As the Double 11 shopping festival enters a protracted phase, have major tech companies made peace with AI?

Over 600 years ago, Zhu Yuanzhang seized his opportunity amid the era of warlords vying for dominance by following the strategy of "build high walls, accumulate ample provisions, and claim the throne at a leisurely pace". E-commerce platforms today certainly do not need to copy the ancient military tactics word for word, but in the current context of slowing traffic growth, fierce price competition and increasingly rational consumers, these nine characters still hold practical significance.

The e-commerce industry has been "held at gunpoint" by AI for two full years.

Two years ago, the narratives around AI were extremely grand. Some people claimed that the search box would soon become a thing of the past; others predicted that consumers would no longer need to open Taobao, JD, or Douyin — they only needed to say "help me buy a washing machine" to a dialog box, and AI would search for products, check reviews, compare prices, collect coupons, and finally complete the order directly for them.

If that day really comes, the business logic that e-commerce platforms have painstakingly built over the past decade — search ads, paid ranking, traffic distribution, merchant operations — seems to be uprooted entirely. Whoever controls the AI entry can intercept users; whoever first turns AI into a shopping entry may reshuffle the entire industry.

As a result, AI was described as the new Imperial Seal that symbolizes sovereign power. All platforms were waiting to see who would be the first to hold this seal above their head during this year's Double 11 shopping festival.

But when the big promotion actually kicked off, things did not go according to the script. The anticipated platform war never broke out, and AI did not suddenly kick open the door to the old world. It is more like a new colleague assigned to different departments by every platform: some are arranged to help users make choices, some calculate inventory in warehouses, some boost conversion in content scenarios, and some even walk into factories to study how the next batch of products should be produced.

This does not look like a revolution at all, but rather a redivision of labor.

Lujiu Business Review believes that the most noteworthy point of this year's Double 11 is not how stunning AI functions any platform has released, but that the leading tech giants have changed their attitude towards AI: from the "who will disrupt whom" mindset two years ago, to today's "who can make good use of AI". This reconciliation has at least three layers of meaning: recognizing the capability boundaries of AI, re-identifying their own survival genes, and starting to recalculate the cost of competition.

If we use the suggestion Zhu Sheng gave to Zhu Yuanzhang more than 600 years ago to sum it up, it is exactly: build high walls, accumulate ample provisions, and claim the throne at a leisurely pace.

01 Build High Walls: AI Did Not Tear Down the City Walls, But Moved Inside Them

First, let's take a look at where the five platforms have positioned AI in their business.

What Taobao and Tmall most need to solve is the problem that consumers are finding it increasingly difficult to make decisions. Buying a washing machine seems to be an ordinary consumption, but before actually placing an order, users have to face a series of troubles: how to choose the capacity, whether the drying function is necessary, which brand is more durable, which parts of the reviews are real user experiences and which are just marketing rhetoric, and which model is truly the cheapest after applying all overlapping discounts. There is no shortage of information, but what is really scarce is the ability to organize information into clear answers.

Therefore, Taobao AI acts more like a shopping decision assistant. Users clarify their budget, family situation and usage needs, and AI helps filter products, compare parameters, summarize reviews, and sort out all discount benefits. For merchants, AI can also participate in hit product prediction, product diagnosis and price management. Its goal is not to turn Tmall into another chat app, but to make existing transactions easier to complete and reduce the hesitation of brand users.

JD's focus is different. When buying a small commodity, consumers may care more about the price; when buying refrigerators, air conditioners and washing machines, what really affects the experience is often whether the product can be delivered, when it can be delivered, whether the old machine can be taken away, and whether the installation is timely. When the number of orders surges during big promotions, inventory, warehouse allocation, delivery and trade-in services will form a complex collaborative test.

These things are not necessarily suitable to be displayed in front of consumers, but they can directly determine the success or failure of a transaction. If AI can predict in advance which regions will see rising demand, help warehouses allocate inventory, arrange delivery capacity, and coordinate old machine recycling, it will further consolidate JD's supply chain advantages. It may not make users feel "this is so smart", but it may make users feel "this is so hassle-free". For JD, that is exactly the value.

Douyin faces a completely different problem. Short videos are good at stimulating interest, but not necessarily good at meeting clear shopping demands. A user scrolling through a short video may be instantly attracted by an air fryer; but when he really wants to buy it, he still needs to search for the model, compare prices, and check reviews. If interest-driven traffic cannot be converted into stable shopping behavior, it is just like water flowing through the fingers — no matter how lively the scene is, it may not precipitate into real business.

