Reference

Containment and AI resolution

The share of contacts a self-service or AI channel finishes without an agent. Vendors and centers define it in several ways.

Also called: containment, deflection, self-service resolution.

Why it matters

It decides what an AI or self-service investment saves. A conversation that ends without an agent is not always a resolved issue.

How it is calculated

Three rates, three denominators
Coverage = eligible demand ÷ total demand. Apparent resolution = resolved ÷ AI-involved conversations. Net automation = durable resolutions ÷ total demandThey are never shown as one another. The gap between apparent resolution and net automation is where most overstatement starts.As published in the AI Deflection Reality Check method, version 1.2.
Durable resolution
Apparent resolution × (1 − repeat and false resolution rate)A resolution that comes back was a deferral.As published in the AI Deflection Reality Check method, version 1.2.

Before you compare

  • Check the denominator: all contacts, contacts the AI could handle, or conversations the AI took part in. The same system can report three rates.
  • A customer who gives up, or calls back later, can still count as contained.
  • Vendor-reported rates use the vendor's own definition.

Published figures

What publishers report for this measure. Each counts something slightly different, so read the definition before you compare; nothing here is averaged across publishers.

Self-service and AI resolutionSurvey of people

Three different questions asked of different people; none is a containment rate. Vendor-defined resolution rates are left out.

  • 14%Issues fully resolved in self-serviceCustomers' own account of their last issue.Gartner, 19 Aug 2024. Gartner Survey Finds Only 14% of Customer Service Issues Are Fully Resolved in Self-Service (survey of 5,728 customers, December 2023).
  • 18%Web chats handled without a human agent, USBenchmark studyReported by centers; 6% in 2020.ContactBabel, 2026. The 2026 US Contact Center Decision-Makers' Guide, key findings.
  • 30%Service cases handled by AIService professionals' own estimate.Salesforce, 13 Nov 2025. State of Service, 2025 (survey of 6,500 service professionals in 39 countries, April to June 2025).
Channel mix and preferenceBenchmark study

What centers handle and what customers say they prefer are different measures; they are shown apart.

  • 62%Live telephony, share of inbound, end of 2024Email about 19%, web chat 8%, telephony self-service 9%.ContactBabel, read on Contact Center Pipeline, 14 Aug 2025. The 2025 US Contact Center Decision-Makers' Guide, as reported in "The Great Contact Center Standoff".
  • 35%Phone as the preferred channel, US adultsSurvey of peopleEmail 23%, live chat 10%, in person 8%, chatbot 1%.YouGov, 13 Mar 2025. How Americans prefer to contact businesses for customer service.

Measure yours