The first AI Double 11, "cats and dogs" fierce battle global marketing

Original source: Light Cone Intelligence

Author: Liu Yuqi

Image source: Generated by Unbounded AI

This year’s Double 11, AI sang the leading role.

In the 2023 Tmall Double 11 press conference, in a 20-minute speech, Dai Shan, CEO of Taotian Group, mentioned AI 20 times, “Taobao is a product of the last Internet technology revolution, and will also seize the next AI revolution.” At the meeting, Dai Shan said that Taobao is gradually starting to test the water intelligently: “This year’s Double 11, we have prepared more than 10 AI tools for merchants to use.”

The first highlight of JD Double 11 is that the whole line is equipped with intelligent customer service “Jing Xiaozhi”, according to JD data, within 10 minutes of the sale, the cumulative number of JD Cloud intelligent customer service consulting services exceeded 1.88 million.

In addition to the two major e-commerce companies, Tencent Advertising uses the hybrid model to support intelligent advertising production; Baidu’s preferred intelligent assistant “Xiaoyou” has been launched in the Baidu APP, which acts as an intelligent shopping guide assistant for consumers and provides the “Double 11 Calendar” with full-platform statistics; Xiaohongshu will also announce an AI smart note-taking assistant during Double 11 to help merchants achieve one-click release of product notes.

Not only the platform is active, but brands, merchants, and MCN companies have also begun to form groups to explore AI tools. A senior marketing industry person believes: ** “Unlike during the Double 11 period in previous years, e-commerce competition was mainly focused on simple and crude price wars, and this year’s AI technology will set off a new round of retail operation efficiency competition.” **

The large-scale implementation of AI large models and series of product capabilities in commercial application scenarios can help retailers quickly and easily complete multiple links such as decision-making, launch, and creation, and achieve growth under more accurate business strategies.

The fifteenth Double 11 is kicking off the first “AI Double 11”.

“Fighting” intelligent customer service

At the end of 2022, several members of JD.com’s technical committee held a meeting to explore how large models can change e-commerce. This is not only a topic for JD.com, but also a must-answer question for every e-commerce platform.

**Judging from the results, the answers given by Taobao and Jingdong all point to intelligent customer service, which has become the first landing point for large models to land in the e-commerce industry. **

In the entire e-commerce chain, customer service has always been the most dependent on manual links. Traditional customer service is usually divided into self-operated and outsourced, which requires 24-hour instant and agile response, and is also the key entrance to guide users to place orders. However, due to the excessive reliance on manual labor, the customer service team has high mobility, tedious and repetitive work, low efficiency, and difficult to meet the diversified needs of customers.

Before the arrival of the large model, the logic of intelligent customer service was more like “keyword search”, when the user triggered the keyword, it would automatically reply to the configured template, but because it could not identify the user’s intent, it was often rigid and rigid, and could not solve the user’s actual problems. After the intelligent customer service is connected to the large model, through the ability to connect to the knowledge base, the originally unanswerable questions become wider through intent recognition and text generation. At the same time, the dialogue has become more professional and natural, and it can really be used by merchants to really reduce operating costs.

“Jing Xiaozhi” is an intelligent customer service tool provided by JD.com for merchants, which was connected to the Yanxi model in May this year, upgraded, and provided online testing for some merchants in 618. According to Light Cone Intelligence, at present, Jing Xiaozhi is not only for self-operated merchants but also for Pop merchants, including 3 versions - “Basic Edition”, “Light Edition (98 yuan / month)”, “Professional Edition (198 yuan / month)”.

“This price is much cheaper than the outsourced customer service team, in the background configuration of machine first, manual assistance mode, the machine will automatically answer the configured questions, such as delivery time, what logistics, delivery location, etc., the machine can not answer, it will be identified and transferred to the manual, filtering 60% of the questions”, a merchant who used Jing Xiaozhi feedback.

At the same time, through the self-developed voice semantic technology and the massive voice interaction generated by 580 million users in the retail scene, the Yanxi platform has launched the “Yanxi Intelligent Outbound Call”, which can actively and batch outbound calls, and the blessing of the large model also allows it to achieve more marketing innovation.

Since the beginning of this year, AI imitation sound technology has been extremely popular, “AI Jay Chou”, “AI Yanzi Sun” has repeatedly appeared on the hot search, Jingdong has conveniently used imitation sound technology in the private domain powder addition activity cooperated with Yili, so that hundreds of thousands of fans have received real voice holiday blessings from spokesperson idols, and the fan rate is as high as 35%. After all, who of Wang Yibo’s fans can refuse the call of “Wang Yibo”?

