How to make smart customer service a productivity tool for enterprises?

About On October 21, 2021, Alibaba Dharma Academy "New Generation Enterprise Intelligent Service Forum" was successfully held in Hangzhou. Wang Weiwei, General Manager of Alibaba Cloud Intelligent Customer Service Business, Product Innovation Center of Dharma Academy, shared the latest progress of Alibaba Cloud’s intelligent customer service, including the omni-channel and full-field coverage of the cloud product matrix, and the solution extending from intelligent services to intelligent marketing scenarios, the first in China Smart Strategy Center; calling on customers and ecological partners to discuss the future development trends of the smart service industry.

1. Cloud intelligent customer service 161a6fa514fea9

1 , development trajectory and application effect: as a productivity tool has experienced a level conversion from the specific effect of the algorithm to the reduction and efficiency increase of in actual combat; around 2015-2016, the characteristics of smart customer service are mainly reflected in Algorithm effect. In the process of continuous solving and verification by customers in various industries, some companies can achieve a 50% cost reduction and efficiency increase effect. This means that the original work of a team of 1,000 people can be solved by 500 people. For the same problem, the remaining agents can invest in marketing, user growth and other levels;

Alibaba Cloud Intelligent Customer Service has recently become the only domestic vendor selected in the "IDC MarketScape: 2021 Global General Conversational AI Platform Vendor Evaluation". Whether it is product strength, revenue scale and long-term development strategy , it has won the recognition of international analysts;

2 . Problems that need to be solved in the after a long period of exploration and development. In addition to the fact that the efficiency improvement and cost reduction from the perspective of the operator have been widely verified; there are still several problems that have not been resolved, such as: B-side Customers are a business chain with different roles such as customers, knowledge operators, sales/customer representatives, agents, managers, etc.; we still have a long way to go in the following three dimensions;

• Perfect customer-oriented experience: customers have a good experience of accurately and quickly solving problems, in the current customer service environment, service cannot be separated from people, and the service process cannot be separated from the participation of agents; the user's subconscious mind is still changing Take manual follow-up solution; whether there is a better solution to promote the integration of man and machine, take care of the user's product experience, make the answer to difficult questions more and more accurate, and the efficiency of the communication process is getting higher and higher;

• Low-cost operations for knowledge operators: used to rely mainly on FAQs or documents, but now incorporates more knowledge structures, such as knowledge graphs, TableQA, and even picture knowledge; in this process, knowledge operators usher in a very large The smart customer service business has just landed in early 2017, and the amount of labeling is very large. Although there is some decline now, it has not yet dropped to the point where operators are very relaxed;

• for personal assistants of account managers: Business changes very rapidly, and knowledge changes rapidly. The attrition rate of the agent team remains high, and some even reach more than 100%. This also means that the agents need to be strong in the service process. Powerful auxiliary tools to solve the problem of efficiency and learning.

2. Alibaba Cloud Intelligent Customer Service Matrix Upgrade and Advantages

In response to these relatively common phenomena we have observed, how does Alibaba Cloud Intelligent Customer Service gradually solve them?

1. Relentless pursuit of a perfect experience from the perspective of customers

• amazing intelligent voice dialogue capabilities; continue to improve the service experience when facing the C-end scene? Q & A personification of the process and to collect complex critical information, such as name, ID number and so on; we now undertake anthropomorphic tone of the original has been greatly improved compared to the same time in extract complex key entities collecting information The above can be completed through multiple rounds of dialogue, which is closer to the actual interactive experience, especially in the long number recognition application, the usability is very high;

• Flexible and changeable Chat UI: In addition to the integrated and complete solution, the customer service experience of the C-end has less attention in the field of intelligent customer service; when facing top customers, a problem will arise. All the capabilities of Alibaba Cloud intelligent customer service are In combination, the front-end part will invite suppliers to participate; after more than a year of polishing, we hope that the user C-end is what you see is what you get, is flexible and configurable, and it provides customers and partners with convenience and operation Interactive experience to quickly meet customer's differentiated product needs;

• Data-driven human-machine integration: agent and the C-end object being served form a better combination relationship, solve problems faster, and improve customer satisfaction. By distribution intelligent and implementation of the recommended capacity, in conjunction with the human scene, based on the customer portraits and historical behavior, the current set of busy lines, such as a comprehensive set of skills developed; we can allow customers to a great extent The problem can be matched to the current agent who is most suitable for solving this problem. While shortening the time to solve the problem, it also takes into account the improvement of customer satisfaction and greatly releases the effectiveness of the agent.

