KPI

Automation rate: AI benchmark

Share of cases, documents or transactions processed end to end without human touch.

How to measure it

Cases completed without any manual step divided by all cases entering the process.

Across the library

Median 76% across 15 deployments, reported range 25% to 99%, plus 2 "up to" values not pooled.

Higher is better. Unit: percent.

Reported values by use case

AI agent for travel and expense report audit

AI for claims triage and straight through processing

AI for payment investigations and exceptions

  • BNYat least 10% · organization claim

Use cases that should track automation rate