Can Ali use AI e-commerce to fight against Pinduoduo?
Image source@Visual China
Text | AI Business Review
"The era of AI e-commerce has just begun, and it is both an opportunity and a challenge for everyone. ”
In response to Alibaba's employees' discussions on Pinduoduo's financial reports and e-commerce, Jack Ma rarely participated in the discussion and spoke on Alibaba's intranet.
Alibaba has always been vigorous and resolute, and has fired the "first shot" of AI e-commerce.
According to the "LatePost" report, Taotian Group has just finished sorting out its AI business, closing from about 20 teams to 4, and at the same time releasing Taotian's own large-scale model product "Xingchen" internally. At present, the AI team of the International Digital Business Group has more than 100 people.
It is reported that on the day that Ma Yun spoke on the intranet, Dai Shan, CEO of Taotian Group, and others jointly led the formation of the team. Taobao will leverage the group's technology and data capabilities to upgrade all existing merchant tools and create new user products and services in the AI era.
On December 20, Alibaba announced that Wu Yongming, CEO of Alibaba Group and chairman of Taotian Group, is also the CEO of Taotian Group. At the same time, Dai Shan, the former CEO of Taotian Group, will assist in the establishment of Alibaba Group's asset management company, which is a new business function after Alibaba's transformation.
From then on, Wu Yongming will serve as the CEO of Alibaba Group, Taotian Group and Alibaba Cloud Intelligence Group.
In September this year, Alibaba's new CEO Wu Yongming established "AI-driven" as one of the group's two strategic priorities in an internal all-staff letter, and in November, in his first earnings call after taking office, "AI" became the most important keyword for him to describe Alibaba's strategy and various businesses this quarter.
So, can AI e-commerce pull back a game for Ali?
Ali grasps the "life-saving grass" of AI e-commerce and urgently improves "human efficiency"
Pinduoduo's market value exceeds Alibaba's, which makes Ma Yun unable to hold back, but what worries him more is Pinduoduo's "strange efficiency".
As of the end of June this year, Pinduoduo had 13,000 employees, while Alibaba had 229,000 employees, and the per capita market value of Pinduoduo was about 17.6 times that of Alibaba.
Even against the entire Internet circle, including Tencent, ByteDance and other companies, Pinduoduo still has the highest human efficiency, which has also caused Pinduoduo's stock price to rise by 69% year-to-date.
Pinduoduo's ultra-high human efficiency is due to its unique management model and business model, such as supply chain management, self-built logistics system, dynamic pricing, social media precision marketing, intelligence, etc., and of course, it is also inseparable from the "996" overtime of employees' ultra-high working hours.
This model has formed a scale and flywheel effect, and it may be too late for Ali to catch up, so he chooses to use AI technology to reduce costs and increase efficiency.
Tsai Chongxin, chairman of the board of directors of Alibaba, said in a letter to all employees that Wu Yongming will serve as the number one position of Alibaba Cloud and Taotian, which will help lead Taotian's transformation with technological innovation, and help ensure the group's unified command and high-intensity continuous investment in the two strategic centers of e-commerce and cloud.
“All in AI”，阿里不止求“快”，但求更快。
An insider revealed that Taotian has discussed the changes of AI to the business many times internally, and has also launched some AI products around Double 11, but with Wu Ma at the helm, this process will be accelerated. "What used to be a multiple-choice question is now mandatory, and lines of business are being asked to rethink their business around key futures such as AI. ”
According to LatePost, in April this year, Zhang Kaifu, head of the cross-border business industry of Alibaba International Business Group, led a team to set up an AI Business team, of which about 1/3 of the model trainers are responsible for the development and operation of model inference applications and infrastructure, as well as specific products.
The algorithm team focuses on three directions: multilingual e-commerce large model training, dialogue models and downstream tasks, and image generation capabilities.
AI Business's AI services have been applied to AliExpress, Trendyol, Daraz and other businesses within Alibaba International, mainly including product information localization, image design, customer service robots, customer service translation and other scenarios.
The three major elements of e-commerce are "people, goods, and places", so how does Alibaba integrate AI into the current e-commerce model?
First of all, in terms of "people", Alibaba analyzes users' shopping history, search records and other data through AI technology, which can accurately grasp the interests and needs of each user, so as to provide them with personalized advertising and promotional activities and improve conversion rates. At the same time, AI e-commerce can provide the most suitable price of goods to users, without having to compare goods, and truly find people for goods. This will not only attract more users, but also increase the user's purchase rate and loyalty.
