AI-DRIVEN PREDICTIVE ANALYTICS FRAMEWORK FOR ELECTRONIC FUNDS TRANSFER, LOAN ORIGINATION, AND AML COMPLIANCE IN DIGITAL BANKING

Abstract

This study addresses the problem that electronic funds transfer (EFT) monitoring, loan origination decisioning, and anti-money laundering (AML) compliance are often governed as separate control silos in digital banking, which limits risk visibility and reduces audit ready decision defensibility. The purpose was to validate an AI driven predictive analytics framework and quantify how Predictive Analytics Capability (PAC) influences EFT monitoring effectiveness, loan origination decision quality, AML monitoring effectiveness, and overall digital banking risk control performance (DBRCP). A quantitative cross sectional, case-based survey was administered across a cloud enabled digital banking environment, yielding 268 responses from EFT operations (31.7%), lending or underwriting (27.6%), AML or compliance (24.3%), and risk, analytics, or IT (16.4%). PAC (20 items) operationalized capability maturity across data integration, data quality, model development and validation, model governance and documentation, and user competence; outcome constructs were measured as Likert 1 to 5 composites. The analysis plan combined descriptive profiling, internal consistency testing, Pearson correlations, and hypothesis driven regression models. Reliability was adequate (Cronbach’s alpha: PAC 0.91, EFT_EFF 0.88, LOAN_QUAL 0.90, AML_EFF 0.89, DBRCP 0.92). Descriptively, respondents rated PAC at M = 3.84 (SD = 0.56), with governance and documentation the lowest dimension (M = 3.68), while EFT_EFF (M = 3.79), LOAN_QUAL (M = 3.73), AML_EFF (M = 3.76), and DBRCP (M = 3.76) were all above the scale midpoint. PAC correlated positively and significantly with EFT_EFF (r = 0.56), LOAN_QUAL (r = 0.52), AML_EFF (r = 0.59), and DBRCP (r = 0.63) at p < .001. Regression results showed that PAC predicted EFT_EFF (beta = 0.48, R2 = 0.31), LOAN_QUAL (beta = 0.44, R2 = 0.27), and AML_EFF (beta = 0.51, R2 = 0.35), all p < .001, indicating the strongest capability to outcome contribution in AML. In the integrated model, EFT_EFF (beta = 0.26), LOAN_QUAL (beta = 0.21), and AML_EFF (beta = 0.37) jointly explained DBRCP (R2 = 0.58), underscoring that coordinated improvements across payments, credit, and compliance drive risk control. Implications are that banks should invest in PAC foundations, particularly governance and documentation, to translate predictive models into consistent operational decisions and demonstrable compliance outcomes.