Alibaba has rolled out a serious improve to its flagship shopper AI software, Qwen App, marking a decisive shift from conversational AI towards what the corporate calls “AI that acts” — programs able to finishing real-world duties finish to finish.
The brand new model integrates deeply with Alibaba’s core shopper companies, together with Taobao, Taobao Instantaneous Commerce, Alipay, Fliggy, and Amap, permitting customers to order meals, e book journey, make funds, and coordinate multi-step duties straight inside a single AI interface. The options are actually open for public testing in China.
“AI is evolving from intelligence to company,” stated Wu Jia, Vice President of Alibaba Group, on the product launch. “We’re shifting from fashions that perceive to programs that act — deeply related to real-world companies.”
Throughout a reside demo, Wu requested Qwen to order 40 cups of bubble tea from an area retailer. The app routinely positioned the order via Taobao Instantaneous Commerce, utilized reductions, accomplished fee in-chat, and scheduled supply — all with out switching apps.
For Alibaba, the improve represents greater than a characteristic launch. It alerts a strategic wager on turning giant language fashions into operational infrastructure embedded straight inside commerce, funds, logistics, and concrete companies.
From Chatbot to Agent
In contrast to conventional chat-based assistants, the up to date Qwen App can execute full workflows from a single instruction. Customers can place meals orders with real-time promotions utilized, plan multi-stop journey itineraries through Fliggy and Amap, full in-chat funds via Alipay, and even make actual cellphone calls to eating places to verify reservations.
A newly launched “Job Assistant,” at present in invite-only beta, expands these capabilities into productiveness {and professional} situations. Based on Alibaba, it could possibly course of as much as 100 paperwork concurrently, generate structured analysis reviews, conduct aggressive evaluation, construct light-weight internet instruments, and coordinate advanced workflows throughout a number of companies.
The Job Assistant is constructed on Qwen’s newest basis fashions and a multi-agent structure designed to orchestrate completely different instruments autonomously.
This mirrors a broader world pattern. Google just lately introduced AI purchasing integrations with main retailers, whereas OpenAI launched “Instantaneous Checkout” in late 2025 for U.S. customers. The trade consensus is more and more clear: AI assistants are shifting from chat interfaces towards autonomous brokers able to finishing transactions and workflows.
Alibaba believes its benefit lies in combining fashions with a completely built-in shopper ecosystem — spanning funds, commerce, journey, mapping, and native companies — reasonably than counting on third-party platforms.
Chasing “Completion,” Not Simply Site visitors
China’s shopper AI market has largely been dominated by ByteDance’s Doubao, which surpassed 100 million day by day energetic customers by specializing in light-weight, high-frequency situations akin to homework checking, gaming guides, and way of life suggestions — notably amongst mass-market and household customers.
Qwen has chosen a special path.
Relatively than optimizing for DAU, Qwen’s inside metrics prioritize whether or not customers actually hand duties over to AI — and whether or not the AI efficiently completes them.
Internally, the Qwen group categorizes person wants into two main domains: “work” (delivery-oriented duties) and “life” (personalised day by day companies).
For work situations, Qwen has recognized greater than 100 core productiveness use circumstances — akin to bill group, PPT era, and market analysis — and applies focused optimization to make sure excessive completion charges.
“If seven or eight out of ten requests could be delivered efficiently, customers really feel the product is principally usable,” a Qwen product supervisor instructed LatePost. “If solely three or 4 succeed, customers will abandon it shortly.”
The purpose is both “excellent supply” — the place outcomes require no modification — or no less than “principally usable,” the place solely minor edits are wanted. Predictability and reliability matter greater than novelty.
For all times situations, personalization turns into the important thing differentiator. As an alternative of merely recommending merchandise, Qwen makes an attempt to know the underlying intent behind requests. For instance, when a person says, “I wish to purchase a mite remover,” the system evaluates whether or not spraying, solar publicity, or various options could higher match the family — factoring in youngsters, pets, and private habits — earlier than recommending merchandise.
This emphasis on intent translation displays Alibaba’s broader ambition: utilizing AI to bridge human wants and complicated service ecosystems, reasonably than merely surfacing content material.
Embedded Funds because the Hardest Downside
One of the vital technically advanced upgrades is native AI fee, enabled via deep integration with Alipay. After specific person authorization, Qwen can execute funds straight inside conversations.
Past easy checkout, Alibaba is experimenting with asynchronous and conditional funds — for instance, directions like “Monitor Shanghai–Beijing flights and routinely buy when the worth drops beneath RMB 800.”
This requires constructing a four-party belief and authorization chain between customers, AI brokers, retailers, and the fee platform — a functionality Alipay itself has by no means deployed at scale earlier than.
Sooner or later, such mechanisms may allow cross-platform, multi-step buying situations, akin to organizing catering and drinks for big group occasions with advanced constraints.
A Startup Inside a Large
Behind the product, Qwen has advanced into one thing nearer to a startup working inside Alibaba.
Since late 2025, greater than 1,000 engineers and product workers from Quark, Amap, Taobao, Alipay, and different items have been mobilized right into a centralized growth effort at Alibaba’s Xixi campus in Hangzhou. Groups are organized by undertaking reasonably than perform, with product managers, engineers, and algorithm specialists sitting collectively to speed up iteration.
Qwen has reportedly shipped dozens of updates inside two months, sustaining a launch cadence of two to a few updates per week — unusually quick for a big company.
Wu Jia hardly ever conducts conventional slide shows. As an alternative, product discussions typically occur round whiteboards, the place groups co-design options in actual time.
A number of Alibaba veterans describe the shift as the corporate “concentrating its forces to combat a decisive battle” in AI — embracing startup-style execution to compete in a quickly accelerating world race.
A Completely different Development Curve
Qwen formally launched on November 17, 2025, when Doubao’s DAU had already exceeded 60 million. Competing head-on for mass visitors was by no means reasonable, and Alibaba doesn’t management China’s largest shopper visitors gateways.
As an alternative, Qwen initially attracted builders, researchers, and enterprise customers who’re extra delicate to mannequin high quality and complicated activity execution. Almost half of Qwen’s present customers are underneath 40, with the bulk concentrated in Tier-3 cities and above, and most customers have interaction with the app for specific productiveness functions.
The expansion technique is obvious: set up excessive activity completion reliability in skilled situations, develop into daily-life companies utilizing Alibaba’s ecosystem benefits, and regularly broaden towards mainstream customers.
Qwen surpassed 100 million month-to-month energetic customers inside two months of public beta, in line with Alibaba.
Notably, Qwen intentionally avoids assigning a set persona or character design. “When you outline a persona, you restrict person expectations,” a product supervisor stated. “Our superb assistant ought to transcend gender, age, or cartoon id.”
From Mannequin Management to Actual-World Infrastructure
Powered by Alibaba’s newest Qwen3 basis mannequin, Qwen App represents Alibaba’s try to convert mannequin functionality into real-world infrastructure — embedding intelligence straight into fee flows, logistics networks, service provider programs, and concrete companies.
The strategic logic is straightforward however bold: within the agent period, whoever controls each intelligence and execution pipelines features structural benefit.
As AI assistants transition from leisure instruments to operational programs, the competitors could now not be about who builds the neatest mannequin — however who can reliably flip intelligence into motion at scale.
For Alibaba, Qwen is turning into that wager.
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the newest breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s developments at this time: learn extra, subscribe to our publication, and change into a part of the NextTech group at NextTech-news.com
