An operational ontology is a typed, machine-readable model of an organisation's world — its entities, relationships, policies, and classifications — that gives AI systems a shared vocabulary grounded in how the operation actually works. It is the load-bearing structure beneath reliable automation: rules, agents, and analytics all inherit its definitions.
The test of an operational ontology is bend, not break: when the domain changes — a new asset class, an amended regulation, a reorganisation — the ontology absorbs the change as data, and every rule and agent built on it inherits the update without a rewrite.
Ontologies also carry jurisdictional nuance that flat schemas cannot: the same building is classified differently under different codes; the same transaction is treated differently under different statutes. Encoding that variation once, structurally, is what makes multi-jurisdiction automation tractable.
This vocabulary is implemented, not aspirational — it describes how the MAIA substrate actually runs. See the platform or read the research.