How the Twin works
See how your Twin turns source evidence into reviewed business knowledge that people and AI can use.
The Twin grows through a clear, repeatable journey. Each stage has a different purpose, and important changes remain visible to the people responsible for them.
1. Connect source evidence
Your Twin connects to the business systems you approve, such as accounting, documents, email, meetings, or operational tools.
Connections normally begin read-only. The original systems remain the source of truth while the Twin records where information came from and when it was observed.
2. Discover what the evidence can support
The Twin looks for useful business concepts in the available evidence. These might include:
- a dataset that brings related records together;
- a relationship between customers, invoices, and payments;
- a business process with recognisable stages; or
- a measure such as monthly revenue or outstanding debt.
At this point they are suggestions, not approved facts.
3. Review the suggestions
A person who understands the business reviews what the Twin found. They can check the proposed meaning, source mappings, exclusions, and exceptions.
The reviewer can approve a useful suggestion or leave it unpublished when more work is needed.
4. Publish approved knowledge
Approved knowledge is published as a clearly identified version. Publishing creates a stable reference that can be checked and prepared for use.
Future changes create a new version rather than silently changing what an earlier answer meant.
5. Activate the intended use
Publication and activation are separate decisions. An organisation administrator or implementation lead activates the exact version only after its required validation and deployment checks pass.
Making that version available to an external assistant requires another explicit decision about what may be shared, which people may use it, and which connection is enabled. A published version is not automatically available in ChatGPT, Claude, or another tool.
6. Put the Twin to work
People can use activated knowledge through the Twin workspace and other routes approved for their organisation, such as ChatGPT, Claude, Codex, or their own software.
Every route reaches the same governed Twin. Connecting another tool does not create a separate copy.
7. Evaluate and improve
Teams can test the Twin with real business questions and record whether the result met the expected standard.
New source evidence may suggest an update. Material changes return to review before a new version is published.
Each stage is a separate decision
Discovering something does not approve it. Approval does not publish it, and publication does not activate it or make it available to every AI tool. Validation, activation, access, and evaluation remain separate controls.
What you can see in your Twin workspace
Organisation administrators can follow the same journey in the Twin workspace:
- Overview shows the Twin’s current state and the next decisions to make.
- What it knows shows governed datasets, relationships, and processes that have been constructed from connected evidence.
- Review changes shows proposed improvements waiting for a decision.
- Source coverage shows which connected records have a governed business view.
- User demand shows recurring themes in what people are asking the Twin to do.
You do not need to understand the underlying technical architecture to review or use the Twin. The service keeps the detailed history so that the result remains inspectable when it matters.
What it knows is not a release checklist
Seeing a governed dataset or relationship in this view does not by itself prove that a published version is active, evaluated, or approved for an external AI connection. Your implementation lead confirms those readiness steps separately.