Skip to main content
墨舜 AI LabMoShun AI Lab
Applications

03 · UNDERSTAND

Enterprise Knowledge & Internal Q&A

Knowledge Fabric

Turning documents, policies, project material and system data into enterprise capability that can be searched, cited and kept current.

01Overview

What this is

An organisation’s most valuable knowledge is usually scattered across documents, mail, group chats and individual experience. A knowledge base is worth building not because things become findable, but because answers carry provenance, permissions are actually enforced, and content keeps up with the organisation.

02Problem

The problem

Keyword search misses semantically close phrasing; pure vector retrieval readily returns passages that are related but not correct. Worse, an answer without provenance cannot be trusted inside an organisation — and no one can trace where it came from.

03Capability

Core capabilities

  • 01

    Document parsing

    Handles structurally awkward contracts, policies, tables and scans while preserving structure.

  • 02

    Hybrid retrieval

    Vector and keyword retrieval combined, covering both near-synonyms and exact terminology.

  • 03

    Citation and provenance

    Every answer names its source document and location, so it can be verified against its origin.

  • 04

    Permission management

    Results are filtered by the requesting user’s permissions; different roles see different scopes.

  • 05

    Continuous refresh

    The base is cleaned and extended continuously rather than loaded once and left.

04Flow

How it works

  1. 01

    Ingest

    Connect document stores, business systems and existing repositories.

  2. 02

    Structure

    Parse and chunk while preserving heading hierarchy and table structure.

  3. 03

    Index

    Build vector and keyword indexes together, carrying permission labels.

  4. 04

    Answer

    Retrieve within permission, then generate an answer with citations.

  5. 05

    Feed back

    Misses and corrections return to the pipeline to extend and clean the base.

05Position

Where it sits across the capability domains

InputOutcome
  1. 01 · SENSESense & InteractAlso touches
  2. 02 · UNDERSTANDKnowledge & JudgementPrimary domain
  3. 03 · CREATEContent & GenerationNot directly involved
  4. 04 · ACTAutomation & ExecutionNot directly involved
  5. 05 · ORCHESTRATEEnterprise OrchestrationNot directly involved
The primary domain defines what this application is responsible for; the domains it also touches show what it has to work with. The diagram describes position only — it is not a statement of delivery scope.
06Use cases

Use cases

  1. 01

    Policy Q&A

    Staff ask about policy, process and entitlement directly, and get cited answers.

  2. 02

    Sales and pre-sales

    Pulls product capability, references and pricing rationale quickly.

  3. 03

    Onboarding

    Turns scattered project experience into a searchable organisational asset.

07Connections

Enterprise system connections

  • Document and file storage
  • Internal wiki / OA
  • CRM and ticket history
  • Data warehouse
  • Identity and permission systems
08Governance

Data, permission and deployment

Permissions are enforced at the retrieval layer, not masked at the presentation layer. Private deployment keeps data in place, and sensitive documents can carry their own access policy and retention period.

09Availability

What is open today

Delivered as an enterprise engagement. The usual path is a narrow pilot on one department’s real material, expanding once retrieval quality is confirmed.

10Related
11Updates
Enterprise Delivery

Apply for access

Want to know what this does in your own context?

Tell us the scenario and the systems already in place. We will start by judging whether it is worth doing at all, then talk about how.

Delivered to enterprise customers as scoped engagements; scope, timeline and integration are defined per project.

Apply for accessailab@moshuntech.com