The success rate of an AI support agent is a number the seller defines. One of those sellers said so in writing, about the rest of the market, in a guide published June 17, 2026.
The guide separates three numbers that get used as if they were one. Deflection: the conversation never reached a person, even if the customer gave up. Containment: the AI kept the conversation, even if the answer was vague and the customer just left. Resolution: the problem was solved and the customer did not come back with the same issue. The guide says the definition of "resolved" can move a reported rate by 20 to 30 percentage points.
In the SaaS support queues we see, that is the difference between a dashboard and a customer. A release goes out, the same question comes back, and a deflection number counts both visits as wins.
Gartner asked on X on September 15, 2026 whether GenAI investments match what customers expect, pointing to more than 3,500 customers who described how AI is changing service, trust and the way they prefer to interact.
Whoever writes the definition of solved owns the number. Today that is the seller, and the customer is the only one who knows whether it was.
Sources
- Fin (formerly Intercom), AI Resolution Rate: What It Is, 2026 Benchmarks, and How to Improve It, June 17, 2026. https://fin.ai/learn/ai-resolution-rate
- Gartner on X, September 15, 2026. https://x.com/Gartner_inc/status/2099857047579656539