AI News

The first year of the AI e-commerce battle: Qwen and Doubao lead, six major platforms compete for new growth in Double 11

36Kr · AIUpdated Nov 16 19:16

AI summary

Qwen Office released an e-commerce edition, and Doubao launched "Ask Doubao on Double 11". Six major platforms—Alibaba, ByteDance, JD.com, Tencent, Kuaishou, and Tmall—concentratedly rolled out AI shopping guides and business operation tools during Double 11. JD Jingmai AI Business Center deployed more than 60 tools, digital human livestreaming covered over 90,000 merchants, and Jingdiandian reduced material costs by 99%. Taobao's general knowledge Query grew 50.4%, and Token was sold as a product on Tmall for the first time.

Body · AI translation

"Double 11" has reached its 18th year, and it is no longer just a shopping festival.

In 2009, it was just a clearance attempt by 27 brands, with 50 million in transaction volume—not even a fraction of what a mid-sized livestream does today. Later it became a nationwide midnight vigil of revelry, then a discount-stacking game so complicated you needed scratch paper to calculate it, and later still it stretched into a month-long endurance contest jointly played by platforms, merchants, and consumers.

Over eighteen years, people have witnessed the heart-racing midnight flash sales and also endured the exhaustion of padding orders late into the night. "Double 11" is like a mirror, reflecting the entire trajectory of Chinese e-commerce from wild growth to deep cultivation of existing stock.

If we rewind to around 2011, the e-commerce wars were once a key touchstone for testing mobile payments. Alipay produced the first mobile payment QR code, WeChat Red Packets ambushed during Spring Festival, and the ride-hailing subsidy war burned payment habits into ordinary people's daily lives. Only after that round of competition did mobile payments truly turn from a tool into infrastructure. Today, e-commerce mega-promotions once again stand at the forefront of technological validation: only this time, what is being tested has shifted from payments to AI.

Starting in mid-October, Alibaba, ByteDance, JD, Tencent, Kuaishou, and Tmall almost simultaneously made their moves. Qianwen Office released an e-commerce edition, Doubao put "Ask Doubao for Double 11" on its navigation bar, Jingmai AI Business Center rolled out more than 60 tools, WeChat Mini Shop paired an AI product-selection assistant the first time it joined "Double 11," Kuaishou split AI into a 7×24 business operations team, and Tmall's Business Manager was upgraded to a version with a "command center." Six directions, the same time window.

In past "Double 11" events, AI was mostly hidden in the background: writing detail pages, polishing customer service scripts, editing short-video materials. This year is different—AI is stepping forward, participating in shopping guidance, business operations, and decision-making. The cards are already on the table; what comes next is to see what old problems each of these products solves, and what new troubles they encounter in their new positions.

The old problems of "Double 11": tiring to buy, hard to sell, slow to grow

With AI collectively entering the field this year, looking back, there are several accounts from past "Double 11" events that have never been squared.

First, on the platform side, after traffic peaked, the rules piled up ever more complex. In the past, platforms competed over who could concentrate user attention within a few days: cross-store discounts, deposit expansion, category coupons, store coupons, and membership coupons layered on top of one another, with rules so complex that guide bloggers had to break them down item by item. Complexity itself was not the goal; complexity was meant to create the feeling of "getting a bargain." But as active user growth slowed, the marginal returns of this playbook grew thinner and thinner. Users got tired of calculating, and platforms also discovered that growth manufactured through rules cannot retain people.

Second, on the merchant side, mega-promotions became a drain on manpower. A merchant operating simultaneously on Taobao, Douyin, JD, and Pinduoduo has operations staff switching between different backends every day: Shengyi Canmou for traffic, Qianniu for orders, Wanxiangtai for ad delivery, Youzan for private domain, and then exporting reports for manual consolidation. During mega-promotions, material production capacity is added on top—two hundred products in a new season, tens of thousands of seeding images, and a designer already at the limit making a hundred images a day; by the time the materials are ready, the peak season is already half over. Customer service goes without saying: once the limited-time free tools during consultation peaks expire, labor costs immediately bounce back. It is not that merchants do not want to use AI; it is that past AI either could only write copy and make spreadsheets, or could not connect to backend data—it looked lively, but once actually used in operations, it broke down.

