千问办公陈宇森：数字员工也得有工号、有权限、能追责

作者：亿邦动力胡镤心文章来源：亿邦动力

2026-09-22 18:14

作者：亿邦动力胡镤心文章来源：亿邦动力

2026-09-22 18:14

邦小白快读

全文速览

总：文章介绍了千问办公新发布的企业上下文和数字员工能力，核心干货是让AI从个人助手进化为能融入企业流程的智能体。

1. 个人用AI写邮件、做PPT虽然快，但企业复杂流程如询盘、报价、合同、审批，AI原先插不进去，原因是Agent不懂业务，不知道何时调用什么数据、走什么规则、找谁协同。

2. 企业上下文解决这个问题：对多年数据进行压缩和结构化，让AI每次基于当前任务的最新真实情况做判断，而且用专属模型把Token成本压得很低。对普通员工来说，以后问AI问题，会得到与当前工作最相关的限定信息，而不是一堆无用知识。

3. 数字员工像真人同事一样有名称、所属部门、负责人、岗位职责，也有授权范围和生命周期。员工能识别它，系统能授权它，管理员能停用它，所有执行记录可追溯到身份。这会让AI做事更规范、可追责。

4. 新增的协作和应用功能让员工能建协作空间，把散落的群聊、文档、知识库信息关联起来，把数字员工拉进来一起讨论；业务系统也能以应用形式接入AI，免去在多个系统间反复切换。普通人日常办公可以关注这些功能，把重复性流程交给数字员工。

总：品牌商可从文章中的企业案例看到AI如何提升品牌运营与销售管理效率，尤其适合门店多、上新快、流程长的品牌。

1. 品牌运营标准可以借助企业上下文快速更新并下发到门店。古茗茶饮将飞书文档、知识库、答疑群、培训日程整理成门店运营知识空间，近万名一线员工能快速掌握新品新标准。当店员询问海报张贴规范时，AI不会返回所有知识，只给当前任务需要的上下文，这能减少品牌信息传递延迟和执行偏差。

2. 品牌销售流程可以由数字员工分工承接。富立卡案例中，销售人员每天处理300多条客户询盘并与6个部门协作，通过查询员工、报价员工、合同员工三个数字员工共享业务上下文，让询盘、报价、合同持续流转，减少品牌商在报价和合同环节的沟通成本。

3. 品牌在渠道和门店扩张中，可将AI能力视为组织基础设施。先让Agent懂业务，再让Agent进流程，最后与员工协同，是品牌实现从个人提效到组织提效的可行路径。文章还提到益海嘉里金龙鱼、一汽丰田等头部企业已接入，说明这套工具正在成为企业标配，品牌商可关注其后续应用方式。

总：对卖家而言，文章揭示了用AI数字员工提升销售转化和运营效率的新工具与商业机会，也提示了接入AI生态的窗口期。

1. 销售效率提升机会：富立卡案例展示了数字员工如何承接高频、重复的客户询盘和报价流程。卖家如果每天要处理大量询盘并与多个部门协作，可以借鉴这种三个数字员工分工的模式，让查询、报价、合同自动衔接，缩短响应时间。

2. 增长市场信号：千问办公将企业上下文作为新产品，并且已有益海嘉里金龙鱼、一汽丰田、方达律师事务所、传化集团等头部企业接入，说明企业级AI数字员工正从概念走向标配。卖家若提供产品或服务，可以尽早考虑如何把自身业务系统接入这类AI应用生态，比如以应用形式接入千问办公，获得企业客户触达渠道。

3. 风险与应对提示：Agent进入业务流程后有授权范围和生命周期，所有执行记录可追溯到身份，这提示卖家在使用AI时要关注权限设置和责任边界，避免因AI误操作引发合同或报价纠纷。同时，企业上下文降低了Token成本，对中小卖家来说，不必担心AI使用成本过高。

