Wofong Technology has an omni-channel customer service system Udesk, AI scene landing expert GaussMind, and integrated intelligent sales

Complete product lines such as the post-service platform ServiceGo and customer experience marketing Cloud Weifeng are committed to applying AI technology to

Enterprise marketing customer acquisition, sales management, customer service and internal sharing services and other scenarios, for consumer goods, manufacturing

Manufacturing, life sciences, automotive, central state-owned enterprises, digital government, finance and other industries to provide customized solutions. Let me introduce you to a standard success story of our company - the AUX Group.

Project background

AUX Group was founded in 1986, the industry covers home appliances, power equipment, medical, real estate, financial investment and other fields. It has been ranked among the top 500 enterprises in China for many years, and has been recognized as a national enterprise technology center, a national technology innovation demonstration enterprise, a national intellectual property demonstration enterprise and a permanent unit of postdoctoral workstation. In the list of China's top 500 private enterprises in 2020, AUX Group Co., Ltd. ranked 97th.

In 2020, the revenue of AUX reached 70 billion, and the expansion of the business scale led to the increase of after-sales service workload. The original after-sales service system was a traditional manual mode, and each system operated independently with manual data statistics. The work was heavy but inefficient, and it was difficult to cope with the rapid business, customers could not call in, work orders were slow, complaints about installation problems could not be handled, and so on. As a result, the number of customer complaints gradually increased, but at the same time, the customer service pressure is very large, and the service quality is difficult to guarantee, so that the brand impact of AUX is in a very passive situation.

Improving the existing service system, comprehensively improving the work efficiency and service system of customer service, and fully guaranteeing and enhancing customer satisfaction have become an important problem facing AUX at present.

AUX internal service is the traditional model, there is no intelligent system, and there are multiple systems, each system operates independently, the after-sales service work is heavy and inefficient. For example: customer service staff receive installation appointment phone - record the relevant content - login work order system - create work order - work order completion - export relevant customer information - telephone return visit - record return visit information - sort out the relevant content form (a large amount of data), the amount of data is very large every day, the pressure is prone to error.

Overall AUX Group is divided into three types of business needs: unified integration: The overall business of AUX needs intelligent robots, call centers, work order systems, quality inspection systems and other multi-product collaboration support, product suppliers have high product perfection requirements, need to have integrated capabilities of service manufacturers.

ASR/NLP customized model training: Because service services have high requirements for address and date recognition, it is necessary to train customized speech semantic models and provide algorithm customization capabilities.

Interconnection with the current business system: the business needs to support work order query and work order flow, and the voice robot needs to support customer identity verification and work order information query during the call, and trigger the generation of work order after the call.

In the first stage of the project, we first conducted a one-month demand survey and business combing on the business departments, after-sales service departments and customer service departments. AUX business increased rapidly, with a turnover of 70 billion yuan in 2020, and after-sales service business also soared. At present, AUX is a traditional after-sales service model, independent customer service platform, independent work order system, manual return visit, quality inspection, statistical data, heavy workload and low efficiency. Due to limited customer service personnel and huge business volume, the return rate of manual outbound calls is extremely low, a large number of customers do not return visits, unable to obtain real feedback from customers. Meanwhile, the customer service quality inspection is random, and the coverage rate is only 1%, which cannot form an effective evaluation. Therefore, we consider building an integrated service platform system integrating intelligent incoming and outgoing call robots, call centers, agent assistants and intelligent quality inspection. Based on business characteristics, it can accurately identify phone numbers and addresses, and the accuracy rate of ASR identification can reach more than 90%, so as to ensure the data docking of service quality with existing customer systems and work order systems. To achieve data acquisition, automatic ordering, automatic return visits and other services, to achieve full quality inspection of customer service staff, objective and fair assessment of customer service quality, and excellent cases to improve service quality.

The agent assistant can assist the artificial agent in real time, record the conversation in real time, monitor the status of the agent conversation, remind the agent in time, and reduce the complaint

Service process, speech reminder, navigation installation, maintenance, reminder and other service process guidance, service speech recommendation, avoid missing business process, ensure service and marketing quality knowledge recommendation, support intelligent training automatic recommendation of related knowledge, real-time quality inspection, sensitive information timely reminder warning, to avoid causing service public opinion.

