Databricks Genie Ontology: What's New, Availability, and Alternatives (2026)

Databricks announced the general availability of Genie Ontology on November 11, 2026. The announcement calls it "an agentic ontology that agents build from your teams' everyday work and people govern." The goal is to give every agent one understanding of the business.
In plain terms, Genie Ontology is a shared layer of business context for AI agents. It learns definitions from dashboards, queries, and notebooks. People then lock down the definitions that must be exact, such as revenue.
This guide covers what changed with the GA release, how the ontology works, what it costs, and who can use it. It ends with other ways to give AI business context, including tools that start from a spreadsheet instead of a data platform. Details come from Databricks' blog, documentation, and pricing page, read on November 16, 2026.
What Databricks announced
Databricks first introduced Genie Ontology at its Data + AI Summit in June. The November 11 post moves it to general availability and adds new sources and controls.
The post's summary lists the main parts of the release. It "introduces Genie Search for enterprise knowledge." It also "extends learned context to BI tools and organizational directories." Finally, it "adds Ontology Manager for curation and brings Unity Catalog Domains and Pages to General Availability."
The same week, Databricks also updated Genie Agents. A separate post says Genie Agents can now analyze unstructured files, connect to more tools, and be shared outside a company.
Databricks frames the whole week around one idea. Companies no longer run one agent for one team. They run many agents across many functions, and each one needs the same understanding of the business. The announcement argues that an enterprise ontology "should therefore be shared infrastructure." In its view, IT and data leaders should own it the same way they own the data platform.
How Genie Ontology works
The documentation calls Genie Ontology "the unified context layer that gives AI agents a business-aware map of your organization." It combines two kinds of knowledge.
Learned knowledge comes from work teams already do. Agents read Databricks dashboards, SQL queries, notebooks, and BI semantic models. They store what they find as Ontology Snippets. The documentation gives examples of snippets:
- Metric definitions. "An 'active user' is a distinct user, deduplicated across all platforms."
- Authoritative sources. For example, which curated agent should answer revenue questions.
- Business rules. "A 'qualified lead' only counts after a demo is booked."
Modeled knowledge is what people define on purpose. It lives in Unity Catalog as metric views, domains, and Pages. The documentation says "Genie treats modeled knowledge as more authoritative than learned knowledge, so it takes precedence when the two conflict."
A ranking step decides which context an agent sees. Databricks calls it OntoRank. The announcement says it weighs "source authority, certification, the expertise and relevance of the people behind the knowledge, popularity, and recency." It then hands the best context to the agent.
Permissions carry through. The documentation says snippets "are gated by Unity Catalog permissions, so users only see snippets derived from assets they're allowed to access."
One limit is worth knowing early. The documentation says "You can't edit learned knowledge snippets directly." To change what Genie learns, you update the source asset, or define the concept as modeled knowledge.
What's new in the GA release
The GA release adds several pieces. Some are generally available today, and others are in Beta or coming soon.
| Component | Status in the November 11 post | What it does |
|---|---|---|
| Genie Ontology | Generally Available | Shared business context for agents, learned and modeled |
| Ontology Snippets | Generally Available | Reusable pieces of context, such as calculations and rules |
| Genie Search | Generally Available | Search across documents, tickets, and wikis, with permissions |
| External knowledge sources | Beta | Learns from BI semantic models; Power BI today |
| Organizational directories | Generally Available | Imports people, teams, and reporting lines, starting with Workday |
| Unity Catalog Domains and Pages | Generally Available | Organize assets by business area and record official definitions |
| Ontology Manager | Coming soon in Beta | A visual interface to review and curate the ontology |
Genie Search reaches knowledge outside the lakehouse. The post says it "supports Google Drive, Jira, Confluence and Workday today, with SharePoint in Beta." It also says Genie Search "ingests each source's access controls alongside its content."
External knowledge sources bring in logic from BI tools. The post says "Power BI is supported today. Tableau and Sigma support is coming soon."