总：文章中的富立卡案例和一些数字化能力，对工厂推进生产、仓储、物流和销售协同有直接启发。

1. 数字员工能帮工厂处理跨部门协作：富立卡是仓储物流包装公司，销售团队每人每天处理300多条客户询盘，并与6个部门协作。千问办公为其配置查询员工、报价员工、合同员工，三个数字员工通过企业上下文共享信息，围绕同一笔业务分工协作，减少工厂内部沟通成本。

2. 企业上下文可把分散在群聊、文档、知识库中的生产和运营信息结构化，让一线人员快速找到与当前任务相关的规则和资料。工厂如果涉及多基地、多产线、频繁更新标准，可参照古茗茶饮的做法，建立类似的知识空间来培训和指导一线员工。

3. 工厂数字化不必从零建系统，可通过千问办公或钉钉打造数字员工，让员工有名称、部门、负责人、岗位职责、授权范围和生命周期，执行记录可追溯。文章还提到企业和SaaS厂商可将业务系统以应用形式接入，工厂原有的ERP、MES等系统可以在保留原界面的同时加入Agent能力，避免重复建设。

总：文章清晰指出了当前企业级AI服务的客户痛点，也给出了可参考的解决方案和生态合作机会。

1. 客户痛点包括：Agent无法深入业务工作流，不懂业务，不知道在什么节点调用什么数据、走什么规则、找谁协同；个人AI提效但组织生产力未线性增长；企业海量信息缺少结构化基础设施，导致Agent难以被调用。

2. 解决方案核心是企业上下文：对多年数据进行压缩和结构化，让Agent基于企业最新真实情况判断；内置专属模型，用极低Token成本完成数据处理，这解决了成本问题。服务商在给企业做AI落地时，可以借鉴这种“数据压缩+业务上下文”的思路，而不是直接把通用大模型接入企业。

3. 技术趋势是数字员工和Agent协同：数字员工有名称、所属部门、负责人、岗位职责、授权范围和生命周期，能被授权、被停用，执行记录可追溯。这要求服务商为客户设计Agent时，不仅考虑能力，还要考虑组织身份、权限和审计。

4. 合作机会：千问办公推出应用功能，鼓励企业和SaaS厂商将业务系统以应用形式接入，保留原有界面和数据的同时加入Agent能力。服务商可以抓住这个生态入口，把垂直行业业务系统接入千问办公，服务头部客户。

总：文章展示了企业办公平台在AI时代的关键做法，包括连接IM、开放应用生态、完善数字员工治理，值得平台商借鉴。

1. 平台需要解决业务数据获取问题：千问办公从诞生起就打通钉钉、飞书、企业微信等IM软件，沉淀群聊、文档等海量业务信息，并将这些信息结构化供Agent调用。这说明跨IM连接和数据治理是平台的核心竞争力。

2. 平台推出协作和应用两类功能：协作空间可将散落在群聊、文档、知识库的信息关联起来，支持人与Agent、Agent与Agent沟通分工；应用功能允许业务系统以应用形式接入并叠加Agent能力。对平台来说，这意味着既要做深度协同，也要做开放应用生态。

3. 平台必须重视数字员工治理：数字员工要有名称、所属部门、负责人、岗位职责，也要有授权范围和生命周期，可被员工识别、被系统授权、被管理员停用，所有执行记录能追溯到身份。这为平台制定数字员工管理规范提供了参考方向。

4. 平台招商或运营可以关注头部企业示范效应：文章列出益海嘉里金龙鱼、一汽丰田、方达律师事务所、传化集团等企业接入，说明平台应重点服务头部客户形成标杆，再向行业推广。

总：文章揭示了AI从个人工具走向组织基础设施过程中出现的产业新动向和新问题，也提供了关于Agent治理和商业模式的研究素材。

1. 产业新动向：千问办公发布企业上下文，旨在解决Agent无法深入业务工作流的问题。文章提出组织生产力并未因个人AI提效而线性增长，这提示研究者关注“个人效率与组织效率之间的转化机制”这一重要命题。

2. 新问题：Agent在真实企业场景落地需要懂业务、懂组织协同，且每次任务执行要有边界、能追溯。文章将企业多年数据结构化定义为基础设施建设，说明数据压缩与上下文调用是当前AI落地的主要瓶颈之一。

