Shulou(Shulou.com)06/03 Report--

One thing that is very different from previous years in this year's Singles' Day is that there are more elements of "human" participation. The e-commerce platform, which used to connect people and goods directly, now has a live broadcast of this purchasing agent. Traffic stars have a trend of carrying goods, and the introduction of external traffic also needs to rely on fan economy. As for logistics, customer service and other links, there is frequent news of staff working overtime all night. This seems to be quite different from the e-commerce economy that reduces complex human processes as we first understood.

Of course, when the great promotion comes, e-commerce practitioners should naturally devote more time and energy to serving users. But can we once again use technology to revolutionize the efficiency of shopping, just as we have changed retail with the Internet?

In this era of new technologies, the answer is, of course, yes. In this Singles Day holiday, in addition to the annual increase in the transaction quota, the growth of technology is also worthy of attention. Among them, JD.com AI is a typical sample of the times.

AI output Logic under High flow pressure

We know that in customer service, logistics warehousing and other links, JD.com has been using AI to improve efficiency. However, since the emergence of the NeuHub JD.com artificial intelligence open platform, the pace of JD.com 's AI output capacity has been greatly accelerated, and its own technical ability has accumulated like a snowball, which has been rapidly improved with its wide application.

This kind of promotion is concentrated under the high flow pressure of Singles' Day. With the influx of a large number of consumer traffic, there will be a huge demand for customer service consultation before, during and after sale, which are assigned to multiple entrances such as category C merchants, JD.com self-management, logistics, after-sales complaints and so on. including voice, text, pictures and other rich forms. Especially at the critical point of great promotion, consumers' mood and patience may be different from those they usually do. Businesses not only want to properly deal with user problems, but also want to provide more high-quality and efficient services to help complete the transformation of traffic.

All these test the ability of AI, including emotion analysis, speech recognition, semantic understanding, knowledge graph and so on. JD.com AI packaged these capabilities into two levels of service.

Among them, the most basic is intelligent customer service, which relieves the pressure of artificial customer service through machine response. During the double 11 promotion period, the 24-hour round-the-clock service provided by JD.com Mall's online intelligent customer service accumulated more than 34.16 million queries during the period from November 11 to November 11, and the intelligent customer service independently solved more than 6.59 million consultations on November 11. JD.com Logistics online intelligent customer service fights Singles Day for the first time this year, receiving a total of 450000 consultations, independently solving 78% of the problems in covered traffic, which is equal to about 3200 manual customer service, ensuring that the manual online connection rate is more than 95%.

We are already relatively familiar with this level of service. generally speaking, intelligent customer service is an integrated output of NLP technology, which allows as many users as possible to get answers in time when traffic is concentrated, and reduces the pressure on manual customer service.

Continue to move forward, we can also see JD.com AI provides the second layer of service, JD.com customer service brain. Different from the independent service strategy of intelligent emotional customer service, JD.com 's customer service brain is more focused on helping manual customer service to improve work efficiency. Compared with the manual customer service that answers almost every consultation without difference in the past, the customer service brain can filter out the invalid consultation from the source by identifying the characteristics of the conversation, and automatically search for the needs of the other party in the course of the conversation. provide information automatic query response program real-time recommendation service to speed up customer service efficiency and ensure service quality. During the Singles Day holiday, the customer service brain accumulated 360000 manual conversations, 1.96 million auxiliary customer service reception conversations, and reduced customer service hours by 45115 hours.

We can see the ability output logic of JD.com AI: not to do everything, but to create a multi-layered solution to facilitate users with different needs to choose applications. To be able to give users the right to choose shows that JD.com AI has sufficient strength, which comes from long-term data and R & D accumulation, as well as sufficient AI application experience.

Not a triumph of empiricism

But in this year's Singles' Day, what we see is by no means a victory of AI's "empiricism", but also a brand-new technological milestone.

As we mentioned earlier, an important feature of this year's Singles Day holiday is the increasing introduction of external traffic to retail platforms. This situation may be seen as a phased reflection of the new retail development, when the retail channel is global, the input of traffic should also be global. One of the technical milestones of JD.com AI this year is the introduction of traffic through technology.