Finally, on the consumer side, decision costs keep rising. Too many choices, rules too convoluted, and the same product bouncing back and forth across different platforms, different specifications, and different post-coupon prices. Users nominally have more choices, but in reality they are led by the platforms' marketing rhythm. Comparing prices requires switching between three apps, calculating coupons requires a calculator, and padding orders requires forcing in a few items that were not needed in the first place. Consumption decisions have become an arithmetic problem rather than a pleasant choice.

Platforms are calculating the traffic account, merchants are calculating the manpower account, and consumers are calculating the discount account. When all three accounts are tallied, they point to the same problem: the efficiency of "Double 11" has been stretched to its limit.

Alibaba chose to enter from cross-platform operations. On October 14, Qianwen Office released its e-commerce edition, positioned as an independent third-party e-commerce Agent that does not require merchants to pick a side first. Its four professional functions are broken down very concretely: a cross-platform operations dashboard places one's own and peers' data side by side, trend insights and selling-point analysis capture direction from social hot words, viral materials are generated in batches, and cross-platform ROI intelligent optimization dynamically diagnoses ad delivery. It has already connected to tools such as Shengyi Canmou, Qianniu, Youzan, and Alimama Wanxiangtai; when merchants switch to "e-commerce operations" mode, they can complete product selection, materials, ad delivery, and review on one workbench. This approach hits the data silos of multi-platform merchants; at this stage, it tests permission, security, and trust mechanisms in cross-platform collaboration more, and smooth integration of account systems still needs time to be worked out.

ByteDance's path is more thorough, directly turning Doubao into a channel. Doubao follows a complete channel path: connecting to Douyin Mall product cards, launching "Doubao helps you choose," incorporating "APP source = Doubao" into independent settlement, and setting up a Doubao operations zone in the Douyin Store backend. This year's "Double 11," Doubao set up a dedicated "Ask Doubao for Double 11" navigation bar, and consumer coupon distribution covers Doubao and other product endpoints. The advantage is a complete closed loop from recommendation to order and settlement; the attribution chain is still being improved, and a new channel needs a period of adjustment from running through to fine-grained accounting. At this stage, merchants can add a source confirmation line in customer service scripts to make the channel account clearer.

JD chose to cover everything with tools. At the end of September, it upgraded Jingmai AI Business Center, with more than 60 tools covering six links: store operations, business opportunity and product selection, material production, marketing and promotion, order fulfillment, and customer service reception. It also released the first batch of "Jingmai AI Expert Teams." Digital-human livestreaming covers more than 90,000 merchants, and the design agent Jingdiandian reduces material production costs by 99%. The toolbox is ready out of the box and the most convenient, better suited to single-point efficiency gains; cross-platform operations linkage still requires people to work with tools.

Tencent's focus is on the infrastructure of small shops. WeChat Xiaodian participated in "Double 11" for the first time, equipped with an AI product selection assistant, smart clipping, and livestream sales customer service, plus livestream product reservations and platform subsidies. Private-domain scenarios are naturally close to transactions, which is its advantage; the e-commerce volume of WeChat Xiaodian is still growing, and the value of the tools will be released along with the scale.

Kuaishou simply split AI into a team. AI digital employees are online 7×24 hours on the merchant workbench, the virtual store manager is responsible for the operation of the whole store, and Cili Xiaoniu understands spoken descriptions and automatically provides promotion plans. The team covers all time periods, but it has high requirements for the quality of merchants' real operating data, and the pace of implementation depends on the data foundation.

Tmall, meanwhile, turns AI into a sentinel. The upgraded Business Manager and AI War Room can automatically inspect negative reviews, monitor abnormal products in campaigns, track competitor dynamics, and predict hit-product trends. The value of the sentinel lies in defense and early warning, but how well it is used depends on whether merchants themselves have the actions to respond to the warnings.

With the six cards laid out, the differences are clear: Alibaba and ByteDance are building colleagues that can do work, JD.com and Tencent are handing out toolboxes, Kuaishou is building a team, and Tmall is building a sentinel. Toolboxes are the least troublesome, colleagues save the most time, and teams and sentinels are the hardest to implement because they need real operating data. No card is omnipotent; the key is where the merchant's own pain points lie.

Is the fight for incremental growth, or for the next-generation e-commerce entrance?

On the surface, this year's "Double 11" is a concentrated appearance of each company's AI products. But looking deeper, two things are happening at the same time behind this.

