What is an Intent Graph?
An Intent Graph is an event-sourced intent fact store that carries structured customer intent from first question to Sales handoff. It stores typed facts (environment: "hospital", budget: "€50k") — not text chunks — enabling every touchpoint to read and write shared context. Unlike RAG systems that retrieve documents by similarity, the Intent Graph stores explicit facts queryable by criteria. Customers never repeat themselves because facts accumulate across touchpoints over time.
Why This Matters
The query has changed. "Air filter hospital" became "Which filter handles high humidity in a hospital AHU with clogging issues?" Parse that question: hospital is an environment, AHU is an asset, clogging is a problem, high humidity is a constraint. Four facts that would take a configurator five screens to extract — delivered in one sentence to an AI assistant.
ChatGPT has the full specification. Your landing page has an anonymous session ID. The gap between those two states is what the Intent Graph solves.
Key Architectural Principles
- Structured over semantic: "Hospital" is an environment type, not a text chunk to retrieve
- Minimal schemas: Six fields that matter beat sixty fields that might
- Deterministic where possible: A product click is a product click — no interpretation needed
- Thin shared layer: Two API calls to integrate, not a platform migration
Technically, the Intent Graph is an event-sourced intent fact store with timeline projection — think "fact ledger with projection API," not Neo4j. We use "graph" to describe the connected nature of facts across touchpoints and time.