AI Reliability Console

Monitor behavior, manage risk, and ship confidently

Metrics

Metric Catalog

Define thresholds that clearly communicate pass/fail criteria.

Add Metric

Pass Rate

Percent of rows that pass overall eval criteria.

statisticalPresetv3Threshold: pass if >= 90
OpenEdit

Critical Failure %

Rate of critical safety or policy failures.

statisticalPresetv2Threshold: pass if <= 3
OpenEdit

Brand Tone Alignment

Rubric score for whether response matches brand voice.

rubricCustomv1Threshold: pass if >= Good
OpenEdit

Resolution Confidence

Confidence that answer fully resolves customer intent.

statisticalCustomv4Threshold: pass if >= 85
OpenEdit

Policy Accuracy

Rubric score for policy adherence in answers.

rubricPresetv5Threshold: pass if >= Good
OpenEdit