> For the complete documentation index, see [llms.txt](https://k-ai.gitbook.io/knowledge-ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://k-ai.gitbook.io/knowledge-ai/reference/glossary.md).

# Glossary

Canonical DKP terminology. Each entry pairs the K-AI term with its closest equivalent in the structured-data world, where applicable.

### Document Repository

*Data world equivalent: Data Warehouse / Data Lake*

**Definition.** The complete document estate of an organisation, considered as a governed asset rather than mere storage. A Document Repository aggregates — physically or logically — the documents held in heterogeneous source systems (SharePoint, Confluence, Notion, Google Drive, Box, network shares, legacy ECM).

**Example.** *An industrial group's Document Repository: 1.2M documents spread across 47 SharePoint instances, 8 Confluence spaces and 12 NAS shares, addressed as a single estate.*

### Document Knowledge Base

*Data world equivalent: Data Lakehouse*

**Definition.** The combination of a Document Repository and an active semantic layer (Semantic Document Layer) that makes documents queryable and auditable by AI. It is the unit of work of the DKP: operations no longer apply to files, they apply to a knowledge base.

**Example.** *The "HSE Procedures Europe" Knowledge Base: 18,000 cleaned and semantically indexed documents, exposed to an AI agent through MCP with source ACLs preserved.*

### Document Product

*Data world equivalent: Data Product*

**Definition.** A coherent set of documents packaged as a consumable product, with a Document Owner (business direction), an explicit business scope, a quality SLA (consistency, freshness, completeness) and a defined lifecycle. The Document-as-a-Product principle is one of the six founding tenets of the DKP.

**Example.** *Document Product "Auto Claims Reference": \~3,200 documents, Owner = Auto Claims Director, 30-day freshness SLA, 95% non-contradiction quality contract, Steward = Knowledge Manager of the Claims division.*

### Document Mart

*Data world equivalent: Data Mart*

**Definition.** A subset of a Document Repository dedicated to a domain or a specific use case. A Mart is a useful, homogeneous slice — by function, by site, by product line, by audience.

**Example.** *Document Mart "Onboarding for new Logistics joiners — France": a curated extract of the HR Repository combined with the Logistics business reference Products.*

### Document Catalog

*Data world equivalent: Data Catalog*

**Definition.** Living inventory of an organisation's Document Products. For each Product: business description, Owner (direction), freshness, quality level, ACLs, lineage, dependencies. The Catalog is the entry point of document governance — without a Catalog, no DKP.

**Important for large groups.** Many customers have already invested in a Data Catalog (Collibra, Atlan, Alation, Informatica) for their structured data. The DKP is not intended to replace it. Two scenarios:

* **Customer with no Data Catalog in place** — the DKP provides a standalone Document Catalog.
* **Customer with a Data Catalog already deployed** — the DKP publishes Document Products as assets in the existing Catalog through the vendor's native APIs. K-AI acts as an enricher, not a competitor. This is the most common scenario in groups above 5,000 employees.

**Example.** *A bank deployed Collibra four years ago for its 2,800 structured datasets. With K-AI, it progressively publishes its Document Products as Collibra assets: "KYC Reference" becomes an asset with its Owner, SLA and lineage. The CDO sees the full structured + unstructured estate in Collibra.*

### Document Lineage

*Data world equivalent: Data Lineage*

**Definition.** Traceability of the origin, transformations, sources and versions of a document. Lineage answers: where does this information come from? What is its reference version? Who modified it? Which Document Products depend on it? Critical for AI Act compliance (proof of provenance of the data feeding an AI system).

**Example.** *Lineage of an HSE procedure: created in 2019, derived from the ISO 45001 guide, modified four times, last reviewed by the SME in March 2026, consumed by 3 Document Products and 2 AI agents.*

### Document Quality

*Data world equivalent: Data Quality*

**Definition.** Measurement of a Document Product's quality along semantic dimensions: internal consistency, non-contradiction with the other documents of the scope, non-obsolescence, completeness (absence of missing subjects), semantic deduplication. Document Quality is the unit of work of the K-AI Audit module.

**Example.** *Quality score 78/100 on the Product "Line X operator manual": 14 conflicts detected, 6 missing subjects identified, 23 semantic duplicates to merge.*

### Document Governance

*Data world equivalent: Data Governance*

**Definition.** The framework of policies, roles, controls and processes applied to the document estate. Governance answers: who can do what on which Document Product? Which quality standards apply? How is compliance proven (GDPR, AI Act, sector-specific requirements)?

**Example.** *Governance of the Document Product "Banking procedures": Compliance Direction as Owner, dual validation on every change, quarterly audit, 10-year retention.*

### Document Owner

*Data world equivalent: Data Owner (Data Mesh sense, not Data Office)*

**Definition.** The business direction that owns a Document Product (HSE Direction, Compliance, HR, Engineering, Sales, etc.). The Owner stands behind the editorial quality of the content, sponsors the Producers of its scope, and arbitrates strategic conflicts. One Owner per Document Product.