**The huge number of consultations on Double 11 has become a training ground for intelligent customer service, just like the challenges and experiences of cloud servers in the Spring Festival Gala that year. **

Obviously, Taobao will not give up this opportunity either. In September, which is approaching Double 11, Taobao announced that it will connect Tongyi Qianwen to Taobao Wenwen, which is different from Jingdong 2B’s idea that Taobao Wenwen provides “intelligent shopping guide” services for the C-end.

For example, a user asks on Taobao Ask, “What should I prepare for camping?” The AI will quickly give text answers, video introductions, product links, and users can also click on the text hyperlink or search for pictures to enter the secondary shopping page. At the end of the answer, there are also 3 related recommendation tabs, click to enter the store product detail page, and you can buy directly.

The C-end oriented intelligent customer service is more like a new entrance, and the large model intelligently screens and matches according to the needs of users, breaking the traditional shelf e-commerce link and directly recommending it to users, saving users 80% of their browsing time.

On the eve of Double 11, Taobao Wenwen also launched a new “big promotion mode”. In the promotion mode, in response to the special needs of users during the promotion cycle, such as “the most worthwhile mobile phone” and “the most cost-effective lipstick”, Taobao Wenwen will synthesize the information of big promotion discounts such as instant reduction and full reduction to generate and provide the most powerful purchase suggestions, so as to reduce the cost of user selection and decision-making, so that users no longer have to “burn their brains” for discounts.

At the same time, Taobao Wenwen has also improved its product recommendation capabilities, supported summarizing product details, user feedback, intelligently generated product evaluations, and added more abundant product display forms such as short videos and live broadcasts.

**However, on the surface, Jingdong and Taobao’s ideas are not the same, but intelligent customer service is just a “gripper”, and clues such as user needs, user habits, etc., learned through intelligent customer service, will eventually be fed back into the operation process of merchants. **

For example, another core function of Jing Xiaozhi is “full-link documentary”, including logistics tracking after delivery, facilitating orders after visits, active reminders after payment, long-term care for member users, and so on.

Taobao Wenwen uses AI to guide users to describe their needs more specifically, while deepening the semantic understanding of the platform, it also aims to further improve the matching efficiency of people and goods.

Guild Wars Marketing AIGC

After the popularity of large models, AIGC quickly became the biggest landing scenario, and “generated content” hit the shortcomings of the high cost of marketing material production that has always been in the e-commerce field.

For most small and medium-sized businesses, the cost of content production is higher, the threshold is higher, and the production of marketing materials for big promotion nodes is an inevitable expense, from the promotional copywriting and promotional posters of the circle of friends to the production of product videos, which requires both professionalism and speed.

**Some large manufacturers with technology accumulation and application scenarios are the first to land. **Tencent’s hybrid model supports intelligent creative creation, adapts to industry and regional characteristics, and realizes the natural integration of text, pictures, and videos. For example, when a user needs to generate a poster about a hiking route involving Asian women at a campsite, they can get an AI-generated poster by entering a prompt word.

At this year’s Baidu World Conference, Baidu launched the AI Native marketing platform “Qingge”, where Robin Li used the creative generation ability of the helm to demonstrate the rapid generation of car graphics and video ads, generate 100 advertising copies in 2 minutes, and produce 1 digital human video in 5 minutes.

Xiaohongshu has also logically launched an AI smart note-taking assistant to help merchants generate titles, texts and pictures based on products when publishing product notes, allowing merchants to realize one-click release of product notes. To this end, Xiaohongshu also announced that it will invest tens of billions of traffic support and 100 million subsidies during the Double 11 period, although it is not yet certain whether the notes generated by the official tool will receive more traffic recommendations.

**Some service providers are also coming. **In June this year, Weimob launched the intelligent assistant “WAI”, and recently launched the “Decisive Double 11” plan, which not only provides content generation capabilities for merchants, but also divides the entire Double 11 into four stages: water storage period, pre-sale period, detonation period and return period, and generates event content for different types of marketing materials required at each stage.

**The above is the most familiar to everyone, and it is also a key part of marketing - idea generation, which uses AI to generate materials used in marketing. In addition, at present, the reshaping of the marketing industry by AIGC technology is also reflected in content understanding, delivery and distribution. **

In the content understanding link, according to Zhang Zheng, an algorithm engineer at JD.com, there are hundreds of millions of products in JD.com, and the content information of each product includes the title, main image, product detail map, user reviews, etc. Based on these data, with the help of NLP, short titles or new titles can be generated through title compression, and labels of image content can be established through main color recognition or product recognition, so as to judge whether the image is of high quality enough and suitable for advertising exposure.

