Patterns
Semantic Drift Patterns
Five recurring patterns of semantic drift, each with a description, the risk it creates, an example, and the governance response that resolves it.
Term Overload
A single surface term is reused across unrelated contexts without distinguishing the meanings. Each user assumes their interpretation is the shared one.
Risk
Reports, integrations, and AI outputs silently combine incompatible meanings. The cost is invisible until a downstream decision is made on wrong data.
Example
The word "Coverage" used identically by Claims, Reinsurance, and Product, with no context label, in the same data warehouse field.
Governance response
Resolve to distinct Canonical Concepts inside one Meaning Space using context labels (Coverage (Claims Limit), Coverage (Reinsurance Cession)). Term Overload is not a contradiction — it is unmanaged ambiguity.
Departmental Divergence
Two departments hold genuinely incompatible definitions of the same governed term, with no shared canonical version and no mechanism to surface the conflict.
Risk
Cross-department reporting reconciles silently and incorrectly. Executives are unaware that two reports answering the same question use different definitions.
Example
Finance defines "Customer" as a billed party; Marketing defines it as any party in the CRM. Both publish customer counts.
Governance response
Treat as a same-context contradiction. Route to governance review, name an authority, decide whether one definition replaces the other or whether the two should be split into separate Canonical Concepts.
Definition Evolution
A definition is refined or clarified over time within the same context, without contradicting prior meaning. The new wording strengthens, not replaces, the canonical concept.
Risk
Treated as drift, evolution causes false alarms and review fatigue. Treated as untracked, it loses the audit trail and history.
Example
A claims handler rewrites "Coverage (Claims Limit)" to add clarification about per-event vs aggregate limits, while preserving the core meaning.
Governance response
Record as a "definition_evolution" event rather than drift. Confidence increases, prior wording remains in provenance, no governance review required.
Regulatory Drift
A definition diverges because external regulation changes its required meaning, but internal governance has not yet updated the canonical concept.
Risk
Compliance reporting uses the obsolete meaning. Audit findings, fines, or restatements follow.
Example
A new EU directive narrows the scope of what counts as a reportable incident; the internal canonical definition still uses the prior broader scope.
Governance response
Open a governance decision referencing the regulatory change as source, version the canonical concept, deprecate the prior version with effective dates.
AI Interpretation Drift
AI agents resolve a business term to a meaning learned from training data rather than to the organization's canonical concept. The drift originates in the model, not the people.
Risk
AI answers are confidently wrong in domain-specific ways. The error is hard to detect because the surface vocabulary matches.
Example
An enterprise assistant interprets "policyholder" using a generic insurance meaning, when the organization's canonical definition narrows it to natural persons only.
Governance response
Require AI agents to resolve terms through the canonical registry and cite the resolved Canonical Concept. Reject answers that bypass resolution for governed terms.