**Why business-side and not CDO-side.** The CDO does not know the content: they cannot judge whether an HSE procedure is valid or obsolete, nor arbitrate a conflict between two legal references. Document ownership must stay where the business lives — this is the original Data Mesh principle (distributed ownership). The CDO carries the Document Authority (see next entry), not the ownership of contents.

**Example.** *The Document Owner of the Product "HSE Procedures Europe" is the HSE Director Europe. They validate the scope, sponsor the SME contributors (Producers), and arbitrate business-level decisions. A Document Steward (HSE Knowledge Manager) handles day-to-day animation on their behalf.*

### Document Authority

*Data world equivalent: Data Office / Federated Governance Team*

**Definition.** Cross-functional teams reporting to the CDO (or to the Chief Knowledge Officer when the role exists separately). The Authority carries the transverse documentary standard, regulatory compliance, platform tooling, and multi-domain federation. It owns no Document Product directly — it is an arbiter, guarantor and enabler, not an editor.

**Why distinguish Owner and Authority.** Without this distinction, editorial ownership (business) and transverse platform governance (CDO) get conflated. Conflation is the classic failure mode of Knowledge Management programmes driven solely by IT or data teams: zero adoption because the business never feels ownership.

**Seven concrete responsibilities of the CDO as Document Authority.** Moving Owner to the business does not weaken the CDO's role — it clarifies and reinforces it:

1. **Definition of the transverse standard.** The CDO sets the rules that apply to every Document Product across the organisation: minimum quality thresholds, freshness policies, Document Contract formats, shared taxonomies, common vocabulary. They own the grammar, not the text.
2. **Transverse regulatory compliance.** AI Act, GDPR, sector requirements (Solvency II, MiFID, ICH, HSE). The CDO is typically the DPO or reports directly to them. The DKP gives them a new estate to bring into compliance — their core competency.
3. **Platform stewardship.** Procurement, deployment, budget management, contractual governance with K-AI and integration partners. The CDO operates (or delegates to IT under their control) the transverse DKP infrastructure, the same way they operate their Data Catalog.
4. **Multi-domain federation.** In a large group with 12 directions and 50 Document Products, the CDO orchestrates cross-reviews, converges standards across domains, and surfaces directions falling behind. They play a transverse animation role no single business direction can hold.
5. **Consolidated view and ExCom reporting.** The CDO is the only role with the full picture: how many Document Products, which ones are at risk, what coverage internal AI agents have, what the aggregated compliance level is. They carry quarterly reporting to the ExCom on behalf of the federation.
6. **Joint C-Level sponsorship.** An HSE Director does not have the political weight to carry a transverse programme alone — they own their scope. The CDO carries the DKP programme at group level and unblocks C-Level arbitrations (prioritisation, budget, inter-direction RACI).
7. **Integration with existing governance stack.** The CDO ensures the DKP integrates with the enterprise Data Catalog (Collibra/Atlan/…), IAM, and observability stacks, with no new silo. They define the extended governance architecture covering structured and unstructured estates together.

**What the CDO does NOT do inside the DKP.** They do not arbitrate business conflicts (that is the Owner). They do not validate whether a procedure is correct (that is the Producer). They do not animate the SME community of a domain day to day (that is the Steward). This very clarification is what lets them be effective: they perform their transverse governance role without overreaching into editorial competencies they do not have.

**Example.** *The Document Authority of a European bank (Group Data Direction, 12-person team reporting to the COO):*

* *Sets the group standard: "every critical Document Product must have a freshness ≤ 30 days, an AI Act trace, and a named Director as Owner."*
* *Operates the K-AI group platform and its integration with Collibra.*
* *Runs a quarterly committee of 18 Document Stewards (one per direction).*
* *Presents the half-yearly "heat map" of Document Products and lagging directions to the ExCom.*
* *Never decides whether KYC policy version 4.2 is correct — that is the Compliance Director (Owner).*

### Document Producer

*Data world equivalent: Data Producer*

**Definition.** Subject-Matter Experts (SMEs) who create and maintain the documents under the authority of the domain's Document Owner. Occasional contributors to K-AI Audit (targeted review sessions, conflict arbitration, filling missing subjects) — not daily users. In practice, the Document Steward drives day-to-day activity and engages Producers when needed.

**Example.** *HSE engineers are the Producers of the Document Product "Industrial-site safety procedures". They are engaged by their Document Steward (HSE Knowledge Manager) to arbitrate conflicts detected during quarterly reviews. They do not live inside the platform.*

### Document Consumer

*Data world equivalent: Data Consumer*

**Definition.** End user of the clean, governed document layer. Two profiles: humans (via Document Discovery) and AI agents (via MCP). Consumers see neither the conflicts nor the duplicates — the platform handled that upstream.

**Example.** *An AI customer-support agent is a Consumer of the Document Product "Incident resolution base" — it queries the MCP and gets sourced, up-to-date answers without contradictions.*

### Document Steward

*Data world equivalent: Data Steward*

**Definition.** Operational arm of both the Document Owner and the Document Authority. The Steward animates the Producer community day to day, tracks per-domain quality KPIs, triggers periodic reviews, escalates structuring arbitrations to the Owner, and surfaces transverse issues to the Authority. Typically a Knowledge Manager inside a business direction, or a Data Steward whose scope has been extended to unstructured content.

