Enterprise

Databricks pitches always-on Genie for business monitoring—but access is still forthcoming

The proposed agent could flag revenue and campaign changes for review, but Databricks says preview and beta access is still coming.

Photo: Wikimedia Commons / Agrawroh, CC BY-SA 4.0 — Databricks red stacked-layer symbol above the company wordmark

Databricks announced its second-generation, always-on Genie agent on November 11, 2026, promising background business monitoring. But it isn’t generally available: preview and beta programs are forthcoming, without an access date.

Users would set goals and boundaries; Genie would delegate to subagents, chase missing inputs and request human decisions through a priority inbox and email. Running on its own compute, scheduled or triggered, it would track progress. Each alert would arrive as an unread email, with updates threaded together.

Databricks lists revenue tracking, customer-usage monitoring and marketing-content recommendations as potential assignments. Actions would remain auditable, subject to Unity Catalog permissions and Unity Gateway approval rules and budgets. Administrators could restrict tools.

Existing Genie and Genie Ontology are generally available; Genie Search and Ontology Manager aren’t yet. Genie Ontology supplies definitions, metrics, people and process context learned from dashboards, queries, notebooks and external business-intelligence tools. An organizational connector adds reporting relationships. Ontology Manager would flag frequently asked concepts needing refined definitions.

The release lists 100+ connectors for ingestion, indexing in place or live queries through MCP, including Google Drive, Jira and Confluence. Genie can generate slides, dashboards and interactive reports that stay current with governed data.

A separate Databricks plugin announcement describes native connections in ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot; other agents can use the MCP server. Genie can route pipeline-attribution questions to a specialized marketing Genie Agent, write and run SQL over permitted data, and return cited answers. Clients supporting the MCP Apps extension also receive interactive visualizations and ontology citations.

TechCrunch reported that co-founder and CEO Ali Ghodsi demonstrated Genie conversations covering a poor sales pitch and features the prospect required before buying. In Databricks’ release, Alan Khalil, associate director of data science and engineering at HelloFresh, said Genie reduced the time to identify winning Meta ads by 75%. That doesn’t establish outcomes for the forthcoming agent.

Databricks describes consumption-based, not per-seat, pricing, with team or user spending limits through Unity Gateway. It provides no numerical price for the new agent.

For revenue leaders at small B2B teams already using Databricks, the possible application is monitoring pipeline and marketing-attribution signals, then investigating stalled deals or campaign shifts sooner. That’s an inference, not evidence of an SMB-ready acquisition product.

As with scheduled sales-pipeline analysis, the evaluation questions are practical: are business definitions sound, are sources visible, and do uncertain findings reach a human? Access to the proposed monitoring agent is still ahead.

Sources