Canonical concept · WikiSure
Semantic Governance
The discipline of approving, versioning, and propagating canonical business definitions so every human, system, and AI agent resolves a term to the same meaning.
Definition
Semantic Governance is the discipline of producing, approving, versioning, and propagating canonical business definitions so that every human, every system, and every AI agent in an organization resolves a given term to the same meaning. It treats meaning as a governed asset — owned, versioned, and accountable — rather than a tacit convention that lives in slide decks and tribal knowledge.
Business context
Enterprises increasingly run on autonomous and semi-autonomous AI agents. Agents act on definitions, not on intentions: an underwriting agent that interprets 'coverage' differently from a claims agent will produce inconsistent decisions at scale. Semantic Governance closes that gap by giving every term one canonical, machine-resolvable meaning — and a governance trail that explains why.
Insurance example
A multinational insurer carries 14 different internal definitions of 'material damage' across product lines, regions, and legacy claim systems. Without semantic governance, an AI-assisted claims triage model trained on one definition silently mis-classifies cases that arrive under another. WikiSure enforces a single canonical 'Material Damage' definition per namespace, with named owners and a versioned audit trail.
Governance example
When a regulator updates the definition of 'eligible counterparty,' the change must propagate atomically across compliance dashboards, customer onboarding flows, and downstream LLM agents. Semantic Governance treats that propagation as a first-class event (CanonicalApproved → DependenciesNotified → AgentsResynced) instead of a quarterly cleanup project.
Related concepts
Frequently asked
- How is semantic governance different from data governance?
- Data governance owns the data — schemas, lineage, access. Semantic governance owns the meaning — the business definitions that data, applications, and AI agents resolve against. Both are required; neither replaces the other.
- Why do AI agents need semantic governance?
- Agents act on the definitions they are given. Two agents reading two definitions of the same term will produce contradictory decisions. A governed canonical definition is the agent-readable contract that prevents that contradiction.
- Who owns a canonical definition?
- A single named owner per definition is mandatory in WikiSure. Shared or 'committee' ownership is the failure mode semantic governance exists to prevent.