Exclusive: Alibaba mobilizes nearly 100 developers to build the Qianwen office e‑commerce version, slated for Singles’ Day pre‑sale.

LatePost Exclusive | Alibaba organizes nearly a hundred people to develop Qianwen Office e‑commerce version, plans pre‑sale on Double Eleven..

Person in charge Huang Pibei led the incubation of Huiwa in 2023, one of Alibaba’s earliest independently developed AI e‑commerce applications, producing marketing content for merchants and influencers across multiple platforms. Early features included virtual try‑on, virtual models, product images and copywriting, later adding batch image generation, detail‑page creation, image‑to‑video and intelligent remixing. As of September this year, Huiwa has 1.5 million users and an ARR (annual recurring revenue) of about 50 million RMB.

Qianwen Office E‑commerce Edition continues the multi‑platform approach: merchants can directly upload product, clothing, and model assets; the operation dashboard aggregates data from multiple platforms and stores, showing anomalies and recommended actions; AI spreadsheets organize product reviews, issues, and next steps.

Data from Taobao and Tmall can be connected directly through Alibaba’s internal systems; external platforms such as Youzan and KeRuyun are being gradually integrated; other e‑commerce platforms require merchant‑authorized login to collect backend data.

A Qianwen Office staff member believes the reason for a dedicated e‑commerce version is that merchants don’t want to keep adding disparate tools—they need a unified Agent work interface redesigned around products, assets, stores, and operational analysis. It can continuously understand a merchant’s brand, products, store, and operational data, and may eventually become the entry point for a merchant’s daily work.

In recent years, AI e‑commerce has been exploring two paths: on the merchant side, the commercial value is clear and focuses on cost reduction; on the consumer side, the focus is on recommendation and price comparison, but progress has been limited.

Earlier this year, Taobao and the Qianwen app (Alibaba’s consumer‑facing personal AI assistant) further integrated. Users can use the “Qianwen AI Shopping Assistant” within Taobao, or directly select and compare products in Qianwen before completing a purchase. Both sides adopt an Agent‑to‑Agent architecture — the Qianwen Agent understands user intent and decides when to invoke the Taobao Agent; the Taobao Agent provides capabilities such as consumption decisions, product information, transactions, and fulfillment.

After this year’s Spring Festival, the joint project team formed by Qianwen and Taobao began collaborating, with core management meeting every two weeks. What content the Taobao Agent returns, the standards it follows, and how it enters Qianwen’s interaction flow are jointly determined by Qianwen and Taobao.

During the 2025 Double‑11 shopping festival, Taobao also launched dozens of AI products for consumers, such as AI Shopping Assistant, AI Universal Search, AI List, AI Pick for Me, AI Lipstick Try‑On, etc. These are primarily experimental and have not noticeably driven user engagement or sales growth.

To converge entry points and improve user experience, in the first half of this year,these products were integrated intothe AI Shopping Assistant coordinated by Ding Jian, mainly including four functions: AI Q&A, AI try‑on, AI saving, and AI recommendation.

A Taobao employee told us that on the consumer side, different teams have tried repeatedly, and so far they still haven’t found a particularly new form; the direct impact on GMV is almost negligible. “We definitely can’t abandon the C‑end,” he said, “but if we keep doing it only within the original Taobao, it’s too difficult.”

This isn’t a problem unique to Alibaba. Whether it’s e‑commerce platforms or chatbots, none have created a new entry point.

Meta’s recently launched personal agent Muse demonstrates another path: users hand their needs to a personal assistant that knows them well, which then completes search, comparison, add‑to‑cart, and purchase across sites. Qianwen’s personal AI assistant “Qianwen Go!” is also about to be released. However, Amazon quickly blocked Muse, exposing the limitation of this approach: products, inventory, accounts, and transactions remain under the control of e‑commerce platforms.

From what we know, in enterprise‑level AI, Qianwen Office, Doubao Work, and WorkBuddy have all set aggressive growth targets.

The three companies each start from their own strengths: Alibaba leverages e‑commerce and DingTalk, organizing nearly a hundred people to turn its strongest e‑commerce business into a separate version; ByteDance relies on content, video models, and Feishu, keeping industry capabilities within the generic system of Doubao Work, with Seed, Seedream, and Seedance all integrated into Doubao Work to handle analysis, image, and video production; Tencent builds on its early product advantage, emphasizing an open platform that lets industry partners develop Buddy applications on the WorkBuddy foundation.

AI office is also shifting from a generic Agent feature race to a competition for each company’s existing businesses, data, and customers. For Alibaba, the first card it plays is e‑commerce.