3. 治理启示：数字员工被赋予工号一般的身份，有名称、所属部门、负责人、岗位职责、授权范围和生命周期，执行记录可追溯到身份。这为研究AI组织行为、Agent责任认定、权限管理和退出机制提供了具体案例，可进一步探讨数字员工的法律主体与追责框架。

4. 商业模式：千问办公给出的路径是先让Agent懂业务，再让Agent进流程，最后让Agent与人协同，并以企业上下文作为第一块基石。同时，协作空间和应用功能允许SaaS厂商和业务系统接入，形成企业与第三方共建的生态模式，可研究平台型AI办公的商业模式与竞争壁垒。

声明：快读内容全程由AI生成，请注意甄别信息。如您发现问题，请发送邮件至 run@ebrun.com 。

我是品牌商卖家工厂服务商平台商研究者帮我再读一遍。

Quick Summary

Summary: The article introduces Qianwen Office's newly launched Enterprise Context and digital employee capabilities. The core takeaway is that AI evolves from a personal assistant to an agent that can integrate into enterprise workflows.

1. AI is fast at writing emails and making PPTs for individuals, but in complex enterprise processes such as inquiry, quotation, contract, and approval, AI previously had no entry point. The reason is that Agents lack business knowledge—they don't know when to call which data, follow which rules, or seek collaboration from whom.

2. Enterprise Context solves this by compressing and structuring years of data, enabling AI to make judgments based on the latest real situation for each task, while a dedicated model keeps token costs extremely low. For ordinary employees, asking AI will return a limited set of information most relevant to the current work, rather than a pile of useless knowledge.

3. Digital employees, like real colleagues, have a name, department, manager, job responsibilities, plus authorization scope and lifecycle. Employees can identify them, systems can authorize them, and administrators can deactivate them, with all execution records traceable to an identity. This makes AI's work more standardized and accountable.

4. The new collaboration and application features allow employees to create collaboration spaces that connect scattered group chats, documents, and knowledge-base information, bringing digital employees into discussions. Business systems can also be connected to AI as applications, eliminating repeated switching between systems. Ordinary people can pay attention to these features for daily work and delegate repetitive processes to digital employees.

Summary: Brands can see from the enterprise cases in the article how AI improves brand operations and sales management efficiency, especially for brands with many stores, fast product launches, and long processes.

1. Brand operating standards can be rapidly updated and distributed to stores using Enterprise Context. Guming Tea organized Feishu documents, knowledge bases, Q&A groups, and training schedules into a store-operations knowledge space, enabling nearly 10,000 frontline employees to quickly master new products and updated standards. When a store clerk asks about poster posting rules, AI returns only the context needed for the task, not all knowledge, reducing delays in brand messaging and execution deviations.

2. Brand sales processes can be handled by digital employees with a division of labor. In the Fulika case, sales staff process over 300 customer inquiries per day and coordinate with six departments. By using three digital employees—inquiry agent, quotation agent, and contract agent—sharing business context, inquiries, quotations, and contracts flow continuously, reducing communication costs in the quotation and contract stages for brands.

3. During channel and store expansion, brands can treat AI capabilities as organizational infrastructure. Having Agents first understand business, then enter processes, and finally collaborate with employees is a viable path for brands to move from individual efficiency to organizational efficiency. The article also mentions that leading companies such as Yihai Kerry Arawana and FAW Toyota have adopted it, indicating this tool is becoming an enterprise standard. Brands should watch its future applications.

Summary: For sellers, the article reveals new tools and business opportunities to improve sales conversion and operational efficiency through AI digital employees, while also signaling the window for plugging into the AI ecosystem.

1. Sales efficiency opportunity: The Fulika case shows how digital employees handle high-frequency, repetitive customer inquiries and quotation processes. If sellers process a large volume of inquiries daily and coordinate with multiple departments, they can borrow this model of dividing tasks among three digital employees to automatically connect inquiry, quotation, and contract, shortening response time.

