Microsoft 的 Genie 新功能:Copilot Cowork、Chat、Teams 与 Excel

在 Microsoft Cowork 中轻点几下即可获得有治理保障的答案:Azure Databricks Genie 插件将 Genie One 连接到 Microsoft Copilot Cowork 与聊天,让用户从整个数据资产中获得有治理保障的洞察。
大多数 AI 助手能读取个人的邮件、文件和日历,却无法判断哪张营收表是经过认证的,也无法判断财务部门如何定义净留存。这导致结果不一致、采用率低,业务用户也缺乏信任。Genie Ontology 为 Genie 解决了这一上下文问题:它把团队明确定义的业务语义与从日常工作中学习到的上下文结合起来,使答案建立在对业务的完整理解之上。
一旦上下文实现统一,下一步就是让它发挥作用。随着 AI 助手和智能体在企业中扩散,组织需要大规模管控数据访问、支出和工具使用。缺少这种管控,敏感数据可能流向不该看到的人,AI 支出变得难以预测,智能体还可能连接到无人批准的工具。在 Genie 中,Unity Catalog 执行每个用户的数据权限,Unity Gateway 则为管理员提供所需的追踪记录、MCP 策略和预算控制,从而在全组织范围内管理 AI。
通过 Azure Databricks Genie 插件,Genie 把关键的上下文与治理能力带入团队已经在使用的 Microsoft 工具中——该插件将 Microsoft Copilot 和 Copilot Cowork 连接到 Genie One。Genie 还可在 Microsoft Teams、Microsoft Excel 和 Microsoft Copilot Studio 中使用。
Microsoft Teams 中的 Genie:在每个频道和私信中提供有治理保障的答案
面向 Microsoft Teams 的 Databricks Genie 应用现已正式可用,把 Genie 带入团队日常做决策的对话之中。用户可以向 Genie 发送私信,或在频道、群聊中 @Databricks Genie,Genie 会在对话线程中回复,答案基于组织受治理的业务上下文。
借助 Teams 中的 Genie,用户可以基于整个数据资产中的知识展开实时协作。例如,一位销售经理可以在频道中发帖:“哪些金额超过 $100K 的未关闭销售线索本周还没有被跟进过?”无需有人手动调取销售管道报告、再与 CRM 交叉核对、翻找会议记录、逐一给销售代表发消息,她现在只需 @Genie。Genie 会在该线程中回复匹配的销售线索、它们的最新活动,以及所引用来源的链接——这不仅为团队节省了时间和精力,也让工作就在问题最初被提出的地方完成。
Teams 中的 Genie 不止于给出答案,它还让用户可以直接在自己的 Teams 工作区中把洞察转化为工作成果:
安排定期分析:点击答案下方的“按计划运行此分析”,或用自己的话提出请求,例如“每周一上午 9 点把这个分析发给我”。结果会发送到你的私信。
处理文件:附上文件供 Genie 分析,并在该线程中收到 PDF 等文档。
在线程中分享图表:Genie 会直接在 Teams 中返回可视化图表,团队需要时图表已经就绪。
Azure Databricks Genie 插件:在 Copilot Cowork 和 Chat 中提供受治理的答案
面向 Microsoft Copilot Cowork 和 Copilot Chat 的 Azure Databricks Genie 插件,把可信的业务答案带入用户可能已经在用的 Copilot 体验中。Copilot 不再需要自己去搜寻正确答案,而是可以把工作委派给 Genie One。Genie 借助 Genie Ontology,基于经过认证的定义、指标视图以及从日常工作中习得的上下文,提供准确的洞察,而这一切都由 Unity Catalog 统一治理。
