Databricks 升级 Genie One 为企业 AI 协作中枢并开放跨工具行动能力

Databricks：Blog·2026-11-11 23:00· 1 小时前AI 评分55

Genie One: The AI cowork home for everyone

AI 导读

Databricks 宣布 Genie One 成为面向企业的 AI 协作中枢，统一搜索、构建成果物与跨工具行动，并由 Unity Gateway 治理。

正文 · 原文

In today’s enterprise, work rarely stays in one application. A single task can involve reconciling a ticket in one system, updating a document in another, and nudging a colleague in a Slack thread while you’re waiting on both. Many of these tools have added their own AI assistants to help you automate and enhance the work you do within them, but they each can help with only its part of the overall task. Without the context of your full workflow and the knowledge spread across the rest of your enterprise, these assistants may speed up individual steps, but you’re still left connecting the pieces and finishing the job by hand.

Earlier this year, we introduced Genie One to help you find insights and automate work. Today, Genie One becomes the enterprise home for AI cowork, where every conversation, artifact, and workflow draws on your enterprise context from Genie Ontology and is governed through Unity Gateway. We also announced today that Genie can now take more autonomous action on your behalf, acting as a true always-on coworker for workflows across teams and tools. This release combines one place for you to find information, build finished work, and act across tools with one unified control plane for admins to govern that work as adoption grows.

One home for AI cowork

Genie One now brings the tools you use to find information, reach colleagues, build finished work and take action into one place.

Genie Search

Genie Search is a new kind of hybrid AI search that unifies context-aware chats, live information retrieval, and enterprise search into one experience. Genie Search starts by creating an index of the knowledge your organization keeps outside Databricks, in documents, tickets and wikis across tools like Google Drive, Jira, Confluence, Workday, and SharePoint. When you ask a question, Genie retrieves from the index to answer quickly and accurately, with fewer tool calls and fewer tokens even while drawing on multiple sources.

Simultaneously, Genie Search supports over 100 MCP connectors, which let Genie retrieve live information and take action, such as checking a Jira ticket’s status or creating one. Genie chooses how to use search, MCP, and SQL for each task based on your query. Indexed content also feeds OntoRank, the method Genie Ontology uses to weigh which sources to trust, so search improves the context behind every answer.

The same index now also powers asset search for items that live outside of Databricks. Sometimes you don’t need a composed answer—just the latest all-hands slides or the ticket for a bug you’re tracking. From the same chat bar where you ask questions and collaborate with Genie today you can now search directly. Results from Google Drive, Jira and Confluence appear alongside Databricks dashboards, tables and Genie Agents, so users can find documents and data in one place.

Organizational Directories

Genie Search also brings organizational data into both a visible organizational directory and Genie Ontology. In Genie One, you can browse team structures, find the right colleague, and understand reporting relationships. The same data gives Genie Ontology context about people and teams, helping Genie identify more authoritative and relevant sources for its answers.

Customize your homepage

Admins and users now also have more flexibility over the homepage experience in Genie One, including the ability to pin specific assets such as dashboards, apps, or Genie Agents. Admins can set company-wide pins to bring key updates and resources to the surface, while users can set individual pins based on what they want quick access to from Genie One.

Artifacts: from answer to finished work

Of course, getting the answer from AI is only half the job. Turning it into a shareable deck, dashboard, or page usually means moving to another tool, and losing context along the way.

Genie One now creates artifacts, including slide decks, dashboards and interactive pages directly in your conversation. You describe the artifact you need, then refine it in collaboration with Genie using the same governed data and context as the conversation. You can inspect the sources behind each number or visualization, check the results, and make changes before sharing, without ever leaving Genie One.

Artifacts also work with scheduled tasks and skills. Scheduled tasks keep the data in an artifact up to date, while skills apply company brand standards.

