China Unicom faced challenges in operating a customer contact center serving users in 31 provinces. Despite using chatbots and agents, customer satisfaction was low. The chatbot had difficulty accurately answering questions due to voice recognition issues. Agents also faced challenges due to different products and services across provinces, resulting in high labor and operational costs.
Improving Recognition and Understanding
With a vision for improvement, China Unicom implemented AI technologies, including improved dialect recognition and intent understanding. This led to higher customer satisfaction and improved accuracy in FAQ and intent recognition. By using emotion recognition, calls were connected to agents in a timely manner. Voice recognition was also improved through standardized labeling and incremental learning.
They also predicted the customer's emotional status using converged data and used phoneme data to increase the quantity of training data for emotion status recognition, improving the efficiency and accuracy of emotion status recognition. Data categorization, handling