Genie 与 Copilot 结合,可以为团队解锁全新的工作流。例如,设想一个客户团队正在为客户会议做准备。客户总监请 Copilot Cowork 提供一份客户会议前的简报,Cowork 便把任务委派给 Genie。Genie 使用经过认证的指标定义,从 Azure Databricks 中提取用量、支出和账户趋势,再通过其 Microsoft 365 连接器读取 Outlook 中最近的客户邮件往来,标记出尚未解决的升级事项。Cowork 把这些洞察转化为一份 Word 简报文档、一份带图表的 Excel 概览,以及一份可直接交付客户的演示文稿。每一个数字都来自 Genie,从而确保交付材料与财务部门和客户团队已经上报的数字保持一致。
Microsoft Excel 中的 Genie:把受治理的答案带进工作簿
许多团队在 Excel 中做规划、预测和汇报,但要把可信数据放进工作簿,往往意味着导出文件、复制数字,然后指望它们与源数据一致。借助 Azure Databricks Excel Add-in 中的 Genie One,分析师可以用自然语言提问,直接从受治理的 Azure Databricks 数据中获得答案,无需翻找数据表或编写 SQL。每个答案都附带引用,标明数字的来源;分析师还可以通过追问来细化分析、恢复之前的对话,并把表格结果直接加载到工作簿中做进一步建模。
Excel 加载项还让团队能够:
基于 Unity Catalog 指标视图和表构建实时数据透视表。拖动字段时结果即自动填充,无需刷新按钮。
安排刷新计划,让周期性报表、预测和记分卡始终与最新数据保持同步。
将最终确定的输入内容(例如预测假设或规划更新)写回 Databricks 表。
这些能力结合起来,把受治理的 Azure Databricks 数据带入团队已在使用电子表格工作流中,无需导出或手动搬运数据。
Microsoft Copilot Studio:构建连接 Genie Agents 的自定义智能体
构建者可以使用面向 Power Platform 的 Azure Databricks 连接器,将 Genie Agents 连接到 Microsoft Copilot Studio。将 Genie 添加为工具后,Copilot Studio 助手就能把有关 Azure Databricks 数据的问题交由 Genie Agents 处理,后者借助经过整理的数据和指令作答,并执行由 Unity Catalog 实施的治理。
强强联合:Microsoft + Genie
业务问题很少只停留在一个工具里,因此答案也不应如此。Microsoft 与 Genie 携手,将来自 Azure Databricks 的受治理业务上下文引入每位员工已在使用的工具:Copilot Chat、Copilot Cowork、Teams 和 Excel。Genie Ontology 确保每个答案都基于你的企业上下文,而 Unity Catalog 和 Unity Gateway 为每个答案和操作执行治理。现在,无论问题从哪里开始,无论是在 Teams 讨论串、Copilot 聊天还是 Excel 工作簿中,用户都能获得同样可信、受治理的答案。
什么是 Genie Ontology?Genie Ontology 是企业上下文层,让 Genie 和其他 AI 智能体对企业业务形成共同理解。智能体从日常工作中构建它,由人员治理,并且每个智能体都能使用它。它将团队定义的业务语义(如指标视图和经认证的定义)与从仪表板、查询和文档中学习到的上下文相结合。Genie Ontology 现已正式可用。
Genie Ontology 如何帮助 Microsoft Copilot 给出更好的答案?它让 Copilot 理解数据背后的业务含义。当 Copilot 或 Copilot Cowork 向 Genie 提问时,Genie 会用经认证的定义来回答,例如财务部门如何定义净留存率,而不是根据表名猜测。团队在每个 Microsoft 工具中都能获得一致的答案。
管理员如何跨 Microsoft 工具治理 Genie?通过 Unity Gateway。管理员借助 MCP 策略、成本控制和监控集中管理 AI 使用情况。MCP 策略决定 Genie 可以使用哪些 MCP 服务器和工具、谁可以访问它们,以及哪些操作需要人工审批。追踪记录会记录 Genie 做了什么以及是谁发起的请求。
我如何开始在 Microsoft Teams 中使用 Genie?