2. "Full link operation tool" for knowledge operators

• Task-based dialogue semi-automatic construction: automatically builds a task-based dialogue flow based on manual conversation logs, including intent, structure, and flow; in multi-round dialogue scenarios. As a knowledge operator within an enterprise, the vertical business may have only a few seats, and when faced with a business with high overall complexity and thousands of sub-scenarios. A person may take on several roles, and the knowledge tree construction of multiple rounds of dialogue becomes a challenging task. Use advanced NLP technology to mine the company’s previous massive communication records, historical logs and other data sources, and automatically generate business maps; on this basis, it is convenient to optimize and adjust business types and frequencies, so that businesses can go online quickly; Emotion recognition will also be online voice recognition soon;

• Automated knowledge mining: from conversation logs, web pages, and documents, and mines knowledge graph triples from unstructured documents. In a non-multi-round dialogue scenario with the same document and log data source, we can dig out FAQs, knowledge graphs and other information, and knowledge operators can store them in the database after quick review, alleviating the lack of knowledge when the business is cold-started;

• Smart sentence recommendation: automatically generates and recommends similar questions based on the sentence recommendation algorithm. When creating a new knowledge point, a lot of related extension questions are needed to improve the accuracy; in the absence of data reference, based on the public and available data set combined with all the available data within the client enterprise, the recommended extension is automatically generated Ask, quickly build knowledge;

• Integrated Voice and Semantic Annotation: Integrated Voice and Semantic Online Annotation System, real-time clustering semantic annotation, real-time annotation data reflow, model training to pull data with one click. The ability to label as a relatively standard configuration, follow-up training, small samples, small traffic test and finally online; separate voice and semantic labeling combined, while labeling the two at the same time, improve the overall efficiency; also support on the basis of the original labeling tools Further enhance the experience.

3. "White Paper Methodology and Courses" for knowledge operators

Alibaba began to establish a customer service business in 2008-09, and established a CCO in 2014. We accumulate and integrate the accumulated experience of front-line product managers, artificial intelligence trainers, and customer representatives, not only to solve the problems encountered by intelligent customer service, but also It is also necessary to solve the customer’s original customer service business problems; after years of exploration and hard work, a set of 3-5 days course system has been created; besides the tools, it can help customers to further improve their efficiency; and improve the operation and landing efficiency.

4 , let the account manager come to "intelligent assistance, dialogue insight and analysis"

With the guidance of this method and the help of tools, knowledge operators can quickly advance the process. long term, we need to solve the customer's product, service and marketing needs. for account managers and frontline agents. What tools and products can we have now that can help everyone?

• Self-evolving intelligent assistance: based on the self-evolving knowledge and SOP mining model of agent feedback and service effect feedback, which reduces agent load while increasing satisfaction and conversion rate. The core solution is to solve the lack of internal knowledge production in the enterprise and reduce the proportion of human-precipitated knowledge; to refine the question-and-answer knowledge, intention knowledge, SOP knowledge, etc. through technology, combined with manual participation in the business; to make the internal knowledge base of the enterprise update and flow in real time;

• Real-time dialogue insight and analysis: analyzes the emotional abnormalities of customers and agents in the process of emotional recognition service through two-way analysis, providing guidance for management and intervention, identifying high-frequency vocabulary and topics in VOC, and quickly understanding customer concerns point. Call content analysis, with the help of NLP capabilities, analyze the main expression content of a large number of calls, and dig deep into the specific content of repeated calls and the main reasons for customer complaints. It can also supervise the execution of speech, intelligently judge the dialogue scene, judge whether the execution of the speech meets the requirements according to the scene, form an analysis summary, and assist the agent in targeted improvement. Finally, it will mine customer statistical data, use online and hotline conversation records as data sources, generate customer statistical data separately, and conduct intelligent marketing on customers based on categories; form an analysis summary to assist agents in targeted improvements. And the conversation content generates customer statistical data to feed back to other business systems to realize the value-added data assets of the call center.

