The same customer may appear in your CRM, billing system and support records under three identifiers. Answering “which orders for this customer need attention?” requires matching records, understanding statuses and checking applicable rules. Connecting applications does not explain these business relationships by itself.
Ontologie provides a shared business model. You define Customer and Order, their properties, relationships and useful operations. People, agents and integrations can use the same definitions to work with data.
Example: tracking a customer's orders
In the walkthrough, Client is the customer type and Commande is the order type. A customer has a stable identifier and a name. An order has a reference, amount, currency and status. Client passe Commande means “Customer places Order”. ACME France and CMD-001 are instances: actual records of those types.
View the diagram description
flowchart LR CRM[CRM : customers] --> Mapping[Mappings and identifiers] ERP[ERP : orders] --> Mapping Mapping --> Customer[Customer type and instances] Customer -->|places| Order[Order type and instances] Documents[Contractual terms] --> Search[Document search] Order --> Agent[Agent : verifiable answer] Search --> Agent Agent --> Decision[Human decision]
The diagram describes a workflow: map sources, inspect records, find terms and decide. It does not imply automatic merging of customers with similar names. You must define and verify how source identifiers correspond.
An agent can retrieve accessible ACME orders and relevant contract passages. A useful answer includes record references, the period examined and missing information. You can check records in the Instances tab and open the cited documents.
How it complements your applications
| Need | What the model provides | Your responsibility |
|---|---|---|
| Understand different schemas | Shared types and properties | Define mappings and units |
| Connect customers and orders | Explicit relationships | Choose stable identifiers and cardinality |
| Query records and documents | Shared context for retrieval and agents | Check evidence and freshness |
| Reuse an integration | Published operations with a contract | Check the catalogue and permissions |
| Prepare an intervention | Available tools and authorized actions | Check inputs and resulting state |
Databases still provide storage and queries. The business model explains what the data means across use cases. This is useful when a decision involves several applications or teams need to agree on the meaning of “active order”.
Where is the data stored?
A source application can remain the business system of record while the platform stores synchronized copies, native instances and document indexes. Access depends on the connector and data type. The Sources tab shows linked sources; the Instances tab distinguishes native records from external records.
Correct an external record in its source application and synchronize again. Edit a native instance in the platform when your permissions allow it. Imported documents have their own processing and indexing lifecycle. Check each stage before treating an agent's answer as the current state of a source application.
Start with a measurable objective
Choose a narrow question: “find orders for customer CLI-001 and explain the delivery condition”. Start with two types, one relationship and a few demonstration records. Expand after this journey works.
- Define essential properties and their meaning.
- Create one Client instance and one Commande instance.
- Check the relationship and provenance.
- Ask a reading question using an agent with the necessary capabilities.
- Compare the answer with the records and documents it used.
Your first outcome is an answer you can verify. Writing requires an available capability, suitable permissions and a check of the resulting changes. Instructions in a prompt do not replace these controls.
