Artificial Intelligence and Operational Risk Management: Current Applications, Challenges, and Research Gaps

Abstract

Operational risk — exposure to loss from failed internal processes, systems, people, or external events — has become one
of the most active application areas for artificial intelligence (AI) in enterprise and financial risk management. This review
examines current AI applications, persistent implementation challenges, and open research gaps in operational risk management.
Industry survey evidence indicates that four in five banks now deploy AI for operational-risk management, with cyber risk the
fastest-growing use case, even as governance concerns and unresolved questions about return on investment temper further
deployment (Risk.net, 2026). Academic and applied literature documents AI’s core contributions through anomaly detection
and predictive analytics (Leo et al., 2019; Heß & Damásio, 2025), AI for IT operations (AIOps) for automated root-cause
analysis and log-anomaly detection (Notaro et al., 2021), and enterprise-architecture approaches that embed operational-risk
controls directly into configurable workflows (Basireddy, 2022b) and metadata-driven platforms (Basireddy, 2022a). The
review further considers real-time data-processing requirements for operational-risk detection (Sannidhanam, 2021; Event
Streaming Architectures for High-Volume Transaction Processing, 2022) and the emerging role of autonomous AI agents in
continuous infrastructure risk monitoring (Sannidhanam, 2025). Despite substantial adoption, the review identifies persistent
challenges — data quality, model interpretability, and a widening gap between AI deployment speed and organizational
governance capacity (Risk.net, 2026) — and research gaps including limited empirical evaluation of AIOps effectiveness in
production environments, underdeveloped frameworks for governing autonomous risk-remediation agents, and insufficient
integration between operational-risk AI models and broader enterprise risk management architecture.