管理员将 Databricks Genie 应用添加到 Microsoft Teams,并为回答选择默认的可见范围。随后,用户可以向 Genie 发送私信,或在频道或群聊中 @Databricks Genie。首次使用时,他们需要登录 Databricks 并选择一个工作区。
Teams 中的 Genie 支持 Genie One 和 Genie Agents 吗?支持。Genie One 是默认选项。它会检索你的整个数据资产来回答每个问题。你也可以将某个会话或频道固定到特定的 Genie Agent。之后,问题会通过 Genie Agentic API 发送给该智能体,因此回答会限定在该智能体经过整理的数据和指令范围内。
Genie 在 Microsoft Teams 中能做什么?Teams 中的 Genie 远不止回答问题。你可以:
发现并可视化洞察:在私信或频道中获取来自 Databricks 数据的答案、表格和图表。
搜索公司知识:从已连接的来源中引入上下文。
在其他工具中执行操作:通过 MCP 连接器创建或更新文档、对工单发表评论,或更新记录。Genie 在做出更改前会征求你的批准。
安排周期性分析:按计划运行分析,并在私信中获取结果。要设置,请让 Genie“每周一上午 9 点把这份分析发给我”。
处理文件:附加 PDF、Word 或 CSV 文件供 Genie 分析,并可拿回 PDF 等文档。
原文较长,译文为节选;译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(What's new for Genie in Microsoft: Copilot Cowork and Chat, Microsoft Teams, and Excel.)
Governed answers in Microsoft Cowork, in a few clicks: The Azure Databricks Genie plugin connects Genie One to Microsoft Copilot Cowork and chat, so users get governed insights from across their data estate.
Most AI assistants can read a person's email, files, and calendar, but it can't tell which revenue table is certified or how finance defines net retention. This leads to inconsistent outcomes, low adoption, and a lack of trust among business users. Genie Ontology solves this context problem for Genie by combining the business semantics teams explicitly define with context learned from everyday work, so answers are grounded in a full understanding of the business.
Once that context is unified, the next step is putting it to work. As AI assistants and agents spread across the business, organizations need to control data access, spend, and tool usage at scale. Without that control, sensitive data can reach the wrong people, AI spend becomes hard to predict, and agents can connect to tools no one has approved. In Genie, Unity Catalog enforces each user's data permissions, and Unity Gateway gives admins the traces, MCP policies, and budget controls they need to manage AI across the organization.
Genie brings both essential context and governance into the Microsoft tools your teams already use with the Azure Databricks Genie plugin, which connects Microsoft Copilot and Copilot Cowork to Genie One. Genie also works in Microsoft Teams, Microsoft Excel, and Microsoft Copilot Studio.
Genie in Microsoft Teams: governed answers in every channel and DM
The Databricks Genie app for Microsoft Teams is now generally available, bringing Genie into the conversations where teams already make decisions. Users can send Genie a direct message or mention @Databricks Genie in a channel or group chat, and Genie replies in the thread with answers grounded in the organization’s governed business context.
With Genie in Teams, users unlock real-time collaboration with knowledge from across their data estate. For example, a sales manager can post in a channel, "Which open leads over $100K haven't had a follow-up this week?" Instead of someone manually pulling the pipeline report, cross-checking it against the CRM, searching meeting notes, and messaging reps one by one, she can now simply tag @Genie. Genie answers in the thread with the matching leads, their latest activity, and links to the sources it used, not only saving the team time and effort, but also allowing the work to happen right where the question was originally asked.
Genie in Teams goes beyond answers, enabling users to turn insights into work right from their Teams workspace:
Schedule recurring analyses: Click “Run this analysis on a schedule” under an answer, or ask in your own words, for example "send me this analysis every Monday at 9 AM." Results arrive in your DM.
Work with files: Attach a file for Genie to analyze, and get documents such as PDFs back in the thread.
Share charts in a thread: Genie returns visualizations directly in Teams, so the chart is ready when your team needs it.