On the one hand, AI is pulling incremental growth, and the increment comes from "transactions that would not have happened otherwise." Tmall President Jialuo gave a set of data at the press conference: an active consumer buys about 15 categories on Tmall per year on average, while normal consumer demand may be 25 to 30. A large amount of demand is dormant—a person who loves hiking may have a need to protect their knees, but he does not know that something like a "patellar strap" exists. Search can only capture already explicit demand, and it is hard for people to search for things they do not yet know. What AI needs to do is match products to people who "may need them but have not yet thought of them." On JD.com's side, the AI shopping assistant "Dongdong" and the digital human JoyAI, which enables real-time video shopping, are also shortening the decision-making chain. Taobao's general knowledge Query grew 50.4%, indicating that consumers have begun to treat e-commerce apps as a place where they can ask questions, not just a place to enter product names. These changes point to the same judgment: what AI brings is not just an increase in conversion rate, but a batch of transactions that would not have happened otherwise being awakened.

On the other hand, AI is reconstructing the playbook, and both the entrance and the distribution logic are changing. In the past, merchants fought for keyword exposure; when users searched for "dress," all merchants competed for rankings around that word. Now when users search for "what to wear for a seaside vacation," AI breaks the demand into a combination of category, style, material, function, and budget, and all products suitable for this scenario have a chance to be seen. The unit of traffic distribution has changed from "words" to "scenarios," and what merchants need to fight for is no longer keyword ranking, but whether their products can be understood by AI as the answer to a certain demand. At the same time, Token was placed on the "Double 11" shelf for the first time. Model vendors such as Zhipu, MiniMax, and Qwen opened flagship stores on Tmall to sell packages, and AI services themselves became commodities. This means the boundary of participation in the major promotion has extended from physical consumer goods to digital means of production. Selling Token today is like selling funds and cloud servers during "Double 11" back then—both are pushing new things in front of the public for the first time.

But the challenges are equally glaring. First, the attribution chain still needs to be completed; AI-recommended orders are mixed with natural transactions, and incremental accounting still needs channels and merchants to improve together. Second, the cost of review is easily underestimated; machine generation is fast, but manual verification still takes time, and the efficiency bottleneck ultimately falls on people. Third, the trust mechanism is still being explored; commercial labeling and consumer discernment in conversational recommendations still need clearer norms. Fourth, permissions and boundaries need to be clarified; when Agents participate in budgets, products, and orders, accuracy, auditing, and responsibility division still need to be polished jointly by the industry.

So for AI, this year's "Double 11" is not a simple feature display, but a public extreme stress test. With the real operations of millions of merchants, advertising budgets often in the hundreds of millions, and various emergencies, whether AI can withstand it and truly help merchants settle that account—once "Double 11" is over, some of the answers will emerge.

In closing

The drumbeat of "Double 11" has sounded for eighteen years. From 27 brands to millions of merchants, from midnight flash sales to a month-long run, from discount arithmetic problems to official direct discounts, each edition answers the same question: how to make transactions more efficient.

This year's new answer is AI. Six major companies submitted their papers at the same time; some built colleagues, some handed out toolboxes, some built teams and sentinels. The cards are different, but the direction is the same: let AI move from the back office to the front stage, from an auxiliary tool for writing copy and making spreadsheets to a production tool that penetrates real operating links.

But AI is not magic. It can help you compress four months of material preparation into fifteen days, help you fish out buried potential products from more than four hundred products, and help you guess what users may need before they speak. It cannot judge for you what counts as a good image, cannot bear for you the inventory risk of choosing the wrong product, and cannot cover for you when AI recommendations go wrong. Machines run processes and improve efficiency, while people guard judgment and warmth—this is probably the more realistic way of collaborating at this stage.

In the 18th "Double 11," what is really being fought over may no longer be traffic, but who understands demand better. AI has built the shelf where the money flows, but whether opportunities can be turned into good business ultimately still depends on products, prices, and services. Whether it is a free carnival or paid intelligence, calculating how many hours those fees actually bought back is what merchants care about most when they look at the bill at the end of the month.

The AI exam of "Double 11" has already handed out the papers, and the results remain to be revealed.

This article comes from the WeChat public account "Di Yi Xin Sheng," author: Di Yi Xin Sheng, published by 36Kr with authorization.

Source

36Kr · AI