Deflection rate counts the share of calls the agent handles end-to-end. A high number sounds impressive, but on its own it tells you nothing about quality. A bot that deflects every call by saying "we cannot help, goodbye" has a perfect deflection rate.
The metric to pair with deflection is resolution rate, the share of calls where the customer's actual issue was solved. A useful vendor reports both, and ideally a CSAT score against the deflected slice so you can see whether the people who got auto-resolved actually liked the experience.
The honest deflection number for an SME use case (booking, rescheduling, basic Q&A) is around eighty percent. Anyone claiming more is usually counting "the agent said something" as a deflection.
Vorel reports both deflection and resolution rate, and surfaces every deflected call in your CRM so you can audit the slice.