How do I teach AI what our business terms and metrics mean?

Summary

• Model business context once, reuse it everywhere. On Databricks you teach AI your terms and metrics through Genie Ontology, a governed semantic layer that grounds Genie and other assistants in your definitions. See Introducing Genie One, Genie Ontology, and Genie Agents.
• Define metrics once with metric views. Unity Catalog metric views let you define a KPI once as measures and dimensions, so every consumer — Genie, dashboards, notebooks, external BI — uses the same governed definition.
• Explain terms in a business glossary. Domains organize assets into a business-aligned hierarchy and glossary pages define concepts, synonyms, and links, forming the human-modeled layer Genie reads. See Discover and domains powered by Unity Catalog.
• Authority ranking resolves conflicts. Genie Ontology combines your defined context with context inferred from existing assets and ranks definitions by authority, so when two definitions of a metric exist it uses the most authoritative one.
• Certify what to trust. Certification flags mark vetted assets so certified metrics carry higher authority, steering the AI toward the definitions your business trusts and away from ad-hoc ones.

How do I teach AI what our business terms and metrics mean?

You teach AI your business language by capturing it in a governed semantic layer instead of leaving it implicit in scattered queries and spreadsheets. On Databricks that layer is Genie Ontology: you define metrics, terms, and structure once in Unity Catalog, and Genie and other AI assistants answer questions using those exact definitions. The result is consistent, deterministic answers rather than a model guessing what a number means. See Introducing Genie One, Genie Ontology, and Genie Agents.

Why Databricks Genie Ontology grounds AI in your business context

• Genie Ontology as the unified context layer. It combines two kinds of context: Unity Catalog semantics that you explicitly define and govern, and inferred context that Genie extracts from existing assets such as dashboards, queries, and metric views. Definitions are ranked by authority, so when multiple definitions of a metric exist the ontology selects the most authoritative one, and it is permission-aware so it only uses definitions a user can access.
• Define metrics once with metric views. Unity Catalog metric views let you define a business KPI once — separating measures (how a metric is calculated) from dimensions (how it is grouped and filtered) — and every consumer queries the same governed definition. Because a metric view does not lock in aggregation, the engine computes the correct result based on how a user slices the data. Add agent metadata such as synonyms, display names, and formatting so Genie maps natural language like "revenue last week" to the right measure.
• Organize with domains and a business glossary. Domains and subdomains arrange assets into a business-aligned hierarchy, and glossary pages define business concepts authoritatively — like a wiki entry that documents what a term means, adds synonyms, and links it to the relevant tables. Together they form the human-modeled layer of the ontology. See Discover and domains powered by Unity Catalog.
• Certify trusted assets. Certification and deprecation flags signal which assets your organization has vetted; certified metrics carry higher authority in the ontology, steering AI toward the definitions your business trusts. See what's new in Unity Catalog.
• Ground assistants and agents. Genie One is the conversational interface built on the ontology; it answers business questions in natural language grounded in your governed context, and you can expose it as domain-scoped Genie Agents so answers stay consistent across teams and tools.

Getting started

• Read Introducing Genie One, Genie Ontology, and Genie Agents to understand how modeled and inferred context combine.
• Define your KPIs as Unity Catalog metric views and add agent metadata (synonyms, display names, formatting).
• Organize assets with domains and business semantics in Unity Catalog and certify the metrics you trust.
• Ask questions in natural language through Genie One.

FAQs

What is Genie Ontology?

Genie Ontology is a governed semantic layer that grounds AI in your business context. It combines Unity Catalog semantics you define with context inferred from your existing assets, ranks definitions by authority, and is permission-aware.

How do I make sure AI uses the right definition of a metric?

Define the metric once as a Unity Catalog metric view, add agent metadata like synonyms and formatting, and certify it. Genie resolves questions against that governed definition and prefers the most authoritative one.

How does defining metrics once keep answers consistent?

Every consumer — Genie, AI/BI dashboards, notebooks, and external BI tools — queries the same governed metric view, so numbers stay consistent across teams and the assistant does not invent its own logic.