**Using AI’s understanding of advertised products can also tap users’ potential business interests. **

For example, Tencent’s hybrid training model accurately refines and captures the product tags and categories corresponding to the user’s language description, so that it can quickly match and recommend people and goods, and ultimately improve the conversion rate. In the user consultation process, AI can also improve the user’s experience by extracting some common sense knowledge, answering and promoting products to the user.

Compared with foreign countries, the deepest link of China’s current intelligent transformation is concentrated in content creative generation, which also has a lot to do with the manufacturer’s All in large model, and the multimodal functions of some large models such as Wensheng pictures and Wensheng videos are naturally deeply related to the advertising and marketing industry.

However, referring to Salesforce’s AI intelligent product Einstein, which covers the whole link of customer insight, marketing forecasting, personalized generation, recommendation, intelligent analysis and decision-making, etc., the domestic marketing field is still in the stage of focusing on local and technological upgrades, and needs to be further integrated with its own business.

“Decisive Battle” Smart Delivery

The competition of e-commerce is becoming more and more fierce, and the growth of public domain platforms has reached the ceiling, so a series of problems such as expensive traffic and high quotations of big anchors have followed, and marketing channels and costs are undergoing huge changes.

**And these marketing problems are concentrated in the annual Double 11. This year’s Singles’ Day, the whole process of marketing, from content understanding, idea generation to delivery and distribution, is being reshaped by AI. **

“Any brand and enterprise launch, have to find professional pitchers, they are worried about how to spend the money, traders to respond almost 24 hours a day, look at the data of each time period, the effect is good, the effect is good, the effect is not good, to reduce the investment”, a trader to the light cone intelligence preached.

In the face of multi-channel delivery, the manual operation of merchants is cumbersome, and during the delivery period, someone also needs to “keep an eye on the market” at all times, or the delivery plan may “run away”, and the management efficiency is tested.

Countless advertising engineers and product managers are racking their brains to optimize the advertising effect in different links, but they quickly reach a bottleneck with manpower against the “system”. **The essence of reshaping marketing is a more concise and efficient marketing method, and the balance between cost and efficiency is extreme. **

For example, manual delivery is often “post-investment optimization” rather than “pre-investment prediction”, which not only makes labor costs high, but also brings the problem of “cost fluctuations” in delivery, the person in charge of a brand e-commerce told Light Cone Intelligence: **The loss of enterprises in the delivery has reached 30% of the total cost. **Through intelligent delivery, the overall delivery situation can be regulated, and the plans/units of the same business type in the account can be aggregated to assess the cost centrally, so as to improve the stability of conversion costs.

More importantly, the traditional advertising engine is funnel in terms of technical structure, from advertising recall, advertising rough layout to creative optimization, and then to advertising fine arrangement and final auction, this process is a multi-level complex structure, and layer by layer of loss, the effect is decreasing layer by layer, what can be done manually is to reduce the loss as much as possible, but it can never do zero loss.

**The advent of intelligent marketing abandons the funnel structure of the past, and makes the entire advertising system more direct in an end-to-end way, and the creative and decision-making processes in the middle are handed over to the large model, reducing cost consumption in a flat way. **

Domestic e-commerce platforms and large manufacturers are actively exploring the road. Before Double 11, Baidu launched the “Qingge” full-stack intelligent marketing platform based on Wenxin Yiyan, creating a generative orientation system based on a large model, breaking the traditional logic of user interest orientation and search keyword targeting, and understanding user needs through large models and matching user needs, rather than simply and rudely “throwing stones to ask for directions”.

Backed by its own Jingzhun Tong intelligent delivery platform, JD.com has launched an intelligent launch new customer plan, new product promotion and explosive product plan. In terms of new customer planning, dynamic pricing can be adjusted based on the dynamic optimization of CTR, CVR and other indicators. In terms of new product promotion, intelligent product selection and independent product selection are combined, and the strategy is customized according to SKU granularity for precise delivery; In terms of explosive product planning, it provides a variety of control options of “expected ROI” and “expected CPC”, and intelligently matches keywords, crowds, and ideas. It can be seen with the naked eye that almost all marketing platforms are making efforts in intelligence, and this year’s Double 11 test is like the first stone thrown into the water, trying the depth. Dai Shan once gave Taobao’s intelligent reform period: “Within 5 years, Taobao will realize a comprehensive intelligent reform of merchant tools.” **And this is not only for Taobao’s time, but also for the entire e-commerce industry, pressing the intelligent countdown button. **