Databricks adds a ChatGPT plugin for querying governed business data

ChatGPT users can now ask questions of business data through Databricks Genie rather than start their analysis inside Databricks. Announced on November 11, 2026, the new plugin connects the assistant to Genie’s governed data service. Databricks says Genie interprets business terms, writes and runs database queries, and returns answers with source citations.

The ChatGPT connection is part of a coordinated release of native Genie plugins for Claude, Microsoft Copilot Cowork and Grok Bot. The plugins use Genie’s MCP server, the connection through which external assistants reach its business context. Agents without a native plugin can also use that server directly.

Giving the assistant a business map

Databricks frames the release as an answer to a problem with connecting AI agents to more company information: access alone does not tell an assistant where the right answer lives.

Its proposed solution is Genie Ontology, a shared layer of business definitions, relationships and source context. That layer supplies meaning as well as data—for example, the company’s agreed definition of an active user. Databricks says organizations can maintain those definitions once and make them available across assistants connected to Genie.

An assistant can hand a business question to Genie for investigation. Or it can keep responsibility for a larger assignment, ask Genie for missing information, then continue using its own tools to edit files, draft plans or post to Slack.

Genie does the query work

For a handed-off question, Genie identifies relevant data and writes and executes SQL, the language used to query databases. Databricks says retrieval stays within information the user is allowed to see. Questions needing specialized analysis can go to domain-specific Genie Agents, such as marketing or finance.

The scope extends beyond tables: Databricks says Genie can find relevant information across documents. For clients supporting the MCP Apps extension, it can also return interactive visualizations and Ontology citations inside the conversation.

OpenAI’s plugin listing suggests several requests:

• Investigate a metric change: identify the main drivers, show the query path and note caveats.
• Draft and review SQL: explain joins and filters, and list checks for missing or unexpected values.
• Explain a dashboard: summarize changes against the previous period, likely drivers and follow-up checks.

Access depends on the workspace

OpenAI cautions that availability depends on the plugin, the user’s plan and workspace settings. Databricks also says platform releases are staged: an account might not receive an update until a week or more after its initial release date.