Data Governance · 5 min read

Governed Genie AI expands across Teams, Copilot, Excel

Databricks has made Genie for Teams and Genie Ontology generally available, extending governed Azure Databricks context to Copilot, Excel, Copilot Studio and Microsoft Foundry.

By Grace Kim986 words

Key takeaways

  1. The Databricks Genie app for Microsoft Teams is generally available.
  2. Genie Ontology, the enterprise context layer, is also generally available.
  3. A sample Teams question targets open leads worth more than $100,000.
  4. Microsoft Foundry supports Genie One, while Unity Gateway exposes Foundry models in Azure Databricks.
  5. Work IQ access over MCP is slated to arrive, with no release date supplied.
What's new for Genie in Microsoft: Copilot Cowork and Chat, Microsoft Teams, and Excel.
What's new for Genie in Microsoft: Copilot Cowork and Chat, Microsoft Teams, and Excel. — AI-generated

By Grace Kim

Databricks has made its Genie app for Microsoft Teams generally available, extending governed Azure Databricks answers to Teams, Copilot Chat, Copilot Cowork and Excel. The announcement offers no adoption figures, pricing or rollout dates, so breadth is the news, not scale.

The case is context, not another chat box. Most AI assistants can read email, files and calendars, yet cannot tell which revenue table finance certified or how net retention is defined. Databricks links that ambiguity to inconsistent outcomes, low adoption and low trust among business users.

Authors Kiriana Stukas and Toussaint Webb define Genie Ontology as “the enterprise context layer that gives Genie and other AI agents a shared understanding of your business.” The layer combines explicit metric views and certified definitions with context learned from dashboards, queries, documents and everyday work. Agents build it, people govern it, every agent can use it, and it is now generally available.

What can Teams do now?

The native app supports direct messages or @Databricks Genie mentions in channels and group chats, then replies in thread with governed context and source links. Admins add the app and set default response visibility; users sign in and choose a workspace on first use.

Genie One is the default and searches across the data estate. Teams can pin a conversation or channel to a specific Genie Agent through the Genie Agentic API, scoping answers to curated data and instructions. The app supports charts, file uploads and MCP actions, works in Microsoft 365 Copilot, and requires no Copilot Studio license.

Databricks uses a six-figure example: a sales manager asks, “Which open leads over $100K haven't had a follow-up this week?” Genie returns matching leads, latest activity and source links in the thread, replacing a manual sweep through pipeline reports, CRM records, meeting notes and reps.

For Private Link workspaces, an account admin creates a context-based ingress policy for Teams connectivity. Every question runs as the asking user, so Unity Catalog and source permissions restrict data access. Databricks processes each query inside the customer’s workspace; the service routing messages between Teams and Databricks stores neither questions nor answers.

How does Copilot use Genie?

The Azure Databricks Genie plugin lets Copilot and Copilot Cowork delegate questions to Genie One instead of searching independently. Setup takes a few clicks: enable the plugin, select Connect and sign in with Azure Databricks credentials. Databricks says users need neither an OAuth app registration nor manifest edits.

This lets Copilot use finance’s certified net-retention definition rather than infer meaning from table names. The plugin draws on Genie Ontology’s metric views, certified definitions and learned context, all governed by Unity Catalog.

The account-team example is ambitious. Cowork delegates a customer-meeting brief to Genie, which pulls usage, spend and account trends from Azure Databricks using certified metric definitions, then reads recent Outlook customer-email threads through the Microsoft 365 connector to flag open escalations.

Cowork produces a Word brief, an Excel overview with charts and a client-ready presentation; every figure comes from Genie. Tidy workflow. Same old definition risk.

What changes in Excel and Copilot Studio?

In the Azure Databricks Excel Add-in, Genie One answers plain-English questions with governed Azure Databricks data without requiring analysts to hunt for tables or write SQL. Every answer includes source citations. Analysts can ask follow-ups, resume conversations and load table results directly into a workbook for further modeling, reducing the export-copy-and-pray routine.

Builders can connect Genie Agents to Copilot Studio through the Azure Databricks connector for Power Platform. Adding Genie as a tool lets a Studio assistant delegate Databricks questions to agents using curated data and instructions under Unity Catalog governance.

Use the native Teams app by default; choose Copilot Studio for custom agents combining Genie Agents with tools outside Azure Databricks. The native route avoids an extra agent layer.

How are access, spend and tools governed?

As assistants and agents spread, Databricks identifies three control gaps: sensitive data reaching the wrong people, unpredictable AI spend, and agents connecting to tools nobody approved. Unity Catalog enforces each user’s data permissions. Unity Gateway gives administrators traces for monitoring, MCP policies and budget controls across the organization.

MCP policies determine which servers and tools Genie may use, who can access them and which actions need human approval. Traces record what Genie did and who asked. The controls matter more than the polished demos, though Databricks provides no benchmark proving they work.

Where does Genie run next?

Genie spans Copilot Chat, Copilot Cowork, Microsoft 365 apps, Teams, Excel, Copilot Studio, Power Platform and Microsoft Agent 365. In Microsoft Foundry, agents can use Genie One for governed Databricks data while acting as the signed-in user. The connection also runs backward: Microsoft Foundry models become available in Azure Databricks through Unity Gateway.

Genie will soon connect to Work IQ over Model Context Protocol, or MCP. That planned route would add email, meetings, documents, Teams messages, and people and organizational context beside governed Databricks data. No launch date is supplied.

Requests would run under the signed-in user’s Microsoft Entra ID identity, preserving Microsoft 365 permissions, sensitivity labels and compliance policies while Unity Catalog governs Databricks data.

What should data teams do?

Pilot the native Teams app for broad governed Q&A, reserving pinned agents for deliberately scoped domains. Certify metric definitions in Genie Ontology before exposing Copilot, then test Unity Catalog permissions, Gateway MCP policies, approval rules, budget limits and traces with sensitive roles. Use the Excel add-in when analysis ends in a workbook, and Copilot Studio when custom orchestration warrants another layer.

Treat Work IQ as roadmap, not capability. Databricks provides no adoption metrics, price, performance benchmark or Work IQ release date, leaving the verdict unchanged: the company has broadened Genie’s reach, while each answer still rests on the definitions and permissions underneath it.

docs.databricks.com (Original)

Price tags

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering companies, deals and strategy in Analytics Digest is a professional daily at Analytics Digest.

18 articles

Read also

« Previous articleNext article »