# Data + AI Summit 2026: Genie One, business semantics and controls for agents

> At Data + AI Summit 2026 Databricks made Genie One, Genie Agents and Genie Code generally available, added domains and a business glossary to Unity Catalog, extended Agent Bricks and put guardrails, spend controls and tracing for agents into Unity AI Gateway. Here is everything announced, with release stages.

By Brandon Gross. Published June 18, 2026.  
Tags: databricks, genie, agents, data agents, semantics, unity catalog, mcp.  
Canonical: https://kenseme.com/blog/databricks-data-ai-summit-2026-genie-one/

Data + AI Summit 2026 ran in San Francisco the week of June 15, and Databricks made one argument all week: business users, engineers and agents should work from the same governed data and the same business definitions. I agree with the argument. What stood out to me is how much of the stack Databricks wants to own to get there.

The announcements fall into five groups: Genie, business semantics, the agent platform, governance and control over agents, and the data platform underneath.

Stages below are as Databricks stated them. Where the press release and the documentation disagree, both are given. Many items were announced with no stage, and that is said rather than assumed.

## Genie

The naming moved twice this spring. Databricks One, the interface for business users, was renamed Genie in April, then Genie One on June 9. Capabilities did not change with either rename.

1. **Genie One**, generally available. An assistant for business users on web and mobile that answers questions over Databricks data and connected apps, writes documents and reports, and runs schedules and alerts. Chat and documents in Genie One are generally available; the iOS and Android apps are in public preview per the documentation.
2. **Genie Agents**, generally available. They replace Genie spaces. A saved conversation becomes a reusable agent that keeps its data sources and instructions, can take actions and can reason over unstructured content.
3. **Genie Code**, generally available. An agent for data engineering, machine learning and analytics work. A full-page view for running several threads in parallel is in beta. Genie Code can read MLflow runs, check serving endpoints and move GPU jobs to AI Runtime. Scheduled tasks that run without supervision are coming soon. An auto-approve mode lets it approve tool actions itself, with a classifier blocking risky ones.
4. **Genie Ontology**, in public preview since June 4 per the documentation; the summit press release gave no stage. It builds and ranks a map of business knowledge from data, dashboards, notebooks, pipelines and workplace apps. See [our earlier post](/blog/databricks-genie-ontology-preview/).
5. **Genie ZeroOps**, private preview in the coming weeks. A background agent that watches jobs, pipelines, tables and ML workloads, finds root causes and proposes fixes for a person to approve.
6. **Genie App Builder**, private preview shortly after the summit. Builds apps on governed data from a plain-language description.
7. **User skills**, in public preview. Personal, reusable tasks that extend Genie One chat.
8. **Thinking traces in agent mode**, in public preview. They show which definitions Genie applied to reach an answer.
9. **Connectors.** Google Drive, Jira, Slack, Confluence, SharePoint and others; Databricks cited more than 50 apps and systems. No stage stated.
10. **Where Genie shows up.** A managed MCP server for Genie One (beta), custom MCP servers inside Genie One (public preview), a Slack app (public preview), a Microsoft Teams app (beta), and a connection to Microsoft 365 Copilot Cowork through the managed MCP server (beta).
11. **Sharing Genie Agents outside the company**, in beta, as point-in-time snapshots through OpenSharing.
12. **Pricing and budgets.** No per-seat price: each user gets up to $10 of usage free per month, then pays for use. Account admins can set Genie budgets and alerts.

## Business semantics

Unity Catalog Business Semantics, built on metric views, became generally available on April 2 ([our post](/blog/databricks-metric-views-ga-business-semantics/)). The summit added the pieces around it.

1. **Domains**, in public preview. Assets are grouped into business-aligned categories so agents get context scoped to one part of the business. AI-suggested domains are coming later in preview.
2. **Business glossary**, preview coming soon. Business terms and taxonomies, defined or imported, linked to data.
3. **Metrics updates.** Databricks described metric views as ready for the Open Semantic Interchange and headed for open source in Apache Spark and Unity Catalog OSS. Multi-fact relationships in dashboards and metric materialization are in public preview. Importing Power BI and Tableau models is in beta.
4. **Simpler metric view syntax**, in public preview.
5. **Relationships between dashboard datasets**, in private preview.
6. **Certified assets first.** Genie One now ranks certified tables and metrics ahead of uncertified ones. No stage stated.

