Metadata-Centric Enterprise Architecture for AI-Enabled Risk Management and Cloud-Based Workflow Automation

Abstract

AI-enabled risk management spreads its logic across many places: risk taxonomies in spreadsheets, data definitions in databases, model assumptions in project documents, workflow rules in application code, and infrastructure settings in deployment scripts. When these drift apart, no one can say with confidence which data fed a risk score, which model version produced it, or which rule routed the case. This article proposes a metadata-centric enterprise architecture in which that logic is held as governed, versioned metadata in one repository and read at run time by every layer. Business metadata defines risk categories and appetite. Data metadata records sources, definitions, and lineage. Model metadata documents each AI model's purpose, data, and limits. Process metadata defines configurable workflows. Operational metadata defines service priorities and recovery rules. Transactions flow through an event streaming platform to AI risk services and a configurable workflow engine, while an intelligent scheduler and a self-healing layer use operational metadata to keep the platform efficient and available. Using a design-oriented approach grounded in eleven studies published between 2013 and 2023, the article maps the evidence base, defines five metadata domains, sets out a governed metadata change process, and shows how lineage answers audit questions. It illustrates the design with an insurer managing claims fraud and underwriting compliance risk. It argues that metadata, not code, should be the control surface of AI-enabled risk management.