeks
- Are agents skipping required steps to gain speed?
- Does ticket growth erase the handling-time gain?
- Support Issue: error-related tickets rose 12% in the same window
- Stakeholder Feedback: Product User reports skipping optional verification steps under surge
Unresolved — limit the conclusion to Tier-2 non-surge weeks and request one more test cycle before scoring impact.
“Support agents complete escalations without engineering help”
- 312 escalations sampled — 63% completed without an engineering ticket
- IT confirms routing rule matched Allowed Exceptions on every excluded case
- Config-change escalations should move into Allowed Exceptions
- Should config-change types be formally excluded?
- 2 of 41 sampled escalations still needed an engineering handoff for a configuration change
Upheld, with a caveat — config-change escalations should move to Allowed Exceptions.
“Works for every escalation type”
- Product User testimony: correct behavior on two of three sampled scenarios
- Claim scope must reference Allowed Exceptions
- Any remaining undocumented exception types?
- Allowed Exceptions lists 3 escalation types explicitly out of scope
- Security flagged multi-tenant billing disputes as a hard exclusion
Rejected — replace with a scoped claim referencing the approved Allowed Exceptions list.
One conflict is Unresolved, one Upheld, one Rejected — reviewers are not forced to pick a side.
Saves the final interpretation, applicable conditions, reviewer identity, and score impact for each conflict topic.