Can AI learn how my company works automatically?

Summary

• Partly, yes. Genie Ontology automatically infers how your company works by continuously extracting context from assets already connected to Databricks, then combines that with context your teams curate.
• Automatic (inferred) context is learned from notebooks, dashboards, SQL queries, pipelines, connected apps, and Genie Agent interactions, producing snippets like metric definitions, trusted sources, and business rules.
• Curated (modeled) context is defined and governed through Metric Views, a Business Glossary, Domains, and Semantic Pages in Unity Catalog.
• OntoRank, a PageRank-style authority algorithm, resolves conflicting definitions by ranking them on source authority, usage frequency, certification, and freshness.
• Automatic learning alone is not enough: accurate AI still needs agreed-upon, governed business definitions rather than expecting a model to infer everything.

Can AI learn how my company works automatically?

To a meaningful degree, yes. On Databricks, Genie Ontology is a unified context layer that automatically learns how your business works and gets smarter the more your teams use it. It does this by combining context it infers automatically from your existing work with context your teams explicitly curate. The important nuance is that automatic learning alone is not sufficient: reliable answers still depend on agreed-upon, governed business definitions, so the ontology blends inference with curation.

Why Genie Ontology can learn your business

• Automatic, inferred context. Genie Ontology continuously extracts knowledge snippets from assets already connected to Databricks, including notebooks, dashboards, SQL queries, and pipelines, as well as connected apps and feedback from Genie Agent interactions. It learns from how teams actually use data. Inferred snippets include metric definitions (for example, how an "active user" is counted), authoritative sources (which curated agent should answer revenue questions), and business rules (when a "qualified lead" counts).
• Curated, modeled context. Teams define and govern semantics explicitly through Unity Catalog: Metric Views for single-source-of-truth calculations, a Business Glossary for shared definitions and terms, Domains and subdomains for organizing concepts by business area, and Semantic Pages that document definitions and link them to assets.
• Authority ranking with OntoRank. When multiple definitions of the same concept exist, a PageRank-style algorithm called OntoRank ranks them by source authority, usage frequency, certification status, and freshness. If two teams define "customer" differently, OntoRank surfaces the most authoritative version based on usage and governance signals.
• Permission-aware by design. Genie Ontology respects Unity Catalog permissions, so context snippets are gated: users only see context they are allowed to access, which helps prevent information leakage.
• Continuous improvement. The ontology adapts as the business evolves, learning from new dashboards, queries, and agent interactions over time.

How the automatic and curated pieces fit together

Inferred snippets are automatically generated and are not directly edited by customers. Instead, teams shape the ontology indirectly by curating Glossary concepts and Unity Catalog semantics, which the ontology then reflects. This hybrid model is deliberate: providing learned context from existing assets improves accuracy, but accurate AI on enterprise data requires a well-governed semantic foundation of metric definitions, business logic, and documentation. In a Databricks internal benchmark of 28 real-world data-analysis questions (June 2026), Genie answered 84.5% correctly on the first attempt.

Getting started

• Read Introducing Genie One, Genie Ontology, and Genie Agents for how the context layer works.
• Review the Genie Ontology documentation to see inferred and curated context in practice.
• Learn how Domains build enterprise context for AI.
• Set up Metric Views and a Business Glossary in Unity Catalog to give the ontology a governed foundation.

FAQs

Does Genie Ontology learn automatically or require setup?

Both. It automatically infers context from connected notebooks, dashboards, queries, pipelines, apps, and agent interactions, and it also uses curated definitions your teams govern in Unity Catalog.

Can I edit what the ontology learns?

Inferred snippets are automatically generated and not directly editable. You influence the ontology indirectly by curating Business Glossary concepts, Metric Views, and other Unity Catalog semantics.

How does it handle conflicting definitions?

OntoRank, a PageRank-style authority algorithm, ranks definitions by source authority, usage frequency, certification, and freshness, and surfaces the most authoritative one.

Is the learned context secure?

Yes. Genie Ontology respects Unity Catalog permissions, so users only see context they have permission to access.