Confidence threshold
A policy value used to decide whether an AI result can proceed automatically or should be rejected, re-routed, or sent for human review.
Why it matters
Thresholds are useful only when the confidence signal itself has been evaluated against real operating data.