How many AI bubbles are there in smart customer service?

Pictures from the Internet

Text 丨 Intelligence Relators

Author 丨 Shen Lang

The capital tide, the bubble of the AI ​​industry is broken. At this time, the market is more concerned about the application of technology than before. Even applications such as smart customer service are already very popular and relatively mature, and they are repeatedly questioned, verified and iterated, and trapped.

To this day, the smart customer service track is still lively, full of various players, and the increasingly fierce competitive situation is to promote the advancement of technical application, but also promotes the forefront of this field. Foam.

Smart customer service, there are bubbles?

In the current landscape of domestic enterprises, customer service is a very important business link, and smart customer service is also generally becoming the primary choice for corporate business transformation.

According to the "Intelligent customer service foreseeing future & smart customer service trend to develop white paper" based on T research and Ronglian Qimo. In the domestic market, companies' demands for customer service are generally high. 88.6%of companies have their own customer service business, of which 38.4 is 38.4 %Of companies have brought customer service to the cloud, and 22.1%of companies have further constructed AI smart customer service based on the cloud.

The widespread deployment of intelligent customer service has enabled users to contact all kinds of smart customer service in their daily consumption process, such as voice, text, robots, etc., which has become a normal state of consumption life.

However, this normal does not bring consistent praise. This year, the China Youth Daily Social Survey Center jointly conducted a smart customer service satisfaction survey. The results showed that 95.7%of the respondents had used smart customer service. In the case of so wide audience, only 41.3% The interviewees feel that smart customer service is easy to use, and the satisfaction is only 40 %. More respondents still favor the original artificial customer service.

Where is the question? According to the survey feedback, in the process of using smart customer service, the respondents were dissatisfied or encountered more questions. They were mainly answered hard machinery (62.1%) and cannot accurately understand questions (52.2%). To put it simply, the lack of artificial customer service for intelligent customer service, and cannot accurately understand the questions of users, let alone solve the problem personalized, so that the efficiency of solving the problem is too low.

This is realistic, but it seems that the data given by the market is not consistent. Regardless of whether it is an intelligent customer service solution provider or the company that has deployed smart customer service, the market data given is quite high. For example, the accuracy and success rate of customer consultation and the success rate have basically reached more than 90%, and some can even soar to 95% It is also high.

It is not known whether the contradiction between this data and the data is created by the difference in statistical dimensions. But it is clear that between this contradiction, it is easy to produce the foam of the entire smart customer service track. On the one hand, the company is busy deploying the smart customer service system. The problem of dislocation of supply and demand.

Can smart customer service completely replace artificial customer service?

For the technical application trend of artificial intelligence, clouds are described by "three waves" from science and technology. Among them, the first wave is mainly driven by a single point technology explosion. Most of the experiences brought by linearity such as face recognition and voice interaction. The second wave is the industrialization and scenario experience of multi -point technology integration driver, and has more active perception ability, clearer cognitive ability, and so on.

At present, the entire AI industry is in the state of the first wave and the second wave superimposed, as is smart customer service. Therefore, at the moment, smart customer service gives people a very linear feedback, but as the entire multi -point technology integrates closed -loop construction, smart customer service is slowly "spiritual", more temperature, more temperature, and more more intelligent.

Judging from the solutions provided by Netease Qiyu, JD.com's smart customer service, and Tencent Enterprise Points, this path is relatively clear.

First, "intelligence" is not the air tower, and must have rich corpus accumulation as the basis.

In the process of communication with "intelligent relativity", NetEase Qiyu mentioned that in the field of intelligent customer service, the accumulation of data and answering effects is a enhanced reinforcement circuit for mutual effects: more dialogue data input -more Rich and diversified corpus data accumulation -the model is more complete, and the customer service robot is trained to be smarter -covering more users' questioning and questions -more accurate answers to customer service robot output -users get a better experience- —The user higher frequency use -complete more data accumulation.

In this regard, Netease Qiyu also believes that it is not difficult to build a customer service system itself. The difficulty is "intelligence". The matter of intelligence has a lot to do with the input, because the business logic of all walks of life is different. Users ask users to ask. What kind of question, what kind of questioning method is used, it is even more different.

To some extent, in the early days, smart customer service did not exist in the optimal solution that fully adapt to various industries. It must accumulate the statistics according to the needs of all walks of life, and continuously feed massive data in order to achieve "technical logic" technical logic " intelligent". Through the richness of corpus accumulation, the intelligent performance of intelligent customer service solutions for major manufacturers is opening the gap.

