What is an ontology-based operations data platform?
A plain definition, what it is not, and five questions to ask before buying one.
Converge Team, 2026-11-11
An ontology-based operations data platform is software that keeps one shared model of a business (its customers, orders, assets and the links between them) and lets people, applications and AI agents both read that model and change it under rules. The model is the ontology. The rules are permissions, approvals and an audit trail.
That one sentence has three parts, and each separates this kind of platform from something else.
One shared model
Most companies already have the data. What they lack is agreement: the ERP says one thing about an order, the warehouse system another, and a spreadsheet joins the two. An ontology replaces the spreadsheet. It declares object types (an Order, a Vehicle, a Customer), the links between them (this Order belongs to that Customer), and where each property comes from.
The point of declaring it once is that everything else reuses it. A dashboard, a mobile app, a public API and an AI assistant all see the same Order, filtered by the same permissions.
Read and write
A data warehouse or a business intelligence tool is built to answer questions. An operations platform is built so that the person who sees a wrong number can fix it there. The fix is an action: a named change with declared inputs, validation, a check of who is asking, and a record of what happened. An action can wait for a second person to approve it, and an approved action can write the change back to the system the number came from.
Without the write path, people go back to email and spreadsheets to act on what the dashboard showed them.
Under rules
Because people and software change real records through the platform, the rules matter more than in a reporting tool:
- Roles and markings decide who may see or change which objects.
- Approvals hold important changes until someone else signs off.
- An audit log keeps who did what, and cannot be quietly edited.
AI agents follow the same rules. An agent reads the ontology the way a person does and changes data only by calling an action, so it can never do more than the person it works for.
What it is not
| It is not | Because |
|---|---|
| A data warehouse | A warehouse stores tables for analysis; it has no notion of an action or an approval |
| A BI tool | BI shows numbers; it does not let you correct them at the source |
| An ERP | An ERP runs one process end to end; an ontology platform sits across several systems and models how they relate |
| An integration tool | Integration moves data between systems; it does not give people a place to work on it |
| A knowledge graph alone | A graph stores the model; the platform adds applications, actions, permissions and audit on top |
Five questions to ask before buying one
- Can a business user correct a record where they see it, and does the correction reach the source system?
- Who can approve a change, and can anyone approve their own?
- Does a permission set on data hold everywhere, including in an AI answer and in an export?
- Where does the data live, and is one customer's data separate from another's?
- How long from a new source to a working object type? Ask to see it done.
Where Converge fits
Converge is an ontology-based operations data platform delivered as a hosted service: each customer company connects its own systems, keeps its data separate from every other customer's, and its people and AI agents work on one model through approved actions. The overview describes the parts, and the frequently asked questions give the short answers.