Qwen

Alibaba Upgrades Qwen App Into a Task-Doing AI Agent

Alibaba's Qwen app can now order food delivery, book travel, and hunt deals as an agent powered by Qwen3.8, wired into Alipay, Fliggy, and Amap. Inside China's consumer agent war at 100 million MAU.

Alibaba Upgrades Qwen App Into a Task-Doing AI Agent — article cover
On this page6 SECTIONS
  1. From Chatting to Doing
  2. Plugging Into the Alibaba Ecosystem
  3. 100 Million Users and China’s Agent War
  4. The “Physical World” Bet
  5. What It Means for Product Teams
  6. Sources

On January 15, 2026, Alibaba held a launch event in Hangzhou and upgraded its Qwen app from a chatbot into a personal AI agent: it can now order food delivery, book travel, and search out the best deals on your behalf, executing tasks end to end instead of just answering questions. Reuters reported the upgrade as Alibaba’s most aggressive move yet into AI agents, following similar launches by domestic rivals. The front line in China’s consumer AI market has moved from conversation quality to who can actually get things done.

The move matters beyond China. Wiring payments, travel, and maps directly into an agent loop — at hundred-million-user scale — is a first for a major platform, and it previews the product and monetization questions every consumer AI company is about to face.

From Chatting to Doing

The core change is task execution. Say “order dinner” or “train tickets to Harbin next week,” and the agent searches options, compares prices, and places the order rather than returning links for you to click through. SCMP reports the update spans everyday scenarios — shopping, travel bookings, food delivery, ride-hailing, and movie tickets — all drivable by voice. The control surface is a chat window: users can follow each action the agent takes, mark recurring actions as trusted so they run without asking again, and stop the agent at any point. When tasks involve real money, that visibility-plus-interruptibility combination is the baseline — and the most consequential design surface in products like this.

Plugging Into the Alibaba Ecosystem

The agent’s transactional reach comes largely from in-house rails: payment runs through Alipay, travel bookings through Fliggy, navigation and ride-hailing through Amap. For small and medium businesses, the app also ties into Alibaba.com, extending the agent to cross-border sourcing. Reuters’ analysis flags the strategic point: if this works, the Qwen app becomes a significant monetization channel for Alibaba, since every delivery and booking completed by the agent is an in-ecosystem transaction. The agent stops being a cost center and becomes a transaction gateway. Rather than routing traffic out to other apps, Alibaba is letting Qwen grow into the entry point itself — the super-app playbook rebuilt for the AI era.

100 Million Users and China’s Agent War

The scale milestone landed at the same event: Wu Jia, the Alibaba vice-president who heads Qwen’s consumer business group, said the app passed 100 million monthly active users as of January. The public beta only opened in mid-November 2025; in November alone it grew monthly actives by 149%, the fastest of any AI app globally that month, according to tracker Aicpb.com — roughly two months from open beta to a hundred million users, one of the steepest growth curves China’s consumer AI market has produced. Competition is fierce: QuestMobile data puts ByteDance’s Doubao at roughly 150 million MAU as of October, model rivals such as DeepSeek and service platforms like Meituan and Ctrip are fighting for the same users, and SCMP cites Counterpoint analyst Ivan Woo’s view that China’s crowded AI agent market will start consolidating in 2026. Alibaba’s leverage on the model side is openness: China Daily reports the Qwen family is the world’s most-downloaded open model family, the foundation underneath its consumer push.

The “Physical World” Bet

Wu Jia framed the vision bluntly at the event: an assistant’s ability to enter the physical world matters as much as its digital skills, and “it should help people get things done in everyday scenarios.” That sentence is Alibaba’s definition of consumer AI in one line — not chattier, but more capable of doing. The timing ahead of Lunar New Year travel season is no accident: holiday bookings are China’s highest-frequency transaction moments, the ideal stage to demonstrate that an agent is worth trusting — and a high-stakes stress test if anything goes wrong.

What It Means for Product Teams

Three lessons. First, agent UX is fundamentally “action visibility plus interruptibility” — that builds trust faster than conversational polish. Second, ecosystem depth decides agent utility: a general model without transaction rails cannot literally order food; connecting APIs is table stakes, while payment and fulfillment integration is the moat. Third, China is the live rehearsal for consumer agent competition everywhere — the consolidation and monetization squeeze after the 100-million-MAU mark is exactly the agentic battleground we sketched in our 2026 opening outlook. The metrics to watch from here are simple: task completion rates, repeat usage, and how much agent-driven transaction volume Alibaba is willing to disclose next quarter.

Sources

AI-assisted summary compiled from the sources above, reviewed by a human before publishing.

SHAREXEMAIL