Qwen Office Launches Enterprise Context for Agent-Era Workplace AI

Alibaba Group Vice President and Qwen Office CEO Chen Yusheng said at the event that Qwen Office is the industry's first enterprise-level agent product. He said agents must understand business, collaborate inside an organization like a person, and carry out every task within clear boundaries that can be traced. Those three requirements, understanding business, collaborating and being trusted, correspond to questions companies must answer when adopting AI: how much business the agent knows, how it works with colleagues, and how much authority it can be granted.

The first problem is context. Writing reports and organizing spreadsheets have become common capabilities for office agents, but companies have their own business rules. A request to give a customer a quote may involve product systems, discount rights and approval processes that differ by company. Even within one company, a meeting conclusion from three months ago may no longer be valid, and temporary discussions in group chats may not represent final decisions. How much an agent can accomplish at a company depends on its understanding of business context. Qwen Office previously launched MyContext, a personal context project that processes heterogeneous personal work data into context agents can use directly. Enterprise Context extends that capability to the enterprise level. It organizes information scattered across group chats, documents and knowledge bases into structured enterprise context, provides relevant content by task and updates as business changes.

Guming's store operations were cited as an example. Frequent new product launches mean production requirements, material standards and training schedules change constantly, and frontline staff need the latest instructions from multiple channels. Enterprise Context organizes content from Guming's Feishu documents, knowledge bases, Q&A groups and training calendars into a store operations knowledge space. When staff ask about poster posting rules, the agent retrieves the corresponding information and maintains the knowledge space as standards update, reducing time spent searching and asking repeated questions. Qwen Office said that with Enterprise Context and its agent hosting capabilities, companies can build digital employees that understand business and collaborate with human employees on DingTalk and other IM platforms.

Information also originates offline. QwenNote and its first product A2 support voice transcription and voice invocation of Qwen Office, turning offline conversations into records for later work. The device retains only text and meeting minutes, not the original audio, according to the company.

After understanding business, an agent must find its place in the organization. A quote moves from sales to finance, procurement and legal, with each role having its own input and responsibilities. Companies can build digital employees based on Qwen Office and DingTalk, giving agents an organizational badge. They have names, departments, owners and job responsibilities, as well as clear authorization scopes and lifecycles. They can be recognized by employees, authorized by systems and disabled by administrators, and all execution records can be traced to an identity. DingTalk and other collaboration platforms decide which group chats digital employees can enter and which systems they can access. Enterprise Context provides the business information they need. Managed Agents manage how they work. Companies can set goals and boundaries, configure models, instructions and tools, and authorize runtime environments and credentials.

Fulika, a warehousing and logistics packaging company, has applied this approach to sales. Its sales team handles more than 300 customer inquiries per person per day and collaborates with six departments. Qwen Office configured three digital employees for Fulika. They share the information currently needed for a business deal through Enterprise Context, allowing inquiries, quotes and contracts to advance along the same process. The agent's role is no longer an isolated task but a position in the organizational division of labor.

Qwen Office also launched a Collaboration feature that lets colleagues and digital employees discuss, assign work and keep progress and outputs in the same space. Humans and agents, and agents and agents, can share necessary background, and later participants can continue from existing records. The Applications feature allows companies and SaaS vendors to introduce agents while keeping existing pages and data. The agent can understand the user's current operation and progress and help complete tasks. In investment research, for example, researchers can complete analysis, verification and revisions inside the application and save results to the business system.

As agents move deeper into business processes, their permissions grow and the risks get closer to the enterprise core. Chen Yusheng said companies cannot only add defenses around the system; they must consider agent risks such as permission overreach, data leaks and mistaken operations from the beginning of architecture design. The Qwen Office Security Center manages data access, tool calls and task execution. Companies can use sandboxes to limit runtime environments, set approvals and interception for high-risk operations, and handle problems after anomalies based on operation records and recovery mechanisms. With the enterprise permission system, administrators can grant necessary permissions by role, identify sensitive information, audit execution, and trace who initiated a task, what data was used and what operations were performed.

Microsoft's 2026 Work Trend Index found that among AI users surveyed, 66% said AI allowed them to spend more time on high-value work and 58% said they could complete work they could not do a year earlier. The report also said that work methods in many organizations have not kept up with changes in employee capabilities. Gartner predicts that by the end of 2026, 40% of enterprise applications will embed task-specific AI agents, up from less than 5% in 2025. Qwen Office said the related products are now open for invitation testing to companies including Guming, Fulika and Changan Automobile.

The company positions DingTalk and QwenNote hardware as entry points that collect daily communication, meetings and documents. Enterprise Context organizes that information into callable organizational context. Qwen Office then executes tasks and returns results to collaboration scenarios, connecting collaboration entry points, context processing, agent execution and multiple devices into an enterprise AI system.