Third, technological product innovation and mutual achievement of ecological partners

In addition to the product dimensions introduced above, there are many underlying technologies in the technical dimension that people cannot perceive during use, which account for the largest part of the Dharma Academy's intelligent customer service investment. Our various engine capabilities have achieved very good results around the world, greatly reducing the cost of manual annotation; ranking first in the four international lists of WikiSQL, Spider, SParC, and COSQL for a long time, with outstanding results;

• Multi-engine capabilities continue to break records: We built a large-scale pre-trained dialogue model, and the accuracy of spoken language understanding increased by 5%. Significantly reduce the cost of manual labeling by 30%. In the QA question and answer engine, the multi-modal FAQ question and answer surpassed the human benchmark by 80.83% for the first time, with an accuracy rate of 81.26%. In the next step, multi-modal VQA, that is, picture-based question and answer, will continue to be integrated.

• digs out the value of private domain customers, and needs to grow toward products and operations: transforms corporate marketing from experience-driven to data-driven . In the past, we could only rely on human resources to accumulate data and sum up experience; now our algorithm plus past historical data can Insight into more information than manual experience, provide customers with personalized recommendations, match better seats for services; in the future, algorithms will evolve and learn independently with data, use data iterative model capabilities, and service satisfaction and marketing conversion rate are based on the original foundation Further improvement; in retail scenarios such as stores, merchants, product legal person information, etc. We can provide some capabilities to support intelligent customer service, and hope to extend it from service to marketing. Therefore, we put forward service-oriented marketing: using customer operation thinking to build starts with customers and ends with customers " to help customers find more suitable account managers to solve their problems.

• gives the brand a concrete anthropomorphic image: interactive technology is changing in real time, like the recent hot Metaverse metaverse concept; it is impossible for all live broadcasts to be carried by humans. This is not the most economical way for many businesses. We have The interactive form of interactive effects gives a branded anthropomorphic image. In the current scenario, the application of digital human technology can largely solve the problem of closer interaction experience.

IV. Alibaba Cloud Intelligent Customer Service Matrix Architecture

As shown in the figure below, the bottom layer of the product capability architecture of intelligent customer service is the communication ability, the middle layer is the advanced AI technology of each laboratory of Dharma Academy, and the top is the intelligent customer service product matrix; we have a very good degree of openness, and Covers all industries of Alibaba Cloud; in the process of external landing, rich open methods support mutual business connection between customers.

The previous achievements are closely related to ecological partners. In the past few years, dozens of partners have worked with us to provide services to customers. Everyone has worked in various industries for many years, specific to the industry, specific to customer segmentation scenarios, such as tax, legal, financial and other businesses; among them, there are three most important and most important capabilities: delivery capability; product/solution capability; channel Sales capacity .

We effectively combine the product and technical advantages of Alibaba Cloud's intelligent customer service with the comprehensive capabilities of our partners to enable services to create more corporate value and achieve business growth.

Copyright Statement: content of this article is contributed spontaneously by Alibaba Cloud real-name registered users, and the copyright belongs to the original author. The Alibaba Cloud Developer Community does not own its copyright and does not assume corresponding legal responsibilities. For specific rules, please refer to the "Alibaba Cloud Developer Community User Service Agreement" and the "Alibaba Cloud Developer Community Intellectual Property Protection Guidelines". If you find suspected plagiarism in this community, fill in the infringement complaint form to report it. Once verified, the community will immediately delete the suspected infringing content.