Meet your Twin
A shared, trustworthy picture of how your business works, ready for the AI tools you choose.
Your Twin gives AI the business context it normally lacks.
It brings together the meaning behind your customers, finances, operations, processes, and measures. That meaning comes from your real business systems, is reviewed by your people, and stays under your control.
The result is a dependable foundation that different AI tools can use without each one having to rediscover how your business works.
Why businesses build a Twin
Most AI tools are good at general knowledge. They do not automatically know:
- which revenue definition your finance team has approved;
- how a customer moves through your own sales and service processes;
- which system is authoritative when two sources disagree;
- which exceptions matter to your team; or
- whether a figure is current enough to use.
A Twin turns that company-specific knowledge into something reusable. People can work with it in the Twin workspace, and approved assistants such as ChatGPT or Claude can use the same published understanding.
How the Twin works
Follow the journey from connected evidence to knowledge that AI can use.
Build your first Twin
Start with one useful business outcome and grow from there.
Sources and availability
See which modern and legacy systems can contribute evidence and how readiness is confirmed.
Proof from real systems
Review the measured finance examples and a practical evaluation method.
Review and publish changes
Check proposed business meaning and decide what becomes usable.
Understand a Twin answer
See what makes an answer trustworthy and easy to inspect.
What makes it different
Use the systems you already run
The Twin is built from the systems your business already uses. It keeps track of where available information came from while those systems remain authoritative. Modern services and legacy applications can contribute to the same reviewed result without a wholesale migration.
Your people stay in control
Your Twin can suggest useful definitions, relationships, and measures. A suggestion does not become shared business knowledge until an authorised person has reviewed and approved it.
Prepare important calculations once
Where an activation uses a prepared calculation, the approved definition and business logic can be applied before an AI assistant asks the question. This reduces repeated interpretation and helps the same source state produce the same governed result with its evidence attached.
Actual response time and model cost depend on the configured deployment. The important distinction is that the calculation is part of the reviewed Twin context, not invented afresh by each assistant.
Choose models without rebuilding the business context
Your approved business meaning is kept separate from model routing. A deployment can select an approved model suited to a task, or change provider after the required evaluation and controlled rollout, without reconstructing the Twin from the beginning.
An end user or connected client does not select arbitrary models. Model and provider availability are configured for the organisation.
It is designed to say “I don’t know”
When the approved knowledge does not support a question, the Twin should fail clearly rather than produce a plausible guess.
A Twin is not a copy of your business
It is a governed picture of the parts of your business that you have chosen to make useful for AI. Source systems remain authoritative, and the Twin can grow in small, reviewed steps.
Friendly language and technical terms
Most people only need the plain-language description. Technical references use more exact terms when a version or interface contract matters.
| Plain-language term | What it means |
|---|---|
| Source evidence | Permitted records or documents observed from an authoritative business system |
| Proposed business context | A suggested definition, relationship, process, or measure that still needs review |
| Approved business context | Business meaning that an accountable person has reviewed |
| Published Twin product | The stable, versioned technical unit that carries approved context, its contract, and validation evidence |
| Activated version | The exact published version selected for use by a configured deployment |
| Governed answer | A result produced through the approved, read-only path against the selected product version |
These are stages of the same Twin, not separate products.
A simple example
Imagine that several teams use the phrase “monthly revenue,” but each team calculates it differently.
The Twin can bring together the relevant accounting evidence, document the intended definition, show the exclusions, and let a finance owner review the result. Once approved, that version becomes a published piece of Twin knowledge.
People can then ask the same question from their Twin workspace, ChatGPT, Claude, or another permitted tool and receive an answer based on the same reviewed definition. The answer identifies which version it used and its governed evidence references. Where the configured interface exposes a source watermark, it can also show how current the supporting information was.