At the same time, the customer service bot AI bot integrates the capabilities of Tongyi Qianwen large language model, which can interact with users through multiple channels, including text chat, voice call, etc. It greatly reduces the workload and work pressure of manual customer service.
On the "goods" side, by analyzing historical sales data and market trends, the AI system can predict future sales, so as to make inventory management and logistics arrangements in advance. This not only reduces inventory overstock and waste, but also improves the timely arrival rate of goods and improves the user experience.
On the "field" side, it is also Ali's weakest link - Pinduoduo's community e-commerce is in full swing, byte e-commerce has content platforms such as Douyin and TikTok, JD.com is strategically bound to Tencent, Baidu AI e-commerce has a huge traffic pool of search engines, and Ali lacks the most traffic moat.
The AI model has a more humanized experience than previous technologies, whether it is a multi-round dialogue or a Wensheng picture/video, which may help Alibaba make up for diversified and humanized scene content, create a more user-sticky "field", and at the same time greatly reduce human investment.
Not only AI Busines, but also the technical teams of other business units of Alibaba have independently tried and explored some AI large model applications.
On December 11, Ali Da Mo Academy released two large artificial intelligence models, SeaLLM and its conversational fine-tuned version SeaLLM-chat, with 13 billion and 7 billion parameters, respectively. They are capable of handling Southeast Asian languages, including Vietnamese, Indonesian, Thai, Malay, Lao, and Burmese, among others.
DAMO Academy says these AI models are better able to perform tasks that conform to local Xi, style and legal requirements than other large language models trained on English and Latin datasets. According to the Academy, these models could also help the company build chatbot assistants for businesses in the Southeast Asian market.
With the rise of live streaming, cross-border e-commerce companies are facing challenges in recruiting and managing foreign live streamers, and they want to use AI tools to build local services quickly and cost-effectively, according to an analyst.
For example, Lazada, Southeast Asia's largest e-commerce platform owned by Alibaba, aims to reach $100 billion in revenue by 2030 and has 300 million users in the region. In May, Lazada launched LazzieChat, an AI chatbot for consumers in Southeast Asia, to answer consumers' shopping queries and recommend products to them. In July, Lazada launched an internal product for AI marketing assistants and advertising ROI forecasting for operations staff.
The first obstacle for e-commerce to go overseas is the language barrier, including Vietnamese, Indonesian, Thai, Malay, Lao and other small languages with high demand and high cost. Conversational AI bots can be trained through large AI models, which can help solve most customer service problems, and can be continuously optimized and improved manually, which can significantly reduce costs and increase efficiency.
AI e-commerce not only helps Ali's e-commerce platform improve efficiency, but also improves the efficiency of merchants on the supply side.
In early November, Alibaba International Digital Commerce Group said it was piloting its first AI-powered app interface suite, Aidge, based on Alibaba International's proprietary large language models and computer vision models trained on its e-commerce expertise.
Internal tests on Alibaba.com have shown that Aidge can significantly improve order conversion rates, consumer satisfaction, and efficiency for small and medium-sized businesses. Kaifu Zhang, vice president of Alibaba's international website, said that Aidge is committed to simplifying the global operations of small and medium-sized enterprises.
One of the platforms participating in the pilot is Daraz, a South Asian e-commerce platform. Tatiana Palhares, Daraz's chief product officer, said in a statement, "Adopting Alibaba's Aidge AI suite is a seamless process that has the potential to transform our business. ”
During the pilot, Alibaba will test more than 15 APIs with SaaS partners and independent service providers.
Areas covered by the API include:
Consumer Insights: Customer Feedback Analysis and Selling Point ExtractionContent Localization: Product Content Localization Covering Text and ImagesCustomer Service: Multilingual Live Chat, Shopping Assistants, Chatbots, and Customer Service ChatbotsFeaturesDesign: Various design templates that can be used to create marketing ideas or product display exhibitsMarketing: Automatic Search Engine Optimization and Marketing Information Generation
Aidge has also launched several AI-powered agencies that support businesses in sales conversion, customer consultation and request management, and search engine optimization. These proxies have been deployed on e-commerce platforms operated by Alibaba International.
"Today's trial run of Aidge is just the beginning. The novelty of AI technology requires us to work with SaaS partners around the world to not only test, but also co-define their use cases," Daniel Zhang previously added in a statement.
According to a report by investment bank Goldman Sachs, generative AI could drive global GDP growth by 7% within a decade.
With the rise of large-scale model technology, new productivity tools have triggered a new round of productivity transformation, and e-commerce is also being profoundly transformed by AI. Ali can't afford to lose the next e-commerce cycle and faces severe challenges from a number of competitors such as Pinduoduo.