“With artifacts in Genie, we have full creative control over how we present our data. We can get the exact look and feel we want while keeping everything connected to our governed data — it's the best of both worlds."- Guilherme Vivian, Data Manager, Agibank

Take action across your tools

Once you’ve worked with Genie to generate insights and artifacts, you can complete your workflow by updating connected tools via MCP connections. Genie One can read and update systems such as Jira or Asana, so the result of an analysis reaches the system of record without you leaving the conversation. Ask Genie to summarize last quarter's open escalations in Jira and post themes to a Slack channel, or to prep for your meetings and find more focus time in your week. You can choose what actions require your approval, and admins can configure which tools they allow their team to connect Genie to.

"Our current safety software is great, but it's limited to what the investigator knows at the time. Genie integrates that safety information with delivery tickets, maintenance records, hours of service, HR, Dispatch, customer information and even the weather! We now get deeper and more complete information about a particular incident and company-wide trends then we ever could before." - Jeff Emrick, Corporate Safety Director, Ozinga

Share skills that work

Skills let you capture a workflow once and share it with teammates. Every business team has people who get strong results with AI, whether they're automating a weekly pipeline review or a month-end variance summary. With skills, the rest of the team can run the same steps and get consistent results.

You publish Genie skills to Unity Catalog, where colleagues can find and use them. Each skill runs with the permissions of the person using it, so it accesses only the data and tools that person is authorized to use. Teams can also bring skills built in other AI tools into Genie One and keep their workflows in one place.

Work with Genie from anywhere

Employees move between devices and collaboration tools throughout the day. Genie One is now available through native apps and the AI tools teams already use, with the same governed context and permissions wherever they work.

Genie apps for every surface

• Browser extension: Open Genie One with every new Chrome tab. Install the extension from the Chrome Web Store, and opt in to make Genie One your new-tab page.
• Desktop apps: Keep Genie One alongside the desktop apps you use throughout the day, with native apps for macOS and Windows. Ask data questions, work with threads and documents, use Genie Agents and Unity Catalog tables, and run Databricks Apps. Desktop features include back and forward navigation, an unread-activity badge, and automatic updates.
• Mobile apps: Use Genie One on iOS and Android to ask questions, explore and chat with dashboards, and run Databricks Apps. Push notifications let you know when scheduled work is ready, and MCP connections configured on the web carry over to mobile.
• Slack and Microsoft Teams apps: Message Genie directly or @mention Genie in a channel or group chat. Shared answers give teammates context they can build on, and channel owners can pin a Genie Agent for answers tailored to a specific use case. Genie preserves each user’s identity and permissions, so Genie responds based on the data you are authorized to access. Scheduled work, Skills, and MCP connections configured on the web carry over.

“With Genie One embedded in the flow of work, questions no longer die on the table. An executive can ask something they never would have submitted as a request and get a confident answer before the moment passes.”- Brian Muehlenkamp, Sr. Product Manager, Demand Intelligence, Ticketmaster

Genie MCP plugins for the tools you already use

In addition to the native Genie One experiences inside of Databricks, you can also access insights from Genie inside any AI cowork tool of your choice. Genie plugins bring Genie into ChatGPT, Claude, Microsoft Copilot Cowork, and Grok Bot. If you prefer to work in those tools, you get answers grounded in Genie Ontology, with the same Unity Catalog permissions.

Scale confidently with Unity Gateway

However independently an AI coworker acts, people remain accountable for its actions, and that accountability is what lets organizations scale AI confidently. Unity Gateway extends governance from data and context to the actions Genie takes, so admins can see what Genie did, who asked for it and why.

Traces and monitoring

Unity Gateway traces show admins how Genie responded to each request. A trace captures the question, answer, cited objects, response time and actions taken, from running SQL and calling a Genie Agent to publishing a dashboard, scheduling a task or sending a notification. Like audit logs, traces are stored in a table that admins query with their own compute, so they can filter records, join them with other data and build monitoring views.

When a user reports a problem, an admin opens the trace to review the question, SQL, cited objects and thumbs-down feedback. The admin can investigate without a screen share or asking the user to retrace every prompt. Traces also show how Genie is used across the organization, so admins can understand trends and diagnose anomalies over time.

