"Hello, how can I help you?"
"Transfer to manual service."
"The manual agent is busy, please call again later..."
Recently, many consumers have reported to CCTV that when answering user questions, they often encounter AI customer service sending a bunch of fixed words that do not answer the question and cannot solve the specific problem. When they want to contact manual customer service, they will encounter "numerous obstacles". Some platforms even directly cancel the manual customer service option, causing a lot of inconvenience to users.
In this regard, the reporter called the customer service numbers of more than 10 companies, including ticket booking platforms, express companies, and communication service providers, and found that after the call was connected, intelligent voice announcements were made, and it took an average of one minute to transfer the call to manual customer service. In addition, the reporter randomly selected 20 merchants from different e-commerce platforms and consulted customer service on product issues. Among them, 12 stores were answered directly by AI customer service, and 8 stores were answered by manual customer service.
Relevant experts said that nowadays, more and more companies are choosing to use AI customer service. This is to control labor costs and also out of consideration for users’ round-the-clock consultation needs and the large volume of order consultations. However, AI customer service cannot be an excuse for some companies to ignore consumer consultation needs and avoid after-sales issues. Regardless of manual customer service or AI customer service, companies should take solving the actual needs of users as the starting point, respect consumers' right to free choice and improve the service experience for consumers.
Consumer: Why is it difficult to get through to manual customer service?
On New Year's Day in 2024, Wang Rou's friend purchased a round-trip ticket from Beijing to Haikou on a booking platform. After the payment was successful, the page kept showing "ticket pending".
Wang Rou said that they were afraid of affecting their travel plans, so they called the platform customer service to confirm whether the ticket purchase was successful and when the tickets would be issued. "The AI customer service kept replying mechanically, and it was still difficult to find a human customer service. For this simple issue, we communicated from about 11 a.m. to about 4 p.m."
According to the chat screenshots provided by Wang Rou, she repeated the questions she needed to consult more than a dozen times, but received invalid responses each time.
Dialogue between consumers and AI customer service (photo provided by interviewees and published by CCTV)
Wang Rou said that the AI customer service was very enthusiastic but could not solve the problem. After she sent "transfer to manual" four times in a row, the page showed that the transfer to manual customer service was successful, but the manual customer service also had fixed words. "I don't even know whether the customer service is a real person or an AI. As a consumer, why is it so difficult to contact human customer service?"
"Some companies use AI customer service to pretend to be manual customer service, saying that it is a human service, but in fact the robot is answering questions." An operator of an Internet technology company in Beijing said that some Internet companies have a large number of users, and they launch an activity or project, and the project operation The personnel will preset some questions that users may ask in advance, and then prepare some copywriting templates for the dedicated customer service team. Customer service personnel often copy and send them to users. Therefore, whether it is manual customer service or AI customer service, if there are no questions raised by users in the backend database, users often cannot receive effective feedback.
"What makes people angry is that the manual customer service also set a communication time limit." Wang Rou said that after the manual customer service was successfully transferred, she had to describe the problem again, but the customer service staff never responded. A few minutes later, the system suddenly prompted: If no more messages are sent or received, the manual service will end in 60 seconds.
It wasn't just Wang Rou who encountered a similar situation. Recently, Li Meng from Shandong said that because of the issue of changing the express delivery address, she needed to consult customer service, and she was always contacted by AI customer service. Because the AI customer service failed to solve the problem, she kept asking to be transferred to manual service. "After the call was transferred, I was still talking to the robot, and the human customer service couldn't be contacted at all."
Chen Ning once encountered a situation where he could not find human customer service. She told reporters that one time when she was playing a mobile game, she could not hear the sound in the team, so she wanted to ask a question, but the AI customer service on the platform automatically responded, and there was no contact information for manual customer service, and the AI would handle all the situations Listing them does not solve the problem.
Why do consumers and intelligent customer service have poor communication and incorrect answers to questions? Internet technology practitioner Zhang Xiaorong analyzed that at present, the maturity of AI customer service technology is not high, and it is still insufficient for some complex problems, especially in understanding the more complex language and emotional expressions of humans. In addition, if the training data of AI customer service is not rich or accurate enough, it may not be able to accurately understand the user's problem.
Customer Service Company: Helping companies achieve the goal of “reducing costs and increasing efficiency”
Chen Jing, an e-commerce practitioner, told reporters that for merchants, customer service staff have long working hours and low pay, so they have greater liquidity. For business managers, manual customer service requires a large investment in labor costs. In addition, for stores with high traffic, it is not realistic to rely on manual customer service to handle everything.
Chen Jing said that nowadays, e-commerce platforms have certain assessment mechanisms for merchants. If the merchant does not reply to the customer within the specified time, points will be deducted from the merchant. At present, many stores choose to use intelligent customer service to respond first. When AI cannot solve the problem, manual customer service is only a last resort option.
Data from the "2024 China Intelligent Customer Service Market Research Report" show that China's overall market size will be 3.94 billion yuan in 2023, and the compound growth rate is expected to reach 22.6% from 2022 to 2027.