Unity Catalog Pages hold official definitions. Each Page has a named owner. When a question involves a defined concept, the post says "Genie treats the Page as the source of truth."
Ontology Manager is not available yet. Databricks describes it as "the first visual interface" for curating the ontology, coming soon in Beta.
The benchmark Databricks published
Databricks shared results from a benchmark built with Turing. It simulated a Databricks workspace for a fictional UK energy supplier. Genie, using Genie Ontology, was compared with two frontier agents labeled Agent A and Agent B.
| Agent | Pass rate | Average time per run | Average tokens per run |
|---|---|---|---|
| Genie with Genie Ontology | 91% | 100 seconds | 432K |
| Agent A | 45% | 435 seconds | 889K |
| Agent B | 58% | 307 seconds | 1.05M |
The post says Genie "answered correctly in 91% of runs, 1.6 times the rate of the stronger frontier agent." It also used less than half the tokens.
Read these numbers with care. The benchmark was built by Databricks with a partner and run in a simulated workspace. The post itself says the results "reflect an established deployment where teams have accumulated that knowledge." A new deployment with little learned context may perform differently.
Genie Agents got an update the same week
On the same day, Databricks posted an update to Genie Agents. These are the no-code agents that teams build for specific jobs. Three changes stand out:
- Unstructured data. Agents can now analyze PDFs, documents, slide decks, and images in Unity Catalog volumes. Answers can cite "the specific page of the source file." This is in Public Preview.
- MCP connectors. Native MCP connectors are now generally available. The post names "Slack, Google Drive, Google Calendar, Gmail, Glean, Atlassian, Github and Microsoft 365."
- Sharing outside the company. OpenSharing and Genie Clean Rooms, both in beta, let a provider share an agent with partners or customers.
The post gives a sales example. An agent reads pipeline data, pulls account plans from Google Drive, and posts a deal summary to Slack.
Genie Ontology connects these pieces. Every Genie Agent draws on the same governed context, so a metric defined once applies everywhere.
Who can use it and what it costs
Genie Ontology is part of the Databricks platform. You need a Databricks account with Unity Catalog, since permissions and modeled knowledge live there.
Agents outside Databricks can use it too. The announcement says that through Genie MCP, "the same context is available to agents outside Databricks, such as ChatGPT, Claude and Microsoft Copilot."
On cost, the Genie pricing page lists Genie Ontology as "Price coming soon." Its FAQ says "At this time, customers do not incur costs for Genie Ontology generation, storage or retrieval."
Using Genie itself is charged separately. Here is what the pricing page shows for US East.
| Item | What the pricing page lists |
|---|---|
| Genie Ontology | Price coming soon; no cost today for generation, storage, or retrieval |
| Genie Search | Price coming soon; you pay only for AI Functions at existing prices |
| Genie usage | 150 DBUs of free usage per user each month |
| Beyond the free usage | $0.070 per DBU, plus compute billed separately |
The page notes that "150 DBUs = $10.50 in US East; the dollar equivalent varies by region." It also says the free usage covers LLM usage only. Running queries and dashboards still uses compute, such as a SQL warehouse.
The FAQ answers one common question directly. It says "There are no seat-based or user-based fees." Admins can set budgets and alerts for Genie spend.
How to get started
Databricks suggests starting small. The announcement says organizations "can start with the context they already have." Here is a practical order, based on the documentation.
- Check your setup. Make sure your workspace uses Unity Catalog, since snippets follow its permissions.
- Let Genie learn. Snippets come from existing dashboards, SQL queries, notebooks, and Genie Agents. Clean up stale assets first, or restrict access to them.
- Define the critical metrics. Use metric views and Pages for numbers that must be exact, such as recognized revenue.
- Organize by domain. Group assets into domains so Finance and Marketing can keep their own definitions.
- Connect outside knowledge. Set up Genie Search for documents and tickets, and an external source such as Power BI if you use it.
- Check the citations. When Genie answers, open the snippets it used to see where each definition came from.
Step 2 matters more than it looks. The documentation says: "To keep stale or experimental assets out of answers, restrict access to them with Unity Catalog permissions."
