Knowledge Graph
A structured representation of an organization's domain — accounts, products, relationships, events — as nodes and edges that AI systems can reason over instead of retrieving raw documents.
Production AI is not a prompt. It is a system of context, tools, permissions, traces, evals, and feedback loops.
What it is
A knowledge graph models the business itself: which customer owns which account, which product belongs to which contract, which incident touched which service, which employee reports to which manager. It's structured (typed nodes and edges) rather than freeform (documents), which makes it queryable in ways unstructured retrieval cannot match.
Why it matters
Some questions can't be answered well by retrieving documents. 'Which of my customers are affected by the outage on service X?' is a graph traversal, not a search problem. A knowledge graph is the layer that makes those questions answerable directly.
How it works
Built from authoritative operational sources (CRM, billing, observability) using a schema that reflects the domain. Stored in a graph database (Neo4j, TigerGraph) or as a graph view over relational data. Queried by Cypher, GraphQL, or SQL with recursion. Agents can query the graph directly or use it as a retrieval source alongside documents.
How it differs from Source Graph
Source Graph models systems (where data lives, who owns it). Knowledge Graph models the business (what the data means). Both can coexist; Source Graph usually comes first because it doesn't require domain modeling to be useful.
Related resources
A navigable map of every system your data lives in — schemas, documents, code, tickets, events, owners, and permissions — so an AI agent can find the right source and respect the right access boundary.
How an AI agent finds the right document, chunk, or row to ground its answer in — and why the part that matters is the pipeline around the database, not the database itself.
The pattern where an AI agent retrieves relevant context from your data before generating an answer — instead of relying only on what the model learned during training.
The single, authoritative system or table where a piece of information is officially defined — the place every other copy of it must trace back to.