Zhang Xiaorong said that the algorithm model of AI customer service is usually based on natural language processing (NLP) technology, including language recognition, semantic understanding, dialogue management and other modules. Through machine learning algorithms, AI customer service can continuously learn users’ questioning patterns and preferences to provide more accurate answers. When an enterprise uses AI customer service, the initial investment includes the development and deployment costs of the AI customer service system. In the long run, AI customer service is generally more cost-effective than human customer service due to reduced labor costs.
Wang Zhen, a staff member of a technology company in Beijing that provides online customer service, told reporters that during the initial launch of AI customer service, merchants need to think through issues related to pre-sales, after-sales, and product use as comprehensively as possible, and technical staff will organize the issues into different scenarios. Fill it into the knowledge base for reply interception. After the interception is completed, there will be fewer answers from manual customer service.
"Consumers may have different ways of asking a question. We will do some training on similar questions and decipher 50 similar ways of asking, so that the accuracy of the robot's reply is relatively high." Wang Zhen said that when consumers If the questions are beyond the scope of the robot's answer, there will be a cluster of unrecognized questions. These unanswered questions will be pulled out separately for training, and then the answers will be given to achieve more accurate and personalized services.
Wang Zhen revealed to reporters that they provide AI customer service to enterprises at a cost of 30,000 to 50,000 yuan a year, including basic software fees and training fees for scene construction. In addition, if the company needs to switch to manual agents, it will also charge an agent fee of 2,000 yuan per account per year.
He said that the price gap between different versions of AI customer service lies in the cost of chatbots handling a large number of user interactions. For example, AI customer service needs to complete tasks such as order identification and user identity verification, and call the backend knowledge base to provide answers based on user questions. If the internal system needs to connect and identify needs multiple times, this will increase the workload and cost, so the price of robot customer service will be higher.
Wang Zhen said that from a merchant's perspective, replacing manual customer service with AI customer service can significantly reduce the company's labor costs, reduce the amount of customer service workload that requires manual processing, and achieve the goal of "cost reduction and efficiency improvement." Compared with the salary of manual customer service, the cost of AI customer service is much lower. In past cases, some companies were able to reduce customer service personnel by 50%.
Expert: Enterprises must put people first when using AI customer service
The "Consumption Survey Report on Customer Service Convenience under the Digital Background" released by the Jiangsu Provincial Consumer Rights Protection Committee shows that 71.2% of consumers said that intelligent robots "answer questions incorrectly" and "answer questions not intelligently"; 23.6% of consumers said they were unable to Find artificial customer service, or the artificial customer service has "kicking the ball" and other phenomena.
Zhao Jingwu, associate professor at Beihang University School of Law, said that more and more companies are now choosing to use AI customer service. This is to control labor costs and also to consider the 24-hour consultation needs of users and the large volume of order consultations. For example, the use of AI customer service can efficiently respond to repetitive and common questions, and ensure the consistency and accuracy of service quality and service content. However, AI customer service cannot be an excuse for some companies to deal with consumers and avoid after-sales issues.
Hu Gang, deputy secretary-general of the Legal Affairs Committee of the Internet Society of China and a member of the China Consumers Association's legal team, believes that it is understandable for companies to introduce AI customer service to save labor costs. If the company deliberately hides the manual customer service settings or prevents consumers from contacting manual customer service , making consumers unable to solve the problem, which infringes on consumer rights. If consumers are unable to communicate in a timely manner and their specific rights and interests are harmed, they have the right to complain to the consumer association or relevant regulatory authorities and obtain compensation in accordance with the law.
In Zhao Jingwu’s view, AI customer service often causes trouble to consumers. This is essentially because businesses ignore the specific actual needs of users. Enterprises should respect consumers’ free choice and improve their service experience.
"Technology-enabled customer service should be people-oriented." Zhao Jingwu said that judging from the problems reported by consumers, some companies need to simplify the process of switching to manual customer service. For example, options such as "one-click transfer to manual services" can be set up to prevent some special groups such as the elderly and disabled from being able to transfer to manual services because they do not understand the specific operating steps. In addition, enterprises must reasonably adjust the working hours of manual customer service, clarify manual customer service options, and ensure that manual customer service can provide services to users when they need it.
In addition, Zhao Jingwu said that when AI customer service communicates with some consumers, it is difficult to understand consumers' intentions due to factors such as dialect, accent, speaking speed and volume, and it is difficult to carry out efficient conversations. Therefore, companies should continue to increase training on intelligent systems to optimize consumer service experience and service quality.
He believes that in the future, AI customer service will be able to learn and optimize itself. It can continuously adjust and optimize output results by collecting massive user interaction data to ensure the accuracy of responses and provide personalized services. It can also provide users with a certain degree of " Emotional value" to prevent consumers from talking to cold machines.
Hu Gang said that companies cannot only consider cost and efficiency when using artificial intelligence technology. From a long-term development perspective, more attention should be paid to the actual needs and feelings of consumers. Judging from the actual situation, at present, AI customer service, whether in text or voice form, is very inconvenient for some special groups such as the elderly and the disabled. Therefore, relevant departments should strengthen supervision of AI customer service abuse and safeguard the legitimate rights and interests of consumers.