The Ontology Charter
Govern your ontology without slowing your teams down
The Charter sets out who owns each object and link, how the model changes without breaking the applications and agents on it, who may see what across the ecosystem, and what depends on what. We have developed a proven way to enforce it. Stood up with your team in weeks, then run without us.
Why a charter
Make the choice and write it down.
Is the ontology shared infrastructure, or a by-product of each project?
Most enterprises never choose, and end up with as many definitions of “customer” as projects.
The Charter records the choice: a named owner for each definition, rules for changing it, and version control, so people and agents can rely on it.
Governance used to protect reports. Now it also controls what agents do.
- A wrong definition once made a bad dashboard, and a person caught it
- Agents act on wrong definitions directly, across every shared workflow
- EU AI Act, DORA and model-risk reviews now expect evidence of governed meaning
Where does the model live: per use case or first-class?
- Per use case: “customer” defined once per project, duplicated logic, nothing reusable
- Most enterprises choose implicitly, one deadline at a time, and fork
- Two years later there are three versions of the “single source of truth”
Ownership and checks built into the work
- Named owners, versioned change and automated checks, so teams change the model without waiting for a meeting
- Use cases extend the model instead of forking it
- Every extension makes the next use case cheaper
Governance keeps the model in order; putting it to work is an engagement built to hand over.
THE FOUR PARTS
Ownership, change, access, lineage
Rules people and agents can rely on, for the people who run the business and the agents that now act in it.
What the Charter sets down
One document, four parts, each with a named owner and a review cycle. We have run every part inside Foundry ontologies at enterprise scale; the Charter adapts to whatever your semantic layer runs on.
Ownership
Who owns each object and link. Every object type, action and vocabulary has a named steward with real authority over the definition, the data and the consumers.
Change
How the model evolves without breaking the applications and agents on it. Change treated like an API: proposal, deprecation window, migration path, every change tested against what consumes it.
Access
Row and column governance across the ecosystem. Who may see a table; who, or what, may act on it. Access by purpose, with an audit trail that stands up to EU AI Act and DORA reviews.
Lineage
What decisions and agents depend on what. Every reported figure traced to its sources; every agent’s tool call traced to the action it invokes and the definitions it relies on.
Free
Start with the Ontology Health Check
Bring your ontology as it stands: a Foundry ontology, a semantic layer, or the definitions your applications carry today. No deck. On a call, we work through four questions and you leave with a one-page finding: what is governed, what is not, and what to fix before an agent acts on it.
Book the health check- How many live definitions of “customer” exist across your applications and agents
- Who owns each object and link today, and who does not
- Where definitions have drifted apart since they were last agreed
- Which decisions and agents depend on the definitions most likely to move
Is your semantic layer still accurate?
One call shows how the Charter would work for your model.
info@ainurinsights.com