Netease Qiyu knows this logic deeply. In the past 6 years, it has served 400,000+home companies, covering head customers from all walks of life, such as Wal -Mart, Huaxi, ideal car, cotton era, Geely, etc. , Efficiently connected 600 million users, with a total of 7.6 billion rounds of sessions, and the news was 61.5 billion. This is very important. It is based on a large amount of input. Netease Qiyu has accumulated 100 million corpus data and 20+ industry knowledge bases. Combined with the self -developed NLP core algorithm and knowledge base, its customer service robot has achieved the average problem matching The rate of 95%, the problem solution of more than 90%.

Second, the customer service system on the cloud can only be optimized.

Massive data, 100 million levels of corpus accumulation ... such a huge -scale data processing will be a normal state for smart customer service. Then, in the process of iteration in the future, Shangyun is a necessary support for the transition of the customer service system to the intelligent upgrade. Localized storage is difficult to meet the continuous stream of data accumulation.

According to the "2021 China Intelligent Customer Service Market Report" published by the Toubao Research Institute, it is currently in an advantageous competitive position on the smart customer service track. Traditional customer service manufacturers, which are mainly localized deployments, are also transformed in cloud customer service solutions.

Cloud service has become an important consideration indicator in the field of intelligent customer service. In the "Artificial Intelligence Series Report: 2021 China Dialogue AI Market Report in 2021" issued by the Shatin United Touna Research Institute, the main thing to enter the leaders' quadrant is the domestic head cloud manufacturer, such as Tencent Cloud, Jingdong Cloud, Baidu Smart Cloud and other.

Like Jingdong Smart Customer Service, Rhino has two ways: privatized deployment and SaaS deployment. Based on JD Cloud, the dynamic language model update and the automatic dialogue of the dialogue path are used. , Support the stability of the system and the processing of complex tasks. At present, 580 million JD users and 174,000 merchants, such as the recent epidemic prevention and control and investigation solutions that have landed in Beijing. It can be seen that during the iteration of smart customer service, the technical advantages provided by cloud services are significant.

Third, the support of AI capabilities is still the key driver of the advanced intelligent customer service.

The value provided by AI capabilities is advanced, that is, how to screen, recognize, and transform the massive data of the cloud into the application value of smart customer service. This prompted the application level of intelligent customer service to the technical system in the field of artificial intelligence.

Taking the core logic of intelligent service robots as an example, its core algorithm is mainly to identify the intention of the user. Simply put, no matter what kind of services to provide users, the first priority must be perceived and identified by the user's intention. In this regard, Netease Qiyu adopted the current mainstream Transformer structure in the regard to identify algorithms. At the same time, the use of distillation+multi -model integration algorithms not only improved the model reasoning speed, but also improved the accuracy of the intent.

From a technical perspective, the technologies used by intelligent customer service include voice signal processing, natural language processing, dialogue management, and voice synthesis. Basically, the entire technical system is closely around the field of artificial intelligence. This also makes AI capabilities also become a core competitiveness of smart customer service solutions.

At present, smart customer service has completely belonged to a application system with multi -point technology integration. Single -point technology is not enough to support the customer service business needs of an enterprise. More and more complicated needs will continue to force the entire intelligent customer service to upgrade before Have the opportunity to gradually replace the work of artificial customer service in the next time.

Who will define smart customer service?

Judging from the two reports of the Trigeration and Ronglian Qimo and the China Youth Daily Social Survey Center, although the current smart customer service is only 22.1%of the company's deployment, most of them are head enterprises that are closely related to the field of people's livelihood consumption, such as communication Operators, Haidilao, Wal -Mart, Watsons, etc., the consumption public's contact with smart customer service is still common.

Such a general state has completely entered a new stage of the development of smart customer service. In the past, for enterprises, the deployment of smart customer service may be more or a process of rushing trends and showing muscles. Just like advertising, it is necessary to let your users see that their favors have the strength to advertise, so as to shape high -end value to create high -end value Brand cognition.

Today, once the smart customer service does not do well and bring a very poor experience to users, it will easily cause the brand's favorability to decline. With the continuous popularization of smart customer service, the impact will become more and more obvious.

As the People's Daily published a comment, "Smart customer service is not intelligent, and AI has become an obstacle." This is also the same for the shape of the brand and the maintenance of customer relationship -smart customer service is entering a stage of user value -oriented, and what it requires is the value feedback of more accurate and efficient solutions.

Then, if the smart customer service wants to skim off the foam, you must pay more attention to the user's experience effect. Based on the needs of users, the value of smart customer service is defined, rather than a "closed door car".

This is also applicable to today's artificial intelligence industry.

- END -