AI-Driven Credit Scoring and Default Probability Modeling for Basel III Risk-Weighted Asset Optimization in Banking

Abstract

This study addresses a problem: AI-driven credit scoring can strengthen probability of default (PD) estimation, but uneven data quality, governance controls, and explainability reduce model trust and constrain Basel III risk-weighted asset (RWA) optimization. The purpose was to quantify the pathway from AI capability to PD modeling effectiveness and from PD effectiveness to Basel III RWA optimization in a cloud-enabled enterprise case bank. A quantitative cross-sectional, case-based survey was administered; 300 questionnaires were distributed and 212 valid responses were analyzed (70.7% effective response rate). Key predictors were AI-driven credit scoring capability, data quality readiness, model governance maturity, and explainability readiness; PD modeling effectiveness served as the mechanism, and Basel III RWA optimization effectiveness was the outcome. Reliability was strong across constructs (Cronbach’s α = 0.84–0.90). The analysis plan combined statistics, Pearson correlations, and multiple regression with moderation testing. Mean ratings were positive (AI capability M = 3.82; PD effectiveness M = 3.91; RWA optimization M = 3.76 on a 1–5 scale). Correlations supported the framework, including an association between PD effectiveness and RWA optimization (r = 0.66, p < .001). In regression Model 1, PD effectiveness was predicted by AI capability (β = 0.34, p < .001), data quality (β = 0.21, p = .002), governance maturity (β = 0.25, p < .001), and explainability readiness (β = 0.14, p = .018), with R² = 0.57. In Model 2, PD effectiveness predicted RWA optimization (β = 0.51, p < .001) and governance retained a direct effect (β = 0.19, p = .004; R² = 0.49). Moderation results showed a significant AI × governance interaction (β = 0.11, p = .031; ΔR² = 0.02), indicating that stronger governance amplifies the benefits of AI capability for PD outcomes. Implications are that banks seeking Basel III capital efficiency via AI should invest not only in scoring capability, but also in data readiness, governance discipline, and explainability practices so that PD gains translate into defensible RWA optimization.