Are Marketed AI Resolution Rates Real?
Marketed AI resolution rates are frequently inflated, because many vendors blend deflection into the resolution number and measure on favorable workloads. One widely cited example: Intercom Fin has been reported at a real production resolution rate of 45 to 53 percent against a marketed figure near 76 percent. Treat any headline resolution rate as a marketing number until you have measured it yourself on your own tickets with reopen rate attached.
How the number gets inflated
The gap between marketed and real rates comes from three moves: counting deflection (customer did not reach a human) as resolution (problem solved), measuring on cherry-picked or simple ticket sets, and ignoring reopens so premature closes still count. Each move flatters the number without solving more problems. A reported top failure mode is generic answers on workflow-specific tickets, which deflect but do not resolve.
How to verify a vendor claim
- Ask for verified resolution, not deflection, and confirm the definition.
- Ask for the reopen or 72 hour recontact rate next to it.
- Insist on a measurement against your own historical tickets, not a demo set.
- Segment CSAT to AI-handled tickets and check it against your human baseline.
How Resolve stays honest
Resolve reports auto-resolution and deflection separately, never merged, with reopen rate as the north-star honesty metric. The go-to-market motion is built on proving it: ingest a prospect help center and show "we would have resolved this many of your last hundred tickets, with the source cited for each." The claim is auditable ticket by ticket rather than a slide.