According to the demand, the 24-hour telephone inspection notice is required for service stations all over the country, which brings great service pressure to the manual customer service, so we provide intelligent quality inspection solutions. In the 24-hour service station inspection scenario, the robot makes an outbound call to the national service station, and then the robot records and enters the intelligent quality inspection system for service inspection. The scope of inspection includes: whether the service station answers the phone within the specified time, and whether the service station name is accurately reported. Get through the voice robot and the quality inspection system, realize the real-time transmission of the recording after the robot outbound call, intelligent quality inspection, and promote the standardization and efficient service of the service station personnel.

In the connection with AUX work order system, we are divided into incoming scenario and return visit business scenario. There are three types of incoming scenarios: Query information about historical work orders corresponding to incoming numbers or standby numbers in real time: Automatically obtains installation/maintenance device type, service address, and warranty information based on the historical work orders

Automatic address collection and automatic work order generation: When the historical work order is missing, the voice robot automatically carries out address recognition and address structure, and the work order must be collected and automatically generated

Call reminder automatically generates reminder information

The return visit scenario is divided into: after the work order is completed, the automatic trigger to generate the work order AI outbound call back visit: After the installation/maintenance of the work order service is completed, the automatic trigger to generate the work order to return the outbound call task, and the outbound call is quickly completed;

Structured presentation of satisfaction return visit results, standardized data management: after the robot return visit, structured presentation of return visit results, each return visit information results at a glance; Full coverage of completed work order satisfaction return visits.

Automatic recall for disconnected numbers: You can set the number of recalls and the recall time for disconnected numbers

Then it interconnects with AUX CRM system, and the outbound call data is automatically imported into the outbound call task. Data is interconnected and transmitted through interfaces, and outbound data is automatically transmitted, eliminating the need for complex offline import and export operations. The service system directly starts the task without logging in to multiple platforms. At the end of the robot outbound call, the questionnaire information is automatically structured back to the customer's business system, and the questionnaire information is clear at a glance. Finally, it interconnects with AUX' call center, and the call component is embedded in AUX' self-developed customer service system. Listen to call events on the call component and display them on the screen. The customer's own customer service system can directly dial and answer the phone, complete the daily work such as user return visit and hotline service.

The e next stage is the training of the engine, udesk's engine, as its Chinese name suggests, allows the machine to think like a human, avoiding the loss of information caused by a simple series of modules. The engine can obtain more abundant speech semantic information, combine with the robot speech to reproduce the dialogue context, and realize the integration of speech and semantic to realize the iteration of the more intelligent self-learning effect.

Open up customer domain data, ASR professional model optimization

The historical data of the home appliance industry, such as problem speech, knowledge base, FAQ, and work order, are introduced into the recognition engine, so that the machine can do speech recognition in the "limited scene" like a human. To expand the dictionary of rare words in AUX address library; The language model is optimized for AUX address library

Voice data annotation - Achieve customized experience with industry models

From the voice contains a lot of data with business characteristics, accents and verticals with industry characteristics

The more vertical the application scenario, the higher the customization effect. The annotation of user scenario data is significant compared with that of general use. The greater the increase in user data, the greater the improvement.

Application effect

Greatly improve the customer service experience, achieve 7*24 hours of intelligent customer service, fast connection. Robots independently solve complex continuous problems (eg: automatic repair, automatic order), the coverage rate of business problems reaches more than 85%, service efficiency is improved, and consumer service experience is improved. With human-machine collaboration, customer service personnel only need to face less than 20% of the difficult/unknown problems that robots fail to solve. The service focus has shifted to optimizing the service process, optimizing the customer service experience and other aspects of value services, and the value of employees has increased. With the continuous deepening of the understanding of the business, the two sides jointly plan and continue to build, Wofeng Technology will improve the other products of the integrated product line for Okes, while relying on the user portrait and intelligent prediction ability of NLP-PaaS, better activate the data of various channels, and improve the comprehensive user service system such as product marketing, business operation and internal sharing.