Use cases / Knowledge and engineering

AI Knowledge Management

Give AI the context to understand the organization, without flattening access rules.

01

Problem

Knowledge is split across drives, chat and heads. Search returns documents, not answers, and naive AI indexing ignores who is allowed to see what.

02

AI capability

Ingest and index sources, keep permissions attached to content, answer questions with citations and surface stale or conflicting guidance.

03

Data

The agent works from the context you connect, not from the open internet by default.

  • Document stores
  • Wikis and tickets
  • Meeting notes
  • Existing access control lists

04

Agent

Knowledge Agent

Answers questions from permitted content and flags gaps and contradictions.

05

Workflow

  1. 1. Connect sources
  2. 2. Index with permissions
  3. 3. Question asked
  4. 4. Retrieve permitted passages
  5. 5. Answer with citations
  6. 6. Flag gaps for owners

06

Governance

Governance runs inside the agent at runtime: policy changes what it can actually do.

Can

  • Read sources it is connected to
  • Answer within the asker permissions
  • Cite passages
  • Flag stale content to owners

Cannot

  • Reveal content the asker cannot open
  • Edit source documents
  • Index restricted sources without a classification
  • Retain content beyond the retention rule

Requires approval

  • Connecting a new source
  • Raising a source classification

Records

Agent identity, Data accessed, Model used, Output, Tools called, Policy applied, Decision, Approval, Action, Outcome.

Suggested starting autonomy: L3 Supervise

L3 suits read-only answering: it acts within limits and is monitored. Source connection stays human-approved.

How controlled autonomy works

07

Outcome

What you measure, so the agent earns more autonomy on evidence:

  • Answer acceptance by users
  • Share of answers with citations
  • Stale or conflicting documents resolved
  • Permission-leak tests passed

Questions

Does it respect existing permissions?

It is built so retrieval is filtered by the asker permissions; content they cannot open is not retrieved.

Which Swfte product is this?

Cortex provides the knowledge layer, with Connect for source integrations.

Build it on the platform

Related use cases

Centralise your knowledge in Cortex

The desktop AI workspace: 20+ providers, local models, knowledge bases with RAG that cite their sources, MCP tools and agents, with sensitive work staying on the laptop by default.