Unity Catalog Pages: a governed home for your business knowledge in Genie Ontology

Databricks introduced Unity Catalog Pages, a human-curated layer of its Genie Ontology enterprise context system that lets data stewards define authoritative business concepts for AI agents. Pages combine structured fields — an owner, synonyms, and a description — with a rich body holding links, images, tables, and inline references to Unity Catalog and workspace objects, and Genie One cites the Pages used to ground an answer as clickable sources. Databricks said the feature addresses ambiguity that causes LLMs to guess when terms like "active customer" or "completed trip" are defined differently across teams.

Enrich Genie Ontology with your organization’s actual business definitions, turning company-specific terminology into authoritative context for data and AI agents. by Tim Feng https://www.databricks.com/blog/author/tim-feng and Ashish Singh https://www.databricks.com/blog/author/ashish-singh Every AI agent is only as good as the context it is grounded in. Ask an agent a question about revenue, active customers, or churn, and the quality of the answer depends entirely on whether the agent understands what those words mean inside your business. In many organizations, that meaning does not live in one place. It is scattered across Slack threads, Confluence pages, spreadsheets, and the tribal knowledge in a few people's heads, and the same term is often defined three different ways by three different teams. So when an agent hits an ambiguous concept, it does what LLMs do best: it guesses confidently, even when it is wrong. This gap is one of the biggest barriers to enterprises trusting AI with real business questions. This is the problem Genie Ontology https://www.databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents was built to solve. Genie Ontology is Databricks’ enterprise context layer for all AI: it automatically learns how your business works by extracting knowledge from your dashboards, queries, tables, pipelines, and connected apps, and organizes it into a living graph that tells Genie and other agents where to look and what to trust. That automatic understanding covers an enormous amount of ground on its own. But some concepts are too important to leave to inference. When "completed trip" or "active customer" has to be exactly right, you want your own experts to define it once, in a place every person and every agent can rely on. Unity Catalog Pages fill this exact gap. As the newest piece of Unity Catalog semantics, the human-curated layer of Genie Ontology, Pages provide a governed home where your data stewards, with the help of Genie Code, can now define the authoritative meaning of a concept, and Genie treats that definition as the source of truth. Genie Ontology brings two kinds of context together. Alongside everything it learns automatically, it draws on the definitions your teams model explicitly in Unity Catalog semantics. Each modeled piece plays a distinct role: These are complementary, not interchangeable. Metric views tell the agent how to calculate ; Pages tell it what a concept means ; domains tell it where to look; certification tells it what to trust . Together with the knowledge Genie learns on its own, they form a single, governed picture of your business. Now, when a question to Genie One touches on a concept your organization has explicitly defined, Genie can retrieve the corresponding Page and use its definition to help interpret the request, rather than relying solely on inference. For example, if your sales organization has documented what qualifies as an "active customer" and which table to derive that entity from, Genie One can use that exact definition when an analyst asks it to analyze 30-day customer trends during a major sales push—delivering accurate results without guessing. To tie the experience together, Unity Catalog Pages used to ground an answer are cited as clickable sources, so users can inspect the definition, see who owns it, and understand why that context was used. This turns grounding from a hidden AI decision into a verifiable train of thought that users can review before acting on the results. Each Page combines structured fields an owner, synonyms, and a description with a rich body that can hold links, images, tables, and inline references to the Unity Catalog and workspace objects the concept depends on. Pages live in Discover https://www.databricks.com/blog/unified-data-discovery-business-context-unity-catalog , organized under the same domains and sub-domains you can use to structure your data estate. That way, the business context sits right next to the physical assets it describes, rather than in a separate tool. You can also codify the relationships that link a concept to the rest of your data ecosystem, including Unity Catalog assets, dashboards, and even other Pages. In the Related Assets section, you can catalog the specific workspace and Unity Catalog objects that underpin or illustrate the definition. In the Sources field, you can cite the authoritative links and internal objects the Page is derived from, so every definition stays grounded in verifiable evidence. You do not have to write Pages by hand. Genie Code https://docs.databricks.com/aws/en/genie-code/ , Databricks' data-smart AI coworker, can author Pages for you, drawing on a range of supported source material: Unity Catalog and workspace assets, file attachments, links, and MCP-connected tools like Confluence, Slack, Google Docs, or GitHub that you configure in Genie Code's MCP setup. Point it at the right sources, and it can extract and import your business knowledge and terminology in bulk, turning weeks of copy-pasting into a few minutes of conversation. For example, hand Genie Code one of your organization’s key Confluence pages, and it will pull out the concepts and jargon buried inside it and create each one as its own atomic Page. To get started on your first set of Pages, click the "Bulk import pages" conversation starter in Genie Code. Unity Catalog Pages give your enterprise a governed home for the concepts that matter most, and the tools to curate and collaborate on the authoritative meaning your organization relies on. As the latest addition to Unity Catalog semantics, the curated layer of Genie Ontology,, that meaning is served straight to Genie One and your agents, grounding them in consistent, trusted context and connecting your business logic to the data estate where your most impactful work happens. Pages is available today in Beta. Learn more in our product documentation https://docs.databricks.com/aws/en/uc-semantics/pages , and reach out to your account representative to try it out. Subscribe to our blog and get the latest posts delivered to your inbox.