A living graph, not a folder
Most organizations store knowledge as documents: files in folders, records in rows, messages in threads. The information is captured, but the meaning— how things relate, why they matter, what depends on what — lives only in people's heads. When people leave, that meaning leaves with them.
The Knowledge Architecture models institutional knowledge as a living graph of durable objects and typed relationships. Documents don't disappear; they become evidence attached to objects rather than the system of record itself. The result is a structure a machine can reason over and a human can trust.
Objects & relationships
A Knowledge Object is a durable, versioned unit — an asset, a supplier, a standard, a decision. It persists independently of any single document and accumulates history over time. Objects are connected by typed relationships (depends-on, supersedes, owned-by, derived-from) that carry explicit meaning, turning isolated facts into a navigable graph.
Because relationships are explicit, the system can answer questions that no document search can: what would be affected if this supplier failed, which decisions relied on a now-retracted assumption, or how a standard propagated across programs.
Ontology & knowledge fabric
A shared ontology normalizes every source so the same entity referenced in a contract, a filing, and a sensor stream resolves to one object. The knowledge fabric then unifies those objects across systems of record into a single queryable model — without copying them into yet another silo.
Evidence & confidence
Every assertion carries its provenance: the source, the timestamp, and the derivation path. Every derivation carries a confidence signal, letting the system distinguish established fact from working hypothesis. This is what makes conclusions defensible — any answer can be traced back to the evidence that produced it.
Core concepts
The vocabulary that recurs throughout the Knowledge Architecture:
- Knowledge Objects
- Durable, versioned units of institutional knowledge — an asset, a supplier, a decision, a standard — that persist independently of any document that describes them.
- Typed Relationships
- Explicit, meaningful edges (depends-on, supersedes, owned-by, derived-from) that turn isolated facts into a graph a machine can reason over.
- Ontology
- The shared vocabulary that normalizes every source, so the same entity referenced in a contract, a filing, and a sensor stream resolves to one object.
- Knowledge Fabric
- The connective layer that unifies knowledge objects across systems of record into a single queryable model without copying them into a new silo.
- Evidence
- Every assertion carries its provenance — the source, timestamp, and derivation path — so conclusions remain auditable and defensible.
- Institutional Memory
- The accumulated, retained knowledge of the organization that survives reorganizations, turnover, and system migrations.
- Decision Graph
- A structured record of decisions and their rationale, linking options considered, evidence weighed, and outcomes observed.
- Identity Graph
- A resolved map of people, roles, teams, and agents that grounds access, accountability, and attribution across the model.
- Confidence
- A first-class signal on assertions and derivations, letting the system distinguish established fact from working hypothesis.