2. Growth market signal: Qianwen Office has introduced Enterprise Context as a new product, and leading enterprises such as Yihai Kerry Arawana, FAW Toyota, Fangda Partners, and Transfar Group have already adopted it, indicating that enterprise-level AI digital employees are moving from concept to standard. If sellers offer products or services, they should consider early how to connect their own business systems into this AI application ecosystem—for example, integrating into Qianwen Office as an application—to gain access to enterprise customers.

3. Risk and mitigation: After Agents enter business processes, they have authorization scopes and lifecycles, and all execution records are traceable to identities. This reminds sellers to pay attention to permission settings and responsibility boundaries when using AI, avoiding contract or quotation disputes caused by AI misoperations. Meanwhile, Enterprise Context lowers token costs, so small and medium sellers need not worry about high AI usage costs.

Summary: The Fulika case and several digital capabilities in the article offer direct inspiration for factories to advance collaboration across production, warehousing, logistics, and sales.

1. Digital employees can help factories handle cross-department collaboration: Fulika is a warehousing, logistics, and packaging company where each salesperson handles over 300 customer inquiries per day and coordinates with six departments. Qianwen Office deployed inquiry, quotation, and contract agents for it; the three digital employees share information via Enterprise Context and divide work around the same business, reducing internal communication costs in the factory.

2. Enterprise Context can structure production and operational information scattered across group chats, documents, and knowledge bases, allowing frontline staff to quickly find rules and materials relevant to the task at hand. If a factory operates multiple sites and lines with frequent standard updates, it can follow Guming Tea's approach and build a similar knowledge space to train and guide frontline employees.

3. Factory digitalization does not require building systems from scratch. Companies can create digital employees through Qianwen Office or DingTalk, endowing them with name, department, manager, job responsibilities, authorization scope, and lifecycle, with traceable execution records. The article also notes that enterprises and SaaS vendors can connect business systems as applications, so existing factory systems such as ERP and MES can add Agent capabilities while retaining their original interfaces, avoiding redundant construction.

Summary: The article clearly identifies customer pain points in current enterprise AI services, and offers reference solutions and ecosystem partnership opportunities.

1. Customer pain points include: Agents cannot go deep into business workflows, lack business knowledge, and don't know at which point to call which data, follow which rules, or find the right collaborator; personal AI improves efficiency but organizational productivity does not grow linearly; and enterprises lack structured infrastructure for their massive information, making it hard to invoke Agents.

2. The core of the solution is Enterprise Context: compressing and structuring years of data so Agents can judge based on the latest real enterprise situation; an embedded dedicated model processes data at extremely low token cost, solving the cost problem. When implementing AI for enterprises, service providers can borrow this data compression plus business context approach rather than directly connecting a general large model to the enterprise.

3. The technology trend is digital employees and Agent collaboration: digital employees have a name, department, manager, job responsibilities, authorization scope, and lifecycle; they can be authorized, deactivated, and their execution records traced. This requires service providers, when designing Agents for clients, to consider not only capabilities but also organizational identity, permissions, and audit.

4. Partnership opportunity: Qianwen Office launched application features encouraging enterprises and SaaS vendors to connect business systems as applications, adding Agent capabilities while retaining original interfaces and data. Service providers can seize this ecosystem entrance to connect vertical-industry business systems to Qianwen Office and serve top-tier clients.

Summary: The article showcases key practices for enterprise office platforms in the AI era, including connecting IM, opening an application ecosystem, and improving governance of digital employees—practices worth learning from for platform providers.

1. Platforms need to solve the problem of acquiring business data: Since its launch, Qianwen Office has integrated IM software such as DingTalk, Feishu, and WeChat Work, accumulating massive business information from group chats and documents, and structuring this information for Agents to invoke. This shows that cross-IM connectivity and data governance are core competitive advantages for platforms.

2. Platforms are launching two types of features—collaboration and applications: collaboration spaces connect information scattered in group chats, documents, and knowledge bases, supporting human-Agent and Agent-Agent communication and division of work; application features allow business systems to connect as applications and layer on Agent capabilities. For platforms, this means both deep collaboration and an open application ecosystem are needed.