What this means for data teams
Genie Ontology shifts some work from building agents to curating context. Instead of writing instructions into each agent, a team defines a metric once. Every agent then uses it.
That changes who owns the work. Data stewards and domain experts become curators. The announcement says Ontology Manager will let stewards "promote what agents have learned into authoritative Pages directly from a review queue."
It also raises a governance question. Learned knowledge comes from existing assets, including old or wrong ones. The quality of answers depends on the quality of those sources.
For teams already on Databricks, the GA release is worth testing. Pick a domain with clear definitions, such as sales. Compare Genie's answers with a report you already trust.
Cost planning also changes. Genie Ontology itself has no listed charge today, but agents that use it still draw on Genie usage and compute. Budget for the questions people will ask, not only for the ontology. Admins can set budgets and alerts, which helps during a pilot.
Alternatives for giving AI business context
Genie Ontology fits teams whose data lives on Databricks. Other approaches suit other setups. Each detail below comes from the vendor's own pages.
| Option | How it adds business context | Best fit |
|---|---|---|
| Databricks Genie Ontology | Learned snippets plus governed Pages and metric views | Teams on Databricks with many agents |
| Microsoft Fabric data agents | Instructions, examples, and data sources in OneLake | Teams on Fabric and Power BI |
| Power BI semantic models | Measures and relationships defined in the model | Teams whose logic already lives in Power BI |
| Powerdrill Bloom | Files and context saved in a workspace | Teams working from spreadsheets, PDFs, and SQL databases |
Microsoft Fabric data agents. Microsoft's documentation says you can "add organization-specific instructions, examples, and guidance to fine-tune the Fabric data agent." Supported sources include "a Power BI semantic model" and "an ontology." A paid F2 or higher capacity is listed as a prerequisite.
Power BI semantic models. These already hold measures and relationships. Genie Ontology can now learn from them through external knowledge sources, so the two can work together.
For more data-agent options, see our Databricks Genie alternatives guide.
Where Powerdrill Bloom fits
Not every team has a data platform to curate. Many work from exports, spreadsheets, and a single database.
Powerdrill Bloom starts from those files. You upload Excel, CSV, or PDF files, or connect a SQL database, and ask questions in natural language. The ChatBI page says you can "create interactive dashboards, insightful reports, and dynamic visualizations with simple natural language queries."
Context builds up in a workspace. Its changelog says workspaces "give your AI a persistent, memory-powered home." A file uploaded once can be used in every later chat in that workspace. Each workspace also keeps its own skills and MCP servers, so a finance workspace and a sales workspace stay separate. The AI SQL Chatbot handles questions against a database. The AI Report Generator turns the answers into a report.
The pricing page lists a Free plan at $0.00 and Pro at $13.27/mo.
Try Powerdrill Bloom with last quarter's sales export and ask for a revenue report by region.
Frequently asked questions
What is Databricks Genie Ontology?
It is a shared layer of business context for AI agents on Databricks. It combines knowledge that agents learn from dashboards and queries with definitions that people govern in Unity Catalog.
When did Genie Ontology become generally available?
Databricks announced general availability on November 11, 2026. It first introduced the feature at its Data + AI Summit in June.
How much does Genie Ontology cost?
The pricing page lists the price as coming soon. Its FAQ says customers do not currently pay for ontology generation, storage, or retrieval. Genie usage and compute are charged separately.
Can ChatGPT or Claude use Genie Ontology?
Yes, through Genie MCP. Databricks says the same context is available to agents outside Databricks, such as ChatGPT, Claude, and Microsoft Copilot.
Can I edit what Genie Ontology learns?
Not directly. You change learned knowledge by updating the source assets. You can also define a concept as modeled knowledge, which Genie treats as more authoritative.
Sources: Databricks, Announcing GA of Genie Ontology · Databricks documentation, Genie Ontology · Databricks, Genie pricing · Databricks, Genie Agents update · Microsoft Learn, Fabric data agent.