AI can help merchants generate and filter content materials, and can also identify signals that users are shifting from "casual browsing" to "serious comparison", connecting content scenarios with transaction scenarios such as mall and search. What Douyin wants to achieve is to make users not only place impulsive orders after watching a video, but also come back to find the products they need when they have clear demands. In the final analysis, it wants to repair the easily broken link between product discovery and transaction completion.

Pinduoduo's story may be the farthest from consumers' screens, but the closest to the source of business. For factories in industrial belts, the most headache problem is not necessarily the inability to produce, but not knowing what to produce and how much to produce. If the finished products cannot be sold, the inventory will be piled up in the warehouse; when they follow up after seeing a certain product selling well, they may miss the best sales window.

If the transaction data accumulated by the platform can help factories test product performance, judge market demand and arrange production schedules, it will help reduce trial and error costs and inventory backlogs. Pinduoduo's low-price competition seems to be a price competition on the surface, but at its core it is a competition for supply chain efficiency. Where AI can really play a role is not just to lower the price on the page a little bit, but to reduce waste between production, inventory and demand.

Xiaohongshu's core priority is to maintain trust. Users go there to search for a skin care product, a restaurant or a camera, often not just wanting to find the cheapest offer, but also wanting to know "what is the actual experience of ordinary people after using it". When commercial content keeps increasing, real experiences and carefully packaged advertisements are mixed together, the trust that the platform relies on for survival will be consumed little by little.

AI can help identify content semantics, judge the type of notes, and assist the platform to spot spam content and unnatural marketing expressions. But there is also a boundary here: algorithms can assist identification, but they cannot automatically equal the truth. What Xiaohongshu really needs to guard is not the myth that "AI can identify everything", but to ensure that users still believe that the experience sharing on this platform is of reference value.

The five platforms take different approaches, but the underlying logic is the same: they do not treat AI as a hammer to tear down walls, but as a new tool to reinforce their own walls.

In the past, some people believed that once new technology arrives, the accumulated advantages of old platforms will be reset to zero instantly. This judgment ignores a fact: Technology itself does not automatically form a business model. The platform's supply chain, merchant network, content ecosystem, user habits and service capabilities are the "body" formed by long-term operation; AI is more like a newly grown hand. The hand can be more dexterous, but it must obey the command of the body and know its own reach limit.

When platforms no longer see AI as a disruptor that must be eliminated, but integrate it into their own business chain, the so-called reconciliation begins to happen.

02 Accumulate Ample Provisions: Traffic Is Getting More Expensive, First Figure Out Your Own Business

If the first layer of reconciliation is to re-understand AI, the second layer of reconciliation is to re-understand yourself.

Looking back at past Double 11 events, the most common move of platforms is to launch offensives: full reductions, cross-store discounts, red envelopes, subsidies, live streaming, cross-platform price comparison. The battle line keeps extending, and the gameplay keeps getting more and more complex, as if as long as the public attention is pushed high enough, more consumers can be drawn in, and growth can be found from a larger transaction scale.

But the market will not always cater to this imagination. According to the data listed in the original text, from January to August 2026, the national online retail sales of goods increased by 4.3% year on year, among which the growth rate of related physical product categories is even lower; the growth rate of the total transaction volume across the network during this year's 618 shopping festival also slowed down significantly. Behind these figures is a reality that the e-commerce industry is increasingly familiar with: The traffic dividend has not completely disappeared, but it is no longer as easy to obtain as it used to be.

Consumers are better at calculating costs, and merchants also pay more attention to input-output ratio. Subsidies can make users place orders in the short term, but they do not necessarily bring long-term loyalty; a grand marketing campaign may bring a lively performance report, but it may not leave healthy profits. When the ceiling of growth gradually emerges, continuing to charge outward is no longer the only solution.

As a result, AI has become a mirror.

Tmall sees that brand users need lower decision-making costs when facing complex products; JD sees that fulfillment efficiency and service certainty are still its long-term strengths; Douyin sees that there are still breakpoints between content popularity and real transactions; Pinduoduo sees that industrial belts still need more accurate matching between supply and demand; Xiaohongshu sees that once trust is eroded by excessive marketing, it is very difficult to buy it back through a single campaign.

This is why the same general-purpose technology will take completely different shapes after entering different platforms. What really determines how AI is implemented is not the imagination at the press conference, but the specific problems that every enterprise has to face every day.

This also means that Double 11 is evolving from a competition of "who can make more noise" to a multi-week business stress test. Platforms need to pay attention not only to how many people place orders at a certain point in time, but also to whether the products are selected correctly, whether the inventory is overstocked, whether the fulfillment can keep up, whether merchants can make profits, and whether users will come back.