## Agent platform

1. **Agent Bricks** gained an agent memory service built on Lakebase, traces stored in the lakehouse, MCP connections managed through Unity Catalog and agents hosted on Databricks Apps. No stages stated.
2. **Databricks Sandbox.** Isolated virtual machines for running agent code, with narrowed Unity Catalog access. No stage stated.
3. **Omnigent**, an open-source (Apache 2.0) layer for combining coding agents such as Claude Code and Codex, setting cost and permission policies and sharing sessions. The open-source project is in alpha and the managed version in beta.
4. **Supervisor Agent** can now use Unity Catalog tables, vector indexes, volumes and web search as tools, and can nest other agents.
5. **Managed MCP services** for Google Drive, Jira, Confluence, Slack, GitHub and SharePoint, with a central inventory of approved MCP services. No stage stated.
6. **Databricks Apps.** App Spaces apply shared policies to a group of apps and Serverless Micro Apps scale to zero; both are private previews coming soon. Horizontal scaling is in beta; app telemetry to Unity Catalog is in public preview.
7. **AI functions in SQL**, `ai_query`, `ai_extract` and `ai_classify`, generally available.
8. **Vector Search is now AI Search**, adding full-text indexes that need no embeddings.
9. **Lakebase**, the Postgres service, added cross-cloud and cross-region disaster recovery, Git-style branching and agents that run database operations. Databricks also announced LTAP, transactional and analytical work on one copy of the data, coming soon as part of Lakebase.

## Governance, audit and control over agents

1. **Unity AI Gateway** as the runtime control point across models, agents, MCP services and tools. Announced with it, all with no stage stated: AI spend controls (budgets by user, team or tool, hard caps and routing), a single governed trace of model and MCP tool activity, and approval steps for coding agents before actions such as pushing code. Contextual service policies, which screen requests and responses for personal data, prompt injection, jailbreaks and unsafe content, are in beta. Integrations with Okta, Ping, SailPoint and Saviynt are upcoming.
2. **Attribute-based access control.** Tag-driven row filters and column masks became generally available in April, now running as the querying user. Automatic data classification is generally available. Grants on models by tag are in beta. Cross-engine enforcement, so outside engines reading Unity Catalog tables get the same policies, is in beta. Identity and context attributes are in preview soon and tag propagation is in private preview.
3. **Identity.** Automatic identity management with Entra ID is generally available on AWS and Google Cloud; Okta is in public preview. User-to-machine OAuth for service principals is in beta. Context-based ingress, zero-trust access rules for Genie, apps and AI Gateway endpoints, is in public preview.
4. **Monitoring.** Admins can monitor Genie chat prompts (beta). Telemetry from custom model serving endpoints is written to Unity Catalog tables (generally available).
5. **Model access.** Unity Catalog permissions on hosted foundation models are generally available.
6. **Lineage.** External lineage, adding assets outside Databricks to the lineage graph, is generally available.
7. **Recovery.** Managed disaster recovery for Unity Catalog metadata and data is in public preview, as is a configurable recovery window for dropped tables.
8. **Sharing.** Delta Sharing was renamed OpenSharing and moved to the Linux Foundation, now covering agent skills and models. Sharing to any Iceberg REST client is generally available.

## Data platform

1. **Lakehouse//RT**, in beta. Millisecond queries directly on Delta and Iceberg tables under Unity Catalog.
2. **Lakeflow Designer**, the visual data preparation tool, generally available.
3. **AUTO CDC in Databricks SQL**, automatic row expiry and conversion of partitioned tables to liquid clustering, all generally available.
4. **Iceberg** managed and foreign tables and v3 features in Unity Catalog, generally available.
5. **Databricks Runtime 19** on Spark 4.2, in beta, and new Lakeflow Connect connectors in beta.
6. **Azure Databricks.** Querying OneLake through Unity Catalog is generally available, managed Delta tables in OneLake are in public beta and an Excel add-in is in public preview.
7. **Panther.** Databricks agreed to acquire the security operations platform, subject to closing.

