AI-Enabled Financial Risk Management: A Systematic Review of Credit, Market, Liquidity, and Operational Risk

Abstract

Financial risk management is conventionally organized around four principal risk types — credit, market, liquidity, and operational — each historically governed by distinct measurement techniques and regulatory regimes. This systematic review examines how artificial intelligence (AI) and machine learning (ML) methods are being applied across these four domains, synthesizing evidence from banking-sector systematic reviews (Heß & Damásio, 2025; Leo et al., 2019) alongside applied enterprise-architecture and data-infrastructure research addressing how AI-driven risk systems are technically operationalized (Basireddy, 2022a, 2022b; Sannidhanam, 2021, 2025; Event Streaming Architectures for High-Volume Transaction Processing, 2022). Evidence indicates that machine learning coverage is most mature for credit and market risk, with artificial neural networks and ensemble methods showing strong predictive performance, while liquidity risk and certain operational-risk sub-domains remain comparatively underexplored in the academic literature (Heß & Damásio, 2025; Leo et al., 2019). The review further finds that AI-enabled financial risk management increasingly depends on real-time data infrastructure — event-driven and event-streaming architectures capable of processing high-volume transaction data with low latency — as a technical precondition for timely risk detection across all four risk types, and on configurable, metadata-driven enterprise platforms for translating model outputs into operational risk controls. The review concludes by identifying persistent gaps, including underdeveloped liquidity-risk applications, interpretability constraints on regulatory adoption, and the architectural requirements for scaling AI-based financial risk management from pilot models to production-grade, auditable systems.