What is an ontology in AI?

Summary

• An ontology in AI is a formal, shared vocabulary for a domain. It names the concepts that exist, their properties, and the relationships between them.
• It makes meaning explicit instead of leaving an AI system to infer it. An agent that can look up a governed definition of "net revenue" does not have to guess which column or formula the business means.
• An ontology is not a database schema and not a knowledge graph. A schema describes how one database is structured; a knowledge graph stores facts and may use an ontology as its schema layer.
• Genie Ontology (Public Preview) is the Databricks context layer for Genie. It combines business semantics you define in Unity Catalog with context Genie infers from your assets and usage.

What is an ontology in AI?

An ontology is a formal, explicit description of the concepts in a domain, the properties of each concept, and the restrictions on those properties. The Stanford guide Ontology Development 101 uses this definition, and the W3C describes ontologies as "formalized vocabularies of terms, often covering a specific domain and shared by a community of users".
In a business, an ontology might state that a Customer has a name and an email address, that an Order belongs to a Customer, and that an Order contains Line Items. For AI, writing this down makes the meaning explicit and machine-readable, so a model does not have to infer what "customer" or "order" means from table and column names alone.

How do ontologies help AI systems answer questions?

Take the question "What was net revenue for Q3?" Which definition of net revenue applies, which tables hold it, and whether Q3 follows the fiscal or the calendar year are all open questions. An ontology records those decisions once, in a governed place that an agent can consult. Relationships matter as much as definitions: Databricks' guidance on operationalizing Genie Ontology notes that agents join tables to answer questions, and "if they have to guess how tables connect, they will sometimes guess wrong".

What are the key elements of an ontology?

Element	What it captures	Example
Classes (concepts)	The kinds of things in the domain	Customer, Order, Product
Individuals (instances)	Specific members of a class	One particular order
Attributes (properties)	What is known about a class	A Customer has a name and an email address
Relations	How classes connect	An Order belongs to a Customer
Hierarchies	Broader and narrower classes	An Enterprise Customer is a kind of Customer
Definitions and restrictions	The agreed meaning of a term and its allowed values	Net revenue is gross revenue minus returns and discounts (an example definition)

How does Genie Ontology provide business context in Databricks?

The documentation lists Genie Ontology as Public Preview and defines it as "the unified context layer that gives the Genie family of AI experiences a business-aware map of your organization". It combines two kinds of context:

• Modeled context. Unity Catalog semantics lets you define business metrics, terms and organizational structures on top of your catalog data through metric views, domains and Pages. A Page, currently in Beta, is a governed definition of a business concept such as a term, entity or acronym, and Genie One prioritizes it over context it infers automatically.
• Inferred context. Genie automatically extracts snippets from your existing assets and usage, including metric views, dashboards, SQL queries and Genie Agents. Each snippet gets an authority score based on where it was generated, how often it is used and how fresh it is.

Snippets are gated by Unity Catalog permissions, so Genie only uses snippets from assets the user has permission to see, and citation icons show which knowledge sources a response used. Genie One and Genie Code search the same ontology, so context you curate once applies to both.
The Databricks blog frames the split this way: a good semantic model "provides an authoritative core", and an ontology "extends that foundation with the broader relationships, knowledge, and context AI needs".

Where are ontologies used in practice?

Field	Shared vocabulary	What it provides
Finance	FIBO	Defines the things of interest in financial business applications and how they relate, developed in the Web Ontology Language (OWL)
Healthcare	SNOMED CT	A clinical terminology whose core components are concepts, descriptions and relationships
Web content	schema.org	Shared types such as Product and Offer for marking up web pages

FAQs

Is an ontology the same as a data model or database schema?

No. A database schema is the structure of a database, described in a formal language, with its tables, fields and relationships. An ontology is a shared vocabulary for a domain, not the layout of one database. FIBO, for example, can give meaning to data held in spreadsheets, relational databases and XML documents.

What is the difference between an ontology and a knowledge graph?

A knowledge graph is a knowledge base that uses a graph-structured data model. Knowledge graphs may use ontologies as a schema layer, which allows logical inference to retrieve implicit knowledge. The Stanford guide draws a similar line: an ontology plus a set of individual instances of its classes constitutes a knowledge base.

How do I build an ontology?

Ontology Development 101 describes an iterative process: determine the domain and scope, consider reusing existing ontologies, enumerate the important terms, then define the classes, their hierarchy and their properties. For Databricks, the guidance is to start with one high-value domain and strengthen its definitions, sources, governance and evaluation.

Do I need an ontology for every AI use case?

No. Databricks' guidance says you do not have to build all six of its layers at once and should not try to cover the whole business before going live.

Is Genie Ontology generally available?

No. The documentation lists it as Public Preview. According to the AI/BI release notes, ontology snippets became available to all customers, without requesting access, on August 13, 2026.