In the past, when people talked about "accumulating ample provisions", they easily thought of cash, inventory and resources; in today's context, what is really worth hoarding also includes high-quality transaction data, stable merchant relationships, efficient operation links, and service experiences that consumers are willing to use repeatedly. Hype can be manufactured, but operational capabilities can only be accumulated little by little.

03 Claim the Throne at a Leisurely Pace: Whoever Grabs the Entry First Does Not Necessarily Laugh Last

Why do we still need to "claim the throne at a leisurely pace"? The story at the end of the Yuan Dynasty may provide an interesting reference.

Chen Youliang had a strong army and showed his ambition very early; Zhang Shicheng occupied the rich and prosperous Jiangsu and Zhejiang regions, with sufficient money and food, but gradually got trapped in his comfort zone; Zhu Yuanzhang did not have an advantageous starting point, but he did not rush to show all his cards at once. He first stood firm, then judged who was the most dangerous opponent in front of him, and waited until the time was ripe to gradually complete the unification.

Of course, business competition is not historical war, and platforms are not ancient warlords. But this story at least reminds us: Having the most powerful weapon at a certain moment does not mean that you have already mastered the conditions for final victory. Technological leadership can create opportunities, but it cannot replace organizational capabilities, supply efficiency, user trust and business discipline.

Over the past two years, there has been a strong sense of urgency in AI competition: if I don't do it first, others will; if I don't grab the entry, others will take it away. This anxiety is not completely unreasonable, but once everyone takes "grabbing the entry" as the only goal, competition will easily turn into redundant construction and high-cost consumption. Everyone is chasing the same story, but they may forget what they originally relied on to make money.

Last year's food delivery war has already let the industry see the cost of this kind of gameplay. Subsidies, traffic and entry competition can create momentum in the short term, but they do not necessarily bring sustainable advantages. When competitors can all follow up with subsidies, and consumers can switch to another platform at any time, it is worth recalculating whether the money spent has bought user loyalty, or just a few days of temporary hype.

During this year's Double 11, platforms are more likely to draw their attention back: instead of rushing to announce that they are going to rewrite the industry, it is better to first optimize the most difficult, most expensive and most error-prone links in existing businesses. Instead of blindly creating impulsive consumption, it is better to build operational certainty. The former is to seize other people's users, while the latter is to give your own users a reason to stay.

This does not mean that there is no competition between platforms anymore. On the contrary, competition may become more detailed, deeper, and harder to be summed up by a single slogan. What will be competed in the future is not necessarily whose AI assistant is best at chatting, but who can understand user demands more accurately, deliver products more timely, reduce supply chain waste more effectively, and better protect real content. On the surface, everyone is working on AI; at a deeper level, everyone is competing for the irreplicable parts of their own business.

Of course, reconciliation also has its cost. When AI changes from a "disruptor" to a "builder", it is no longer easy to tell a grand story that can excite everyone. The capital market will eventually ask: How much has the efficiency improved? How much has the cost been reduced? Has the conversion rate improved? Have users stayed? If the answer cannot stand the test of operational data, no matter how beautiful the AI narrative will fade.

But this is not necessarily a bad thing. Technology ultimately has to return to business to be tested. For enterprises, instead of continuously increasing investment just to prove that they are standing at the forefront of the trend, it is better to admit that some problems cannot be solved by a single new entry, and some advantages cannot be replicated by just releasing a new model. Knowing what you can do and what you cannot do, knowing when to attack and when to accumulate, is often more important than rushing to claim the throne.

During this year's Double 11, AI did not break down the door of the e-commerce industry, but was invited into the offices, warehouses, content scenarios and factories of different platforms. It seems a little less dramatic, but much more real for business.

Over 600 years ago, Zhu Yuanzhang seized his opportunity amid the era of warlords vying for dominance by following the strategy of "build high walls, accumulate ample provisions, and claim the throne at a leisurely pace". E-commerce platforms today certainly do not need to copy the ancient military tactics word for word, but in the current context of slowing traffic growth, fierce price competition and increasingly rational consumers, these nine characters still hold practical significance.

AI will not automatically claim the throne on behalf of any platform. What it can do is to help a company that already knows where it is going to walk more steadily and further.

In the final analysis, technology is never the answer itself. Whether you can turn technology into your own capability is the real answer.

This time, leading e-commerce giants are not in a hurry to let AI rewrite everything. They first let AI learn how to work in their own world. Lujiu Business Review believes that this is not the end of the revolution, but that competition has finally returned to the essence of business.

This article is from the WeChat official account "Lu