MCP policies, permissions, and cost management

Beyond tracing individual actions, admins can set policies, enforce permissions, and manage costs across Genie One.

• MCP policies: In Unity Gateway, admins control which MCP servers and tools Genie uses and which actions require human approval.
• Budgets and cost management: Admins can also use it to review usage, manage budgets, and forecast spending. Genie One is free for users through January 31, 2027. More detailed cost controls will be available before then, giving admins time to plan.

Permissions and access requests

Genie One and Genie Ontology respect Unity Catalog permissions and each connected source's access controls, so users see only information they're authorized to access. When a question needs data a user can't access, Genie can identify the relevant table or asset and prepare an access request for the user to review and submit. Admins review the request in Unity Catalog and grant or decline it. Users get a direct path to the data they need, and access decisions stay with admins.

Scale AI cowork with confidence, wherever your teams work

Genie One gives you one place to find information, create finished work, and take action across tools. Unity Gateway gives admins the visibility and controls to govern that work as adoption grows.

Get started

Visit our documentation for up-to-date availability of the features outlined above, or contact your Databricks account team for more information.

• Admins: Enable Genie tracing in Unity Gateway, review MCP policies, and connect your first sources through Genie Search. 
• Business users: Open Genie One, search across your tools, and ask it to build your next dashboard or deck. Save what works as a skill and share it with your team. For more inspiration, visit our Genie at Work demo library. 

To learn more about taking action on your unified context layer, register for Data and AI Live: Context is just the beginning with Ali Ghodsi and Greg Brockman.

FAQs

What is Genie One?

Genie One is the Databricks enterprise home for AI cowork, where business users search company knowledge, analyze data, build finished work and take action across their tools. It draws on business context from Genie Ontology and is governed by Unity Gateway.

Can Genie automate business work?

Yes. It combines an autonomous agent, MCP connections, scheduled tasks and shareable skills to carry out multistep work across tools, like summarizing Jira escalations and posting the themes to Slack. You choose which actions need your approval.

Can Genie search my documents, tickets and data at once?

Yes. Genie Search indexes knowledge from tools like Google Drive, Jira, Confluence, Workday and SharePoint, and uses more than 100 MCP connectors for live lookups. Results appear next to Databricks dashboards, tables and Genie Agents.

How does Genie help business users analyze data?

You ask in plain language, and Genie writes and runs the SQL on governed data, then returns tables, charts and citations. Genie Ontology tells it which sources your team trusts, so it doesn't have to guess.

What should I look for in an enterprise AI assistant?

Look for business context, the ability to act, governance and reach. Genie One covers all four: Genie Ontology for context, MCP actions and artifacts for action, Unity Gateway for governance, and native apps for reach.

Does Genie respect my company's data permissions?

Yes. It respects Unity Catalog permissions and each connected source's access controls, so users see only what they're allowed to see. Admins use Unity Gateway to set MCP policies, require approvals and trace what Genie does.

Does Genie work with the tools my team already uses?

Yes. It updates tools like Jira and Asana through MCP and has native Slack and Microsoft Teams apps. MCP plugins bring it into ChatGPT, Claude, Microsoft Copilot Cowork, and Grok Bot., with the same context and permissions everywhere.

How can Genie reduce ad hoc requests to the data team?

Business users get governed answers themselves, right in their own tools. Scheduled tasks and shared skills turn recurring requests into automatic, consistent workflows.

How does Genie understand my business?

Genie Ontology learns your definitions, metrics and business rules from the assets your teams already use, and ranks them by how authoritative they are. Certifying key metrics in Unity Catalog makes it more accurate.

How are Genie One costs managed?

Admins monitor usage and manage budgets in Unity Gateway. Genie One is free for users through January 31, 2027, and more detailed cost controls will be available before then.

来源：Databricks：Blog · databricks.com