The Azure Databricks Genie plugin: Governed answers in Copilot Cowork and Chat
The Azure Databricks Genie plugin for Microsoft Copilot Cowork and Copilot Chat brings trusted business answers into the Copilot experience users may already rely on. Instead of Copilot searching for correct answers on its own, it can delegate work to Genie One. Genie leverages Genie Ontology to provide accurate insights based on certified definitions, metric views, and context learned from everyday work, all governed by Unity Catalog.
Together, Genie and Copilot can unlock new workflows for teams. For example, imagine an account team preparing for client meetings. An account director asks Copilot Cowork for a briefing ahead of a customer meeting, which delegates to Genie. Genie pulls usage, spend, and account trends from Azure Databricks using certified metric definitions, then reads recent customer email threads in Outlook through its Microsoft 365 connector to flag open escalations. Cowork turns those insights into a Word doc brief, an Excel overview with charts, and a client-ready presentation. Every figure comes from Genie, keeping the deliverables consistent with the numbers finance and the account team already report.
Genie in Microsoft Excel: governed answers inside the workbook
Many teams plan, forecast, and report in Excel, but getting trusted data into a workbook often means exporting files, copying numbers, and hoping they match the source. With Genie One in the Azure Databricks Excel Add-in, analysts can ask questions in plain English and get answers from governed Azure Databricks data without hunting for tables or writing SQL. Every answer includes citations that show where the numbers came from, and analysts can refine their analysis with follow-up questions, resume past conversations, and load table results directly into the workbook for further modeling.
The Excel Add-in also lets teams:
Build live pivots on Unity Catalog metric views and tables. Results populate as fields are dragged in, with no refresh button.
Schedule refreshes so recurring reports, forecasts, and scorecards stay current with the latest data.
Write back finalized inputs, such as forecast assumptions or planning updates, to Databricks tables.
Together, these capabilities bring governed Azure Databricks data into the spreadsheet workflows teams already rely on, without exports or manual data movement.
Microsoft Copilot Studio: build custom agents connected to Genie Agents
Builders can connect Genie Agents to Microsoft Copilot Studio using the Azure Databricks connector for Power Platform. Adding Genie as a tool lets a Copilot Studio assistant delegate questions about Azure Databricks data to Genie Agents, which answer using curated data and instructions and governance enforced by Unity Catalog.
Better together: Microsoft + Genie
Business questions rarely stay in one tool, so the answers shouldn't either. Together, Microsoft and Genie bring governed business context from Azure Databricks into the tools every employee already uses: Copilot Chat, Copilot Cowork, Teams, and Excel. Genie Ontology makes sure every answer is grounded in your enterprise context, and Unity Catalog and Unity Gateway enforce governance for every answer and action. Now, wherever a question starts, whether in a Teams thread, a Copilot chat, or an Excel workbook, users get the same trusted, governed answer.
What is Genie Ontology?
Genie Ontology is the enterprise context layer that gives Genie and other AI agents a shared understanding of your business. Agents build it from everyday work, people govern it, and every agent can use it. It combines business semantics that teams define, such as metric views and certified definitions, with context learned from dashboards, queries, and documents. Genie Ontology is now generally available.
How does Genie Ontology help Microsoft Copilot give better answers?
It gives Copilot the business meaning behind the data. When Copilot or Copilot Cowork asks Genie a question, Genie answers with the certified definition, such as how finance defines net retention, instead of guessing from table names. Teams get consistent answers in every Microsoft tool.
How do admins govern Genie across Microsoft tools?
Through Unity Gateway. Admins manage AI usage centrally with MCP policies, cost controls, and monitoring. MCP policies decide which MCP servers and tools Genie can use, who can access them, and which actions need human approval. Traces record what Genie did and who asked for it.
How do I get started with Genie in Microsoft Teams?
