Beyond trust: An evidence-based benchmark for AI decision readiness in high-stakes operational environments

The deployment of artificial intelligence (AI) in high-stakes decision contexts—including cybersecurity triage, insurance underwriting, clinical decision support, and university policy formulation—has outpaced the methodologies organizations use to judge whether a given system is actually ready to support those decisions. This paper introduces the AI Decision Readiness Benchmark (AI-DRB), an evidence-based framework that quantifies readiness across four automated dimensions: risk score validation, reliability, robustness, and explainability and auditability. The dimensions are combined into a single Decision Readiness Score (DRS) on a [0, 1] scale with explicit deployment thresholds. We evaluate AI-DRB on five public-source datasets covering 35,417 instances and on five state-of-the-art language models plus a fine-tuned BERT classifier. We then apply the framework retrospectively to publicly documented AI governance decisions at 142 United States universities to expose gaps between deployment prevalence and assessment rigor, and we show how the resulting score sheet can support governance decisions at the University of Denver. Findings advance practice from aspirational "trust" language toward measurable, auditable readiness criteria.