This public review is in beta calibration. All results represent self-reported diagnostic indicators, not certified or verified audit findings. Missing or unknown evidence triggers automatic deferral under the non-compensatory rule.
Decision Quality Review v2
Has your AI system earned the authority you are giving it? This assessment tests evidence readiness across six independent gates for your specific lifecycle stage.
Currently exploring the pre-loaded illustrative example (Customer Service Copilot).
Inspect How Non-Compensatory Gates Evaluate Consequential Systems
Commercial Credit Underwriting Engine
High statistical calibration (98.4% AUC) and legal compliance, but lacks deterministic manual override for credit scores under macroeconomic shock telemetry.
Clinical Radiology & Triage Copilot
Clinically robust lesion identification, but medical software liability boundaries and statutory SaMD CE-mark/FDA clearances remain uncertified.
Autonomous Candidate Screener & Ranker
High ATS filtering speed, but fails to cryptographically log why candidates were rejected, exposing the board to EU AI Act Annex III statutory liability.
Autonomous Procurement Routing Dispatcher
Highly profitable automated PO routing, but critical cloud LLM provider has no contractual SLA failover or data portability exit protocol.
Step 1: Frame the System & Requested Gate
Evidence burdens are gate-sensitive. A release gate demands stricter runtime and fallback evidence than early discovery.