Admins add the Databricks Genie app to Microsoft Teams and choose the default visibility for responses. Users then send Genie a direct message or mention @Databricks Genie in a channel or group chat. The first time, they sign in to Databricks and pick a workspace.
Does Genie in Teams support Genie One and Genie Agents?
Yes. Genie One is the default. It looks across your entire data estate to answer each question. You can also pin a conversation or channel to a specific Genie Agent. Questions are then sent to that agent through the Genie Agentic API, so answers stay scoped to that agent's curated data and instructions.
What can Genie do in Microsoft Teams?
Genie in Teams does much more than answer questions. You can:
Discover and visualize insights: Get answers, tables, and charts from your Databricks data in a direct message or channel.
Search your company's knowledge: Pull in context from connected sources.
Take action in other tools: Create or update a document, comment on a ticket, or update a record through MCP connectors. Genie asks for your approval before it makes a change.
Schedule recurring analyses: Run an analysis on a schedule and get the results in your DM. To set up, ask Genie to "Send me this analysis every Monday at 9 AM."
Work with files: Attach PDF, Word, or CSV files for Genie to analyze, and get documents such as PDFs back.
Pick the right expert: Use Genie One across your whole data estate, or pin a conversation or channel to a specific Genie Agent.
Give feedback: Rate answers with a thumbs up or down to help improve Genie.
Does Genie in Microsoft Teams support workspaces that use Private Link?
Yes. A Databricks account admin needs to create a context-based ingress policy to allow secure connectivity between Teams and your Databricks workspace.
Is my data secure for Genie in Microsoft Teams?
Yes. Every question runs as the user who asked it, so Genie only uses data that person is allowed to see, based on Unity Catalog and source permissions. Questions are processed inside your Databricks workspace. The service that routes messages between Teams and Databricks doesn't store questions or answers.
Should I use the Databricks Genie app in Teams, or publish a Copilot Studio agent to Teams?
Use the Databricks Genie app in Microsoft Teams by default. It's a native app that chats directly with Genie One, so users get answers from across their data estate with no extra agent in between. It also works in Microsoft 365 Copilot. The app supports pinned Genie Agents, charts, file uploads, and MCP actions, and it doesn't require a Copilot Studio license. Choose Copilot Studio when you're building a custom agent that combines Genie Agents with other tools not available within Azure Databricks.
How do users get started with the Azure Databricks Genie plugin?
Users can get started in a few clicks. In Copilot Cowork, they turn on the Azure Databricks Genie plugin, click Connect, and sign in with their Azure Databricks credentials. The plugin doesn't require an OAuth app registration or manifest edits.
Which Microsoft tools work with Genie?
Genie works across Copilot Chat, Copilot Cowork, and Microsoft 365 apps. It's also available in Microsoft Teams, Microsoft Excel, Microsoft Copilot Studio, Microsoft Power Platform, and Microsoft Agent 365. Genie One in Microsoft Foundry brings the same governed context to agents built in Foundry.
How does Genie work with Microsoft Foundry?
Genie works with Microsoft Foundry in both directions. Agents built in Foundry can use Genie One to get governed Azure Databricks data while acting as the signed-in user. Microsoft Foundry models are also available in Azure Databricks through Unity Gateway.
Can Genie connect to Work IQ over MCP?
Genie will soon connect to Work IQ over MCP, through Model Context Protocol, so organizations that use Work IQ can bring Microsoft 365 context into Genie alongside governed Azure Databricks data. That context includes email, meetings, documents, Teams messages, and people and organizational context. Each request runs under the signed-in user's Microsoft Entra ID identity, so Microsoft 365 permissions, sensitivity labels, and compliance policies continue to apply while Unity Catalog governs the Azure Databricks data.
Get the latest posts in your inbox
原文为英文,正文由自动抓取并保留原文。
原文发布于 Databricks Blog(What's new for Genie in Microsoft: Copilot Cowork and Chat, Microsoft Teams, and Excel.),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。