---
type: "post"
id: "genie-ontology-now-generally-available-supplies-the-structured-e97e90"
slug: "genie-ontology-now-generally-available-supplies-the-structured-e97e90"
title: "Genie Ontology, now generally available, supplies the structured business context behind the agent."
description: "Databricks says Genie Ontology is generally available as a shared layer of enterprise context that agents can learn from and people can govern."
retrieval_nugget: "Databricks says Genie Ontology is generally available as a shared layer of enterprise context that agents can learn from and people can govern. The release adds enterprise search, broader learned context and curation tools; a company-built benchmark reports higher task pass rates and lower token use when Genie draws on that context."
published_at: "2026-11-12"
updated_at: "2026-11-12"
record_date: "2026-11-12"
date_kind: "published_at"
topics: ["mcp"]
entities: ["Databricks","Genie Ontology","Genie Search","Genie MCP","Unity Catalog","Turing"]
related_patterns: ["company-memory-needs-write-loops"]
novelty_label: "new_fact"
what_changed: "Databricks made Genie Ontology generally available, adding governed business context for agents, enterprise knowledge search, curation tools and benchmark results."
prior_refs: []
source_url: "https://www.databricks.com/blog/announcing-ga-genie-ontology"
source_urls: ["https://www.databricks.com/blog/announcing-ga-genie-ontology"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:genie-ontology-now-generally-available-supplies-the-structured-e97e90"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/genie-ontology-now-generally-available-supplies-the-structured-e97e90/","markdown":"https://newruntime.com/posts/genie-ontology-now-generally-available-supplies-the-structured-e97e90.md","json":"https://newruntime.com/posts/genie-ontology-now-generally-available-supplies-the-structured-e97e90.json"}
markdown_url: "https://newruntime.com/posts/genie-ontology-now-generally-available-supplies-the-structured-e97e90.md"
---

# Genie Ontology, now generally available, supplies the structured business context behind the agent.

## Retrieval answer

Databricks says Genie Ontology is generally available as a shared layer of enterprise context that agents can learn from and people can govern. The release adds enterprise search, broader learned context and curation tools; a company-built benchmark reports higher task pass rates and lower token use when Genie draws on that context.

## Summary
Databricks says Genie Ontology is generally available as a shared layer of enterprise context that agents can learn from and people can govern. The release adds enterprise search, broader learned context and curation tools; a company-built benchmark reports higher task pass rates and lower token use when Genie draws on that context.

## What happened
Databricks announced the general availability of Genie Ontology, an “agentic ontology” intended to give agents a consistent understanding of an organization’s business. The company says agents build much of its knowledge from existing work, including Databricks dashboards, SQL queries, notebooks and semantic models in tools such as Power BI. They capture reusable context as Ontology Snippets—calculations, business rules and data relationships—and refresh those snippets as the underlying work changes. Each snippet remains tied to its source asset and inherits that asset’s permissions.

The system combines this learned context with definitions that teams model and govern through Unity Catalog semantics. Databricks lists Metric Views for certified KPIs, Domains for organizing knowledge by business area, and Pages for authoritative definitions of concepts. Its OntoRank algorithm weighs factors including source authority, certification, expertise, relevance, popularity and recency to choose context for a question. The company says Genie MCP makes the same context available to agents outside Databricks, including ChatGPT, Claude and Microsoft Copilot.

The release also makes Genie Search generally available. It lets people and agents search enterprise documents, tickets and wikis alongside governed Databricks assets, returning synthesized answers with citations. Supported sources today are Google Drive, Jira, Confluence and Workday; SharePoint is in beta. An administrator connects sources and enables indexing. Databricks says the service ingests source access controls through Unity Catalog connections and enforces them before returning results. For real-time work, Genie Search can work alongside MCP connectors and SQL.

Other availability changes include Ontology Snippets, organizational directories beginning with Workday, and Unity Catalog Domains and Pages. External knowledge sources that import business logic from external BI platforms are in beta, with Power BI supported today; Tableau and Sigma are described as coming soon. Ontology Manager, a visual interface for assessing coverage and curating knowledge, is coming soon in beta. The announcement does not state a separate price for Genie Ontology or Genie Search.

Databricks and Turing built a benchmark around a simulated workspace for a fictional UK energy supplier. Each of three agents answered the same questions five times per task, with answers graded against a rubric defined before the runs. Genie with Ontology had a 91% pass rate, 100-second average wall time and 432,000 average tokens per run. Agent A recorded 45%, 435 seconds and 889,000 tokens; Agent B recorded 58%, 307 seconds and 1.05 million tokens. Databricks says the established deployment behind the benchmark had accumulated workspace knowledge that the frontier agents had to rediscover.

## What the archive adds
New Runtime’s earlier coverage of Glean’s Agent Identity, dated 2026-10-24, described agents using their own governed accounts rather than a person’s credentials ([earlier note](https://newruntime.com/posts/glean-makes-agent-identity-generally-available-so-agents-808423/)). Genie Ontology addresses a different layer: shared business context and definitions. This is a new product event that connects context governance to the broader enterprise-agent governance question.

## Trend signal
This supports the pattern that company memory needs write and correction loops: context here is learned continuously from work, then curated by people into authoritative definitions. The announcement describes the mechanism, but does not independently demonstrate how well it handles incorrect or stale learned context. ([Company memory needs write loops](https://newruntime.com/patterns/company-memory-needs-write-loops/))

## Why now
Databricks says organizations are moving from one agent in one domain to many agents across functions, each of which needs accurate and consistent business context. The company frames Ontology as shared infrastructure rather than knowledge recreated separately inside each agent. This is the rationale Databricks gives for the general-availability release and its expanded sources and curation controls.

## Why it matters
For teams, the product’s central proposition is to separate reusable business meaning from individual agent prompts or implementations. Definitions such as a certified KPI can be governed once, while learned context from everyday work can cover the long tail that would be expensive to document manually. If this works as described, a team can give multiple agents the same definitions, source permissions and citations rather than maintaining parallel, potentially conflicting context.

The practical trade-off is platform dependence: much of the documented flow uses Databricks, Unity Catalog and its integrations, even though Genie MCP extends access to outside agents. Teams should test whether their most consequential business rules are represented, whether the source permissions are preserved, and whether citations let reviewers trace answers to evidence. The benchmark is encouraging as a vendor-reported result, but its setup is not enough to predict performance on another company’s data or workflows.

## Limits and uncertainty
The benchmark is a Databricks-and-Turing exercise in a simulated workspace for one fictional energy supplier. The announcement does not establish how the results transfer to other industries, workspaces, query types or less mature deployments, nor does it report independent replication. It also does not quantify the effort required to connect sources, curate definitions or keep permissions and learned context correct as systems change. External BI sources remain in beta, some announced integrations are forthcoming, and Ontology Manager is not yet generally available. Databricks does not disclose a separate price in the announcement.

## Skill to build now
Pick a small set of recurring business questions whose answers depend on definitions spread across dashboards, SQL and documents. For each, record the authoritative definition, source assets and who may access them. Once the relevant Genie capabilities are enabled, compare answers with and without the shared context, checking correctness, citations, permissions and token use. That gives your team a local baseline instead of treating the vendor benchmark as a forecast.

## Reader takeaway
Genie Ontology is Databricks’ GA attempt to make enterprise context reusable across agents: automatically learned from work, ranked for authority and access, and supplemented by definitions people govern. The useful question for builders is not whether one vendor benchmark looks strong, but whether the same context layer improves verified answers in their own workflows.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Mcp](https://newruntime.com/topics/mcp.md): Explore the mcp topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)