One month after the launch of Alibaba's AI office product "QwenWork," it has delivered impressive results: user count exceeded 30 million, with more than half being enterprise users, marking a surge in enterprise-level AI applications.

Over the past 30 days, the team maintained a high iteration efficiency, completing a total of 120 version updates, averaging four iterations per day, adding thousands of new features, and launching the international version simultaneously. On the technical side, QwenWork open-sourced its enterprise context infrastructure MyContext, which converts heterogeneous data scattered across systems such as IM, documents, approvals, and emails into context that can be called by Agents. This project has already received over 3,000 stars on GitHub, and is deeply integrated with DingTalk. Users just need to speak one sentence, and the Agent can query customer information, inventory, and provide suggestions across CRM, ERP, and other systems.

QQ20260904-110107.jpg

Regarding performance, QwenWork jointly launched an office-specific model, Qwen3.8-Flash, optimized for multi-step planning and context compression. Testing shows that single-task generation speed increased by about 100%, and average token consumption decreased by 75%. Currently, leading enterprises such as Changan Automobile, Huifu International, and Laoxiangji have integrated this system. At Changan Automobile, the previously manually completed wire harness selection calculation, which took two days, was shortened to 5 minutes, and the vehicle document verification was compressed from tens of minutes to 1-2 minutes. QwenWork stated that it will continue to refine the product, making AI truly penetrate the core business of enterprises and organizations.