**The role that spends the most time in the platform.** Owner = strategic sponsor (occasional). Producer = business contributor (targeted asks). Steward = daily user, animator, first point of contact for Producers.

**Example.** *The Document Steward for the HR domain of a large group: a Knowledge Manager attached to the HR direction, coordinates 18 business Producers across 4 Document Products, follows a weekly dashboard in K-AI Audit, escalates drifting Products to the HR direction (Owner) every quarter and transverse issues (freshness policy, taxonomies, compliance) to the Document Authority (CDO).*

### Document Engineer (DocOps)

*Data world equivalent: Data Engineer*

**Definition.** Technical profile that operates the platform: connectors to source repositories, ingestion and indexing pipelines, quality monitoring, remediation automations. The documentary counterpart of the Data Engineer.

**Example.** *The Document Engineer configures the SharePoint connector, monitors indexing throughput, and automates alerts when a Document Product breaches its SLA.*

### Master Document Management (MDM-D)

*Data world equivalent: Master Data Management*

**Definition.** Management of the enterprise's critical document reference set — the documents that hold authority, the single point of truth. A Master Document is unique, versioned, governed; any other source that drifts from it is considered derived.

**Example.** *The Master Document for a bank's network security policy is unique, signed by the CISO, versioned, and exposed to every compliance AI agent as the source of truth.*

### Semantic Document Layer

*Data world equivalent: Semantic Layer*

**Definition.** The active semantic layer of the DKP that understands relationships across documents and enables contextual querying. At K-AI, this layer is carried by the proprietary Neural Semantic Graph — a semantic graph that detects subtle contradictions which an ECM or a Data Catalog cannot surface.

**Example.** *The Semantic Document Layer detects that procedure A and service note B contradict each other on the response-time threshold, even though neither references the other.*

### Document Mesh

*Data world equivalent: Data Mesh*

**Definition.** Federated governance model in which each business domain operates its own Document Products under a shared standard (carried by the Document Authority), instead of a centralised model. The documentary counterpart of Zhamak Dehghani's Data Mesh.

**Usage note.** Most large groups are not yet mature on Data Mesh — they are still at the Data Catalog stage. When the term Document Mesh feels too advanced for the audience, paraphrase: *"a model where each business domain operates its estate under a shared standard."* The concept stays, the term steps back temporarily.

**Example.** *A multi-division industrial group adopts the Document Mesh: each division (Energy, Mobility, Chemistry) operates its Document Products with its own Owners and Stewards, under a standard carried by the Group Document Authority (CDO).*

### Document Contract

*Data world equivalent: Data Contract*

**Definition.** Formal commitment between a Document Owner and Consumers (human or AI) describing the guarantees offered by a Document Product: scope, freshness, quality level, ACLs, exposure format.

**Example.** *The Document Contract of the Product "Commercial pricing base" commits to: 24h freshness, 99% non-contradiction quality, MCP exposure with SAP-mirrored ACLs, 99.9% availability SLA.*

### DocOps / KnowledgeOps

*Data world equivalent: DataOps*

**Definition.** Industrialisation practices applied to the document lifecycle: continuous ingestion, automated quality tests, monitoring, alerting, documentary CI/CD, automation of trivial remediations.

**Example.** *A DocOps pipeline detects new conflicts every night, automatically opens a Jira ticket for the relevant Producer, and alerts the Steward above a defined threshold.*

### Active Metadata (documentary)

*Data world equivalent: Active Metadata (Gartner)*

**Definition.** Metadata continuously enriched by AI — beyond technical fields (title, author, date), the platform captures the semantic subject, detected conflicts, quality score, dependencies on other documents, and downstream usage (who consumes it).

**Example.** *A document inherits Active Metadata automatically: "covers subject X", "in conflict with doc Y on point Z", "consumed by 3 AI agents", "quality score 82".*

### Document Discovery

*Data world equivalent: Data Discovery*

**Definition.** Ability to find the right document through natural language, without knowing its location, filename or folder structure. This is the DKP's promise on the human Consumer side.

**Example.** *An employee types "procedure to follow when a pressure sensor fails on line X" and receives a sourced answer, without needing to know where the document is stored.*

### Document Observability

*Data world equivalent: Data Observability*

**Definition.** Real-time monitoring of the health of the document estate: newly detected conflicts, emerging uncovered subjects, orphaned documents, user or AI-agent queries left without a satisfactory answer. It is the implementation of the DKP's continuous-observability principle.

**Example.** *The Observability dashboard surfaces: 14 new conflicts this week, 3 recurring user subjects without answer, 1 Document Product drifting on freshness.*

## Brand vocabulary

For the do-list and don't-list of vocabulary applied across this documentation, see the project [CLAUDE.md](https://github.com/KAI-Internal/kai-gitbook/blob/main/CLAUDE.md). The short version: use the DKP lexicon; avoid "RAG" (for K-AI), "chunking", "internal search", and superlatives.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://k-ai.gitbook.io/knowledge-ai/reference/glossary.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.