This year, JD.com AI focuses on updating the "photo shopping" technology, which is based on CV to search pictures, combined with voice interaction technology. Users can take photos or take screenshots when they see products they are interested in on the Internet or in real life, and search for the same item in JD.com by taking pictures and buying them. JD.com AI fine particle image recognition can accurately distinguish the details between goods, as far as possible accurate positioning of the same style rather than just similar models. After updating, photo shopping can automatically locate multiple complete commodity subjects of the image based on object detection technology, and can better help users to select the goods they are interested in among many products in live broadcast, video and other scenes.

It is reported that there will also be a function called "matching shopping" that will appear with photo shopping in the future. JD.com AI will combine product pictures, text descriptions, categories and other information, and integrate image classification, image understanding, image synthesis, fashion analysis, natural language processing, personalized recommendation and other technologies to build a deep learning network to help businesses and users build combinations of fashion products with different styles, preferences and application scenarios. Not only provide users with dress advice to pry purchasing power, but also give merchants choice and product design.

Photo shopping uses API output mode, allowing price comparison e-commerce, smartphones, short video applications and other third-party access to obtain visual identification and search capabilities of goods, as well as diversion commission. At present, the average daily consumption of API for photo purchase on NeuHub platform is more than ten million times. In this year's Singles Day holiday, the photo purchase is a guide for the mall, and the daily GMV is close to 100 million yuan. This mode of obtaining traffic through technical capabilities can be said to be a technical milestone set by JD.com AI on Singles Day this year.

At the same time, Alpha has also had an excellent performance during the Singles Day holiday this year. Alpha is a pre-sales service robot launched by JD.com AI, which can conduct in-depth voice reasoning and knowledge graph Q & A for user inquiries, as well as large-scale cross-category general in-depth semantic understanding. This means that Alpha will be able to consult and answer questions and undertake shopping guides more openly like human beings. During the promotion period, the application of pre-sales service robot can help merchants leave intentional traffic through interaction, otherwise users are likely to turn to other stores immediately when they do not receive a response to the consultation. During the Singles' Day period, Alpha served a total of 12 million people, which is 370% of that of the same period last year. The new intelligent assistant model not only ensures the user experience without wasting every traffic, but also increases 15% orders for merchants and saves 50% customer service costs.

There are many similar milestones. For example, JD.com Logistics launched artificial intelligence phone booking service this year, using conversation robots with voice recognition and voice generation capabilities instead of manually booking the distribution of large goods before delivery. in order to provide customer service pressure and arrange logistics work reasonably in the logistics demand outbreak period. There is also the intelligent writing function that is about to be put online. JD.com AI helps merchants automatically write more brilliant merchandise promotion skills through the study of massive commodity copywriting. In the future, in the promotion nodes such as Singles' Day and 618, merchants can reduce the copywriting pressure in the massive commodity promotion work through intelligent writing.

Behind the miraculous digital growth of this year's Singles' Day, apart from chopping hands, there is no lack of the contribution made by these newly landed AI technologies. With the help of AI, retail is looking for people from people to find people for real goods, and the key to this process is to use technology to open the connection between the two.

The sense of trust from every key node

In the end, we would like to discuss what it means for the whole technology world to apply AI on such a critical node as Singles' Day.

We must admit that although AI has appeared for decades and experienced many ups and downs, it is still an unfamiliar tool for human beings. At this critical moment of the Singles Day holiday, it is not easy to apply such an unfamiliar tool.

In our daily work, if we catch up with critical moments such as work reports and bidding, we will certainly prefer to be accustomed to using software or hardware rather than risk affecting important work. to apply software and hardware that may be more efficient.

Whether it is an intelligent customer service system that continuously outputs technical capabilities, or a photo purchase that sets a new technological milestone. JD.com AI's achievement during the Singles Day holiday is not so much to set up an application sample as to build up a sense of trust. With the API transfer rate of tens of billions of NeuHubs, it tells the world that AI can cope with the big boost, even better than manpower, and bring a better experience to consumers. Not only Singles' Day, but many of the above technologies have been put into use during the 618 period this year. After repeated training, as users gradually build trust in AI, technical services are also slowly becoming unaffected. People no longer notice whether the person answering the consultation call is the human or the AI, only that the problem is solved faster and faster.

In fact, the same is true of trustworthiness that JD.com AI has always emphasized. People's trust in AI comes not only from the boundaries of scientists' expanding technology, but also from users constantly opening up territory and quietly planting AI around everyone.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.