Whattheorganisationknows,withcitationsandlineage.

Knowledge answers questions about documents, standards, history and precedent. It is necessary and it is not sufficient — which is why it sits beside the ontology rather than instead of it.

Illustration of a governed answer assembled from approved sources, with each claim linked back to a cited document.

The boundary

Knowledge is what the organisation knows. Ontology is what the organisation is, how its parts relate, and what may happen next. Retrieval is how knowledge is found; the ontology is what gives the answer a place to belong.

Capabilities

  • Semantic retrieval

    Hybrid search over documents, records and prior decisions, combining embedding similarity with structured filters from the ontology.

  • Vector search

    Managed vector indexes with residency guarantees, versioned embeddings and re-indexing that does not require an outage.

  • Document intelligence

    Extraction from the formats enterprises actually hold: scanned drawings, inspection reports, contracts, correspondence and spreadsheets.

  • Citations

    Every generated claim resolves to a source passage a reviewer can open. Answers without a resolvable citation are surfaced as such.

  • Lineage

    Which source produced a fact, which version of it was current, and when it was last verified.

  • Memory

    Durable context scoped to an object or a case rather than to a chat session, so a decision made last quarter is still available to the next one.

  • Freshness policy

    Explicit rules about how stale a source may be for a given class of decision, and what happens when it is staler than that.

  • Access-aware retrieval

    Retrieval respects the caller's data scope. An agent cannot receive a passage from a document it is not permitted to read.

  • Redaction

    Sensitive spans can be removed before content reaches a model, with the redaction itself recorded in the evidence.

Access-aware retrieval is the control most often missing from otherwise well-built RAG systems. An index that ignores permissions is a permission bypass with a search box on it.

Whichlayeranswerswhichquestion

Which layer answers which question
QuestionAnswered by
What does the standard require in this situation?Knowledge — retrieve and cite the passage
Which assets does this constraint affect?Ontology — traverse relationships
Has this failure happened before, and what was done?Both — incidents from the ontology, reports from knowledge
Is this operator permitted to approve this?Control plane — policy against typed objects
What did we decide last time, and why?Evidence — the recorded decision and its reasoning

Questions

Do we need to move our documents?
Not necessarily. Content can be indexed in place where the source system supports it, which keeps the source of truth where its owners expect it to be.
How do you stop the model inventing citations?
Citations are resolved against the retrieved passages rather than generated as text. A claim whose citation does not resolve is flagged, not rendered as if it were sourced.
Is retrieval enough for regulated decisions?
Rarely on its own. Regulated decisions usually need the structured half: which objects are in scope, which rules apply, who owns the outcome. That is what the ontology and the control plane contribute.

Bring us one operational problem.

You do not need a finished brief. Bring the problem — we will work out the next step together.

Or book a call with the team