AI-Enabled Digital Risk Management and Configurable Enterprise Workflow Architecture in Cloud Computing

Abstract

Artificial intelligence (AI) can strengthen every stage of risk management. It can find unusual patterns in operational data, read thousands of incident reports, predict when risk indicators will breach their limits, and prioritize the cases that matter most. Its value depends on two conditions that are often overlooked. First, AI outputs must lead to governed action, which means they must feed workflows that route, escalate, and record decisions. Second, the AI models themselves are sources of risk and must be inventoried, validated, monitored, and audited. This article proposes an architecture for AI-enabled digital risk management that meets both conditions. AI risk services for anomaly detection, text classification, and prediction run as discoverable services in the cloud and feed configurable risk workflows. A metadata-centric model registry records every model and its governance status, and self-healing mechanisms keep the services available. A model risk governance layer, based on supervisory guidance on model risk and on internal algorithmic auditing, governs the AI models themselves. Using a design-oriented approach grounded in eleven studies published between 2011 and 2023, the article maps the evidence base and each AI technique to a stage of the risk lifecycle. It presents the architecture and a model governance lifecycle, and sets out controls for each lifecycle stage. It illustrates the approach with an airline managing operational risk from maintenance data and safety reports. It argues that AI improves risk management only when its outputs drive governed workflows and its models are managed as risks in their own right.