3. Platforms must prioritize governance of digital employees: they need a name, department, manager, job responsibilities, as well as authorization scope and lifecycle; they should be identifiable by employees, authorizable by systems, and deactivatable by administrators, with all execution records traceable to an identity. This provides a reference direction for platforms to establish management standards for digital employees.

4. Platform recruitment or operations can focus on the demonstration effect of leading enterprises: The article lists adoptions by Yihai Kerry Arawana, FAW Toyota, Fangda Partners, Transfar Group, and others, suggesting that platforms should prioritize serving top clients to create benchmarks before promoting to the industry.

Summary: The article reveals new industry trends and issues as AI moves from individual tool to organizational infrastructure, and provides research material on Agent governance and business models.

1. New industry trend: Qianwen Office launched Enterprise Context to address the problem that Agents cannot go deep into business workflows. The article argues that organizational productivity does not grow linearly from individual AI efficiency, prompting researchers to focus on the important proposition of the transformation mechanism between individual and organizational efficiency.

2. New problems: Deploying Agents in real enterprise scenarios requires understanding business and organizational collaboration, with each task execution having boundaries and traceability. The article defines structuring years of enterprise data as infrastructure construction, indicating that data compression and context invocation are among the main bottlenecks for current AI deployment.

3. Governance insights: Digital employees are given employee-ID-like identities, with name, department, manager, job responsibilities, authorization scope, and lifecycle, and execution records traceable to identities. This provides concrete cases for studying AI organizational behavior, Agent accountability, permission management, and exit mechanisms, and enables further discussion of the legal subject and accountability framework for digital employees.

4. Business model: Qianwen Office's path is to first make Agents understand business, then bring Agents into processes, and finally let Agents collaborate with people, using Enterprise Context as the first cornerstone. At the same time, collaboration spaces and application features allow SaaS vendors and business systems to connect, forming an ecosystem co-built by enterprises and third parties. This opens research on the business models and competitive moats of platform-based AI office solutions.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

【亿邦原创】9月22日，2026杭州云栖大会，千问办公发布面向企业的全新产品——企业上下文（Enterprise Context），并同步推出数字员工、协作、应用、安全中心等能力。

阿里巴巴集团副总裁、千问办公CEO陈宇森表示：“Agent要在企业真实场景落地，既要懂业务，也要像人一样在企业和组织内协同，且每一次任务执行要有边界、能追溯。”

1、个人提效了，组织为什么没提效？

过去一年，AI让个人办公效率显著提升。但一个尴尬的现实是：组织生产力并未因个人提效而线性增长。

问题出在哪？陈宇森给出的答案是：Agent无法深入业务工作流。个人用AI写邮件、做PPT很快，但企业里真正复杂的流程——询盘、报价、合同、审批、投研——AI插不进去。因为Agent不懂业务，不知道在什么节点该调用什么数据、该走什么规则、该找谁协同。

“数据是Agent进入企业业务流程的最重要前提。”陈宇森说。千问办公从诞生起就打通了钉钉、飞书、企业微信等IM软件，沉淀了群聊、文档等海量业务信息。但业界长期缺少将海量信息结构化并供Agent调用的基础设施。

2、企业上下文：给Agent一张业务地图

一家企业可能积累了多年的数据，但员工开始一项具体任务时，需要的不是全部信息，而是与当前任务最相关的那部分。

企业上下文解决的就是这个问题：对数据进行压缩和结构化，让Agent每次都能基于企业最新真实情况作判断。它内置专属模型，可以用极低的Token成本完成数据的压缩和处理。

古茗茶饮是一个典型案例。古茗上新频繁，运营标准需要及时更新并下发到门店。通过企业上下文，飞书文档、知识库、答疑群、培训日程中的内容被整理成“门店运营知识空间”，帮助近万名一线员工快速掌握新品新标准。当店员通过Agent询问海报张贴规范时，Agent连接的企业上下文不会返回所有运营知识，而是根据请求提供当前任务所需的上下文。

有了企业上下文，企业可以基于千问办公和钉钉打造数字员工——数字员工有名称、所属部门、负责人、岗位职责，也有授权范围和生命周期——可以被员工识别、被系统授权、被管理员停用，所有执行记录都能追溯到身份。

