A legal agent may retrieve both an original contract and a later amendment. Both can be authentic, yet only one may govern an obligation during the disputed period. A knowledge graph preserves the relationships and conditions needed to interpret such information, rather than relying on document proximity or similarity alone. section 5.4 section 5.4.1
An enterprise knowledge graph applies a semantic model to reusable, governed assertions. It can represent that a provision creates an obligation, that a judgment establishes a legal proposition, or that a type of evidence can support a factual proposition. An agent can retrieve, traverse, qualify, and cite these relationships. section 5.4.1
Keep the knowledge graph distinct from related architecture:
- An ontology defines the domain’s concepts, relationships, and, where formalized, constraints and axioms. The knowledge graph applies that semantic model to governed knowledge. section 5.4.1
- An entity-relationship model structures records for storage or transactions. By itself, it does not express legal effect, precedential weight, or the conditions under which a relationship is valid. section 5.4.1
- Retrieval supplies candidate information; the knowledge graph expresses reusable meaning and relationships. A context graph records how knowledge, evidence, policy, tools, exclusions, and judgment came together in a particular decision. section 5.4 section 5.7.3
An amended obligation
An amendment changes a supplier’s volume obligations. Updating the source document alone is not enough: update its index status, relationships, supersession metadata, retrieval priority, and any memory or context that still reflects the earlier clause. A close wording match must not make the original clause appear to govern when the amendment applies. section 5.1.1 section 5.7.3
Constrain assertions by their applicable domain, jurisdiction, matter type, source class, effective period, purpose, and agent authority. Distinguish governed relationships from extracted, inferred, disputed, experimental, and superseded ones. Assign ownership and control approval, versioning, lineage, and retirement as the graph changes. section 5.4.1 section 5.1.2
Table 5.3 Pattern card: Enterprise Knowledge Graph
|
Attribute
|
Definition and requirements
|
|
Mandate
|
Preserve the governed concepts and relationships through
which agents interpret enterprise information. In law, this includes
instruments, provisions, obligations, conditions, claims, factual
propositions, evidence, and judicial authorities.
|
|
Scope
|
Apply it to reusable meaning that can affect retrieval,
applicability, evidential sufficiency, explanation, routing, or agent
authority.
|
|
Operating envelope
|
Constrain assertions by domain, jurisdiction, matter type,
source class, effective period, purpose, and agent authority. Distinguish
governed, extracted, inferred, disputed, experimental, and superseded
relationships.
|
|
Boundary of authority
|
It defines enterprise meaning. It doesn't establish live
facts, determine a claim, authorize action, or replace policy enforcement and
professional judgment.
|
|
Failure mode
|
Restrict traversal, downgrade the result, return to
authoritative sources, request validation, or reduce agent when you can't
establish provenance or applicability.
|
|
Evidence & auditability
|
Preserve ontology and graph versions, source and
relationship provenance, approval and assertion status, temporal and
jurisdictional conditions, material traversals, and the decision episode
influenced.
|
Use the pattern card to check the graph’s scope, operating envelope, and boundary of authority. section 5.4.1
The graph defines enterprise meaning; it does not establish live facts, decide a claim, authorize action, or replace policy enforcement and professional judgment. When the governing source cannot be established, qualify the position, seek further evidence, or reduce the agent’s authority. section 5.4.1 section 5.7.3