## What I take from it

Databricks is building every layer an agent needs inside one platform: the definitions (metric views, domains, glossary, Genie Ontology), the agents (Genie One, Genie Agents, Agent Bricks) and the controls (AI Gateway policies, budgets, traces and approvals).

The governance announcements tell you the most. An approval step before a coding agent pushes code, a hard cap on what an agent can spend and one trace of every tool call are controls companies already put on employees. Agents get the same leash as a new hire, and I expect every platform to copy it.

The open question is reach. Metric views are headed for Apache Spark and the Open Semantic Interchange, and Genie now answers inside Microsoft 365 Copilot and Slack through MCP. That helps. But most companies run more than one data platform, and a definition that lives inside one of them still leaves agents everywhere else guessing.

## Sources

1. [Databricks launches Genie One (press release, June 16)](https://www.databricks.com/company/newsroom/press-releases/databricks-launches-genie-one-all-new-agentic-coworker-every-team)
2. [Introducing Genie One, Genie Ontology and Genie Agents (Databricks blog, June 16)](https://www.databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents)
3. [What's new in Genie Code (Databricks blog, June 17)](https://www.databricks.com/blog/whats-new-genie-code-data-ai-summit-2026)
4. [Introducing Genie ZeroOps (Databricks blog, June 16)](https://databricks.com/blog/introducing-genie-zeroops)
5. [What's new in Unity Catalog at Data + AI Summit 2026 (Databricks blog, June 16)](https://www.databricks.com/blog/whats-new-unity-catalog-data-ai-summit-2026)
6. [AI governance at Data + AI Summit 2026: Unity AI Gateway (Databricks blog, June 16)](https://www.databricks.com/blog/ai-governance-data-ai-summit-2026-whats-new-unity-ai-gateway)
7. [Agent Bricks at Data + AI Summit 2026 (Databricks blog, June 16)](https://www.databricks.com/blog/agent-bricks-dais-2026)
8. [Introducing Omnigent (Databricks blog, June 13)](https://www.databricks.com/blog/introducing-omnigent-meta-harness-combine-control-and-share-your-agents)
9. [Governed apps on Databricks (Databricks blog, June 16)](https://databricks.com/blog/enabling-governed-vibe-coding-enterprise-apps-databricks)
10. [Platform security and compliance at Data + AI Summit 2026 (Databricks blog, June 17)](https://www.databricks.com/blog/whats-new-databricks-platform-security-and-compliance-data-ai-summit-2026)
11. [What's new in the AI platform (Databricks blog, June 17)](https://www.databricks.com/blog/whats-new-ai-platform-agents-ml-engineering-our-deep-learning-platform-and-new-capabilities)
12. [OpenSharing and Marketplace (Databricks blog, June 15)](https://www.databricks.com/blog/announcing-new-opensharing-and-marketplace-capabilities-ai-era)
13. [Databricks launches LTAP (press release, June 16)](https://www.databricks.com/company/newsroom/press-releases/databricks-launches-ltap-first-lake-transactionalanalytical)
14. [Databricks launches Lakehouse//RT (press release, June 16)](https://www.databricks.com/company/newsroom/press-releases/databricks-launches-lakehousert-bring-real-time-analytics-directly)
15. [Databricks agrees to acquire Panther (press release, June 16)](https://www.databricks.com/company/newsroom/press-releases/databricks-agrees-acquire-panther-further-establishing-security)
16. [What's new in Azure Databricks (Databricks blog, June 16)](https://www.databricks.com/blog/unifying-data-and-governance-agentic-era-whats-new-azure-databricks)
17. [AI/BI release notes 2026 (Databricks documentation, entries through June 18)](https://docs.databricks.com/aws/en/ai-bi/release-notes/2026)
18. [Platform release notes, June 2026 (Databricks documentation, entries through June 18)](https://docs.databricks.com/aws/en/release-notes/product/2026/june)