ServiceNow’s internal L1 case study: denominator before headline

ServiceNow’s own customer story, accessed October 26, 2026 (page does not visibly date the results): Rewriting the IT playbook. The IT team reports an internal L1 deployment: 15% of L1 incidents qualified for its bounded autonomous route, and the AI specialist resolved 46% of eligible incidents. If these rates share one reporting cohort and all eligible incidents were attempted, approximately 6.9% of the full L1 incident intake was resolved by AI. This is arithmetic on reported shares, not a claim that only 6.9% of all employee requests were handled. The same story’s '99% faster' is a speed claim, not a 99% success rate.

Competing figures are not time-series evidence. A September 29 vendor announcement claims Now Assist 'handles nearly 90% of ServiceNow employee IT requests' and 'resolves 99% of those interactions.' 'Handled requests' in that announcement and 'L1 incidents eligible for the AI Specialist' in the undated case study may differ by workflow, stage, meaning of resolution and observation period. There is no public reconciliation. Avoid joining 90% with 46%, or claiming improvements from one to the other. An earlier September 15 executive blog likewise discusses 90% and 99% but gives no cohort definitions.

Missing link. The case study names virtual-assistant routing, step-by-step help and selective autonomous escalation, and quotes team leaders; it supplies no original request IDs, requester confirmation, reopen rate, false closure count, active human support time or before-and-after quality comparison. The implementer playbook recommends tracking reopens, coverage, intervention and real action versus suggestions, demonstrating a measurable gap between a dashboard 'resolved' and a durable employee outcome. See audit and internal closure question.