Abstract
Financial AI systems can retrieve records and generate forecasts, yet still fail to understand whether two identifiers describe the same instrument, which event changed an exposure, or which policy constrains a decision. This technical paper presents an ontology-driven architecture that places machine-readable business meaning between data platforms and AI.
The semantic control plane
The proposed control plane combines an enterprise financial ontology, modular domain ontologies, a temporal knowledge graph and executable validation. Source records retain lineage; entity resolution establishes durable identity; mappings align local fields to governed concepts; SHACL validates required relationships and values; and graph APIs provide context to analytics and agents.
Financial alignment
The enterprise model reuses FIBO for financial concepts and aligns transaction-lifecycle semantics with the FINOS Common Domain Model where appropriate. PROV-O represents evidence and lineage, OWL expresses formal semantics, and SHACL converts governance requirements into executable constraints.
AI activation
GraphRAG can retrieve connected evidence rather than isolated chunks. Agents can inspect the instrument, issuer, event, portfolio, risk and control path before producing an answer or action. The ontology does not replace models; it gives them shared identity, rules, context and explainability.
Conclusion
An AI-ready data platform is not only a platform with more features. It is a platform whose data carries explicit, reusable and governed meaning.