Govern · Clean · Activate
The three disciplines that define a Document Knowledge Platform. Each maps to capabilities the platform must implement; together they form the operating loop that takes a document estate from passive accumulation to active, AI-consumable asset.
Govern
Govern is the discipline of treating documents as managed business products rather than files in folders. The platform maintains a Document Catalog — the inventory of the document estate — and each entry is a Document Product with an explicit Document Owner (the business direction accountable for its editorial quality), a Document Authority on the transverse side (the CDO team carrying the cross-domain standard and regulatory conformity), and Document Stewards animating the day-to-day discipline. The contract between producers and consumers is materialised as a Document Contract that commits on quality, freshness, SLA, and ACL. Every change is tracked by Document Lineage, so that any downstream answer — human query or AI agent call — can be traced back to its sources.
In K-AI today, the Document Catalog is materialised through the Document Owner view in the K-AI Audit web app, plus the metadata exposed by the Instance API Documents endpoints. A dedicated Catalog UI is on the roadmap.
Govern enforces Principle 1 (Document-as-a-Product) and Principle 4 (Federated governance): editorial ownership stays where the expertise lives — in the business directions — while the transverse standard is carried centrally by the Document Authority. Ownership is distributed, governance is federated.
Mechanisms:
Document Catalog
Document Product
Owner / Authority / Steward / Producer roles
Document Contract
Document Lineage
See: Roles model for the role definitions and Glossary for the Catalog, Product, Contract, and Lineage entries.
Clean
Clean is the discipline of continuously verifying that the documents in the catalog are internally consistent, up-to-date, and free of orphan or obsolete material. K-AI Audit operates this loop: it detects conflicts between documents covering the same subject, identifies divergent duplicates and obsolete passages, surfaces missing subjects (topics the business needs covered but the corpus does not address), and pushes mandatory questions back to the Document Producers for arbitration. Recommendations are tracked to closure — a detected conflict either gets resolved by a Producer's decision, escalated to the Owner, or explicitly ignored with rationale — and the catalog reflects the resolution.
Clean enforces Principle 2 (Clean before consume) and Principle 5 (Continuous observability): no consumer — human or AI agent — should consume an unaudited document, and the audit itself is a continuous flow rather than a one-shot preparation. Quality is not a state, it is a stream.
Mechanisms:
K-AI Audit
Conflict detection
Missing subjects
Mandatory questions
Document Observability
See: K-AI Audit for the operational surface and Audit MCP tools for the programmatic surface.
Activate
Activate is the discipline of exposing the cleaned, governed corpus to consumption. Humans reach it through Document Discovery — natural-language access for human Document Consumers. AI agents reach it through K-AI MCP, the platform's Model Context Protocol surface, which integrates directly with Claude Desktop, Cursor, Le Chat, and any other MCP-compliant client. ACLs from the source systems are mirrored at the exposure layer so that no consumer ever sees content they are not entitled to, and every query produces a traceable lineage — which documents answered, in which versions, against which ACL context — usable both for audit and for compliance.
Activate enforces Principle 6 (AI-Ready by design) and Principle 3 (Semantic over syntax): documents are exposed in a form directly consumable by AI agents, with the right protocols and the right guarantees, and the platform reasons on meaning (consistency, recency, contradictions) rather than on schema or file format.
Mechanisms:
K-AI MCP
REST API
Document Contract
ACL mirroring
Query-level lineage
See: K-AI MCP for the AI-agent surface and the Instance API for machine-to-machine integrations.
Six founding principles
Document-as-a-Product. A document is not a file in a folder; it is a product with an owner, an SLA, and a lifecycle. The Document Product abstraction transforms a passive document estate into a governed asset. The pattern is inherited from Data Mesh and transposed to the documentary world.
Clean before consume. No consumer — human or AI agent — should consume an unaudited document. The platform makes operational the rule "no consumption without audit". Cleaning is a continuous discipline, not a one-time preparation step, because a corpus clean today will not be clean in six months if regulation changes, internal decisions accumulate, or adjacent procedures evolve.
Semantic over syntax. Document quality is not measured against schema compliance, file format, or template conformity. It is measured in meaning: is this document consistent with the rest of the corpus, is it up-to-date, does it answer a real knowledge need? The DKP reasons semantically where the ECM reasons syntactically.
Federated governance. Documentary governance cannot be centralised in a single team — business contexts are too diverse, specialisms too deep. The DKP federates: each business domain (Document Owner = direction) operates its own Document Products with its own Producers, under a common standard carried by the Document Authority (typically the CDO team). Ownership is editorial and lives in the business; transverse governance is central. This is the documentary counterpart of federated computational governance from Data Mesh.
Continuous observability. Document quality is a flow, not a state. The platform monitors continuously: new conflicts detected, emerging subjects not yet covered, documents that have become orphans, user queries left without a satisfactory answer. Observability is what allows the cleaning loop to remain meaningful over time.
AI-Ready by design. The platform exposes documents directly consumable by AI agents — with the right protocols (MCP), the right guarantees (ACL preserved at every step, lineage traced per query, freshness known), and the right semantic layer underneath. AI-Readiness is not a feature added on top: it structures the architecture from the ingestion layer up.
Next steps
5-layer architecture — how the three disciplines map onto the conceptual layers.
K-AI Audit — the operational surface for the Clean discipline.
K-AI MCP — the AI-agent surface for the Activate discipline.
Last updated