Skip to main content
Back to News
news/AI Automation

Databricks Turns Genie Into an Always-On Enterprise Agent

Databricks relaunches Genie as an autonomous enterprise agent that runs long jobs unsupervised, grounded in a governed ontology and capped by Unity Gateway.

Stefan Trbojevic

Stefan Trbojevic

11 November 20263 min read
LinkedIn
Teal-lit colonnade representing an always-on enterprise agent layer

The takeaway

Genie's relaunch is a bet that context and governance, not raw model capability, decide whether autonomous enterprise agents ship or stall.

Why it matters for builders

Ontology-driven context, spend caps and permission inheritance are now the three architectural decisions that decide whether an autonomous agent survives enterprise review.

Databricks Turns Genie Into an Always-On Enterprise Agent

Databricks used Wednesday to relaunch Genie, its data assistant, as something considerably more ambitious: an autonomous AI coworker that takes ownership of long-running business jobs, checks in only when it needs input or approval, and keeps working after the chat window closes. The company announced the change in a blog post on November 11, positioning Genie as an always-on agent for teams rather than a question-and-answer interface on top of a lakehouse.

What actually changed

The core shift is from request to responsibility. Genie now runs on schedules and triggers, tracks progress on multi-step work, follows up on missing inputs and acts when new information arrives. Teams reach it through Genie One, email, Slack and Microsoft Teams,.

CEO Ali Ghodsi pitched it by showing TechCrunch his own usage instead of a canned demo: dissecting the weakest parts of an employee satisfaction survey and summarising a sales pitch that went badly. At 12,000 employees, he argues, he no longer hears bad news directly. "Anyone who has more than 150 employees, which is a lot of companies, a lot of people can get value out of this," he said, citing Dunbar's number.

The ontology is the real product

Autonomy without context produces confident nonsense, so Databricks put most of the announcement into the layer that supplies context. Genie Ontology maps definitions, metrics, relationships and organisational knowledge, largely built by agents from dashboards, queries and notebooks people already use, then curated by humans. Authoritative definitions live in Unity Catalog, and the OntoRank algorithm gives them precedence so every agent works from the same governed understanding.

Genie Search pulls results from Google Drive, SharePoint, Jira, Confluence and Workday alongside Databricks tables. MCP connectors let it read from and act in Slack, Zoom, Microsoft 365, Jira, GitHub and calendars, with custom MCP extending to ServiceNow and PagerDuty. Reusable workflows can be published as skills that live in Unity Catalog and run with the invoking user's permissions.

Glowing teal glass cube representing a governed context layer on a dark surface

The control problem, addressed head-on

Every always-on agent story hits the same two questions: what can it touch, and what does it cost. Databricks answers both with Unity Gateway, which lets admins control tool access, trace actions and set spending limits. Ghodsi says costs can be contained through usage caps and by swapping in cheaper open-weight models behind the gateway. The team-level Genie Agents now also read PDFs and slides with page-level citations, and can be shared externally through OpenSharing.

What it means for builders

Three things are worth copying. Ontology beats prompt engineering, because the durable advantage is a governed map of business definitions. Budget caps belong in the product, not the runbook, since unsupervised loops are expensive by default. And permission inheritance matters more than raw capability: skills that run as the calling user keep autonomy inside existing policy.

Share𝕏

The Automation Brief

Read 5 AI stories instead of 50.

The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.

No noise. Unsubscribe anytime.

Editorial notes

Reported by

Stefan Trbojevic

Edited by

n8n Lab Editorial

Published

11 November 2026

Updated

11 November 2026

AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.

n8n Lab is an independent service provider. We are not affiliated with, endorsed by, or sponsored by n8n GmbH. “n8n” is a trademark of n8n GmbH and is used here only to describe the platform-specific implementation and automation services we provide.