"Social + e-commerce + AI" new business model,
Pinduoduo is one step ahead
Pinduoduo's market value beats Ali, SHEIN has a market value of more than 600 US dollars to go to IPO, and today's e-commerce has evolved a new way to play, and Ali is indeed lagging behind in this regard.
Pinduoduo's market value surpasses that of Alibaba, the core of which is due to the new business model of "social + e-commerce + AI".
Temu, which was founded in September last year, has now developed into a "locomotive" business of Pinduoduo. TEMU has entered North America, Australia, Europe, Asia and other markets, and has been launched in 47 countries in one year. CICC estimates that Temu's revenue in the third quarter of 2023 will be about 16.1 billion yuan, accounting for 23% of Pinduoduo Group's revenue.
Let's take a look at how Temu achieves savage generation.
Temu attracts a large number of users through vigorous advertising and marketing, and then trains algorithms based on these user data to make targeted recommendations.
On the Temu platform, the algorithm "customizes" the endless items you want to buy for you, just like walking into a digital department store and only wanting to buy one item, but walking out with a dozen items.
Algorithms are also able to make recommendations based on your browsing or past purchases, knowing which products are related to each other and what kind of people are likely to buy those combinations of products – only if you have enough new purchase data, enough customers, and enough new products for a buying algorithm to work.
If a store only has 1,000 or even 10,000 items, it's not enough to provide a personalized shopping experience for the end user.
Unlike traditional "search-style" shopping, this "discovery" shopping provides a more solid foundation for social online shopping. Social add-ons allow you to ask your friends for their opinions on an item or rank an item you're considering buying.
It is worth mentioning that Temu uses AI technology not only for social e-commerce, but also for dynamic pricing algorithms to control the cost of goods.
Temu is able to study the supply and demand of a particular product, market prices, and real-time demand to ensure competitive pricing, especially for a specific period. If you're an avid shopper, Temu will be a lot cheaper for you if you can wait a little longer.
Huang Zheng, the founder of Pinduoduo, can be said to be born in the era of AI e-commerce, having previously experienced games and e-commerce entrepreneurship, and is well versed in social networking and e-commerce, and is also an AI elder.
At the 2020 Titanium Media T-EDGE Science and Technology Life Festival, Huang Zheng shared in his speech that "social + e-commerce + AI" was quite profound and forward-looking:
"The current distributed AI structure is very different from the previous centralized AI, the centralized AI in the past is trying to replace the brain work with AI, people only need to ask the machine, it is faster than you, and it feels useless; ”
AI e-commerce is not only using technology to improve efficiency, but also using technology to improve everyone's personality, which can make people connect more deeply, have a better way of expression, and can also influence more things through their own personality.
In essence, Pinduoduo is to integrate technology and life, integrate technology with people's emotions, and try to use technology to make people happier, rather than simply making people more efficient.
It is reported that Pinduoduo has set up a large model team of dozens of people, with an annual salary of nearly 1.3 million yuan for the largest model positions.
epilogue
In the past 20 years, the e-commerce industry has been evolving, from traditional e-commerce, to mobile e-commerce, to live e-commerce, and next, "AI e-commerce".
Behind the changes, the e-commerce platform is becoming more and more efficient and humanized.
2023 is the end of the traditional e-commerce cycle and the beginning of the AI e-commerce cycle.
Goldman Sachs predicts that AI will save the retail industry $54 billion a year in costs by 2025, while generating $41 billion in new revenue.
At present, Alibaba, Baidu, Pinduoduo, SHEIN and other large manufacturers have opened the door to the era of AI e-commerce.
A new e-commerce war is about to begin, and this round of cycle will be a comprehensive contest of "technology + management efficiency + business model". In the future, whether it is Pinduoduo, Alibaba, SHEIN, or other e-commerce platforms, the research and development and application of AI large models will become more radical.
This AI e-commerce war will last for years or even decades, and it will be interesting to see who will win and who will be out first.
Reference Links:
https://mp.weixin.qq.com/s/L_cHkF7ndHtEesVpMa6R-ghttps://mp.weixin.qq.com/s/L_cHkF7ndHtEesVpMa6R-ghttps://finance.sina.com.cn/stock/estate/integration/2023-12-13/doc-imzxwhkx2277573.shtmlhttps://xueqiu.com/3332132759/148481527https://zhuanlan.zhihu.com/p/669332248https://zhuanlan.zhihu.com/p/670548652