3、协作与应用：Agent之间也能分工

仓储物流包装公司富立卡的销售团队每人每天需处理300多条客户询盘，并与6个部门协作。千问办公为它配置了三个数字员工：查询员工匹配客户需求与产品，报价员工结合客户信息和规则测算报价，合同员工承接前序信息生成合同并按企业规则发起审批。

三个数字员工各司其职，通过企业上下文共享当前业务所需信息，围绕同一笔业务分工协作，推动询盘、报价、合同持续流转。

这意味着数字员工也不是单打独，查询、报价、合同三个员工之所以能围绕同一笔业务持续流转，靠的是它们之间共享上下文、分工协作。但现实中，一笔业务的完成往往还需要更多角色加入——人、多个Agent、跨部门的信息，散落在不同的群聊和文档里。

这正是千问办公推出“协作”功能的原因。

过去，项目资料和进度分散在不同群聊和文档中，团队成员反复同步。数字员工加入后，协作扩展到人与Agent、Agent与Agent之间。基于协作功能，企业可以创建协作空间，关联散落在群聊、文档和知识库中的信息，按需加入团队成员和数字员工。员工可直接与同事或Agent讨论，Agent之间也能相互沟通分工，产出和进度统一沉淀在空间中。

当然，并非所有工作都适合在聊天框里完成，千问办公同时推出“应用”功能，企业和SaaS厂商可将业务系统以应用形式接入，在保留原有界面、数据的同时加入Agent能力。以金融投研为例，过去研究员要完成“比较几家公司、形成一份投研简报”，AI只能给出分析结果，之后还要回到金融系统核对数据、保存任务。现在所有任务均可在千问办公“应用”中完成，无需在两个系统间反复切换。

当Agent既能与人协作、又能嵌入业务系统完成闭环，企业真正关心的就不再是能不能用，而是谁已经在用。大会现场，又有一批头部企业宣布接入千问办公：益海嘉里金龙鱼、一汽丰田、方达律师事务所、传化集团、中国数联、泽大律师事务所、横店东磁、国邦医药等。千问办公已成头部企业的“标配”。

从个人提效到组织提效，千问办公给出的路径是：先让Agent懂业务，再让Agent进流程，最后让Agent与人协同。

企业上下文，就是这条路径的第一块基石。

本文首发于亿邦动力官方网站

【版权提示】亿邦动力倡导尊重与保护知识产权。未经许可，任何人不得复制、转载、或以其他方式使用本网站的内容。如发现本站文章存在版权问题，烦请提供版权疑问、身份证明、版权证明、联系方式等发邮件至run@ebrun.com，我们将及时沟通与处理。

已点赞

好文章,值得被更多人看到

FAQ回顾

什么是千问办公的企业上下文？

企业上下文是千问办公面向企业发布的产品，用于对企业在群聊、文档、知识库等场景积累的海量信息进行压缩和结构化，让Agent在执行任务时能按需调用最相关的数据。它内置专属模型，能以极低Token成本完成数据处理，帮助Agent基于企业最新真实情况做判断，是企业从个人提效升级到组织提效的基础设施。

数字员工在企业中如何运作？

数字员工是基于千问办公和钉钉打造的企业级Agent，拥有名称、所属部门、负责人、岗位职责、授权范围和生命周期，能被员工识别、被系统授权、被管理员停用，所有执行记录均可追溯到身份。例如富立卡配置了查询、报价、合同三个数字员工，分别处理客户询盘匹配、报价测算和合同审批，通过企业上下文共享信息，协同推进业务流程。

企业如何利用Agent提升组织生产力？

千问办公给出的路径是先让Agent懂业务，再让Agent进流程，最后让Agent与人协同。首先通过企业上下文将分散的业务信息结构化，让Agent理解业务规则和数据；再通过数字员工、协作空间和应用功能，使Agent深入询盘、报价、合同、审批等复杂流程，与员工及彼此协同，从而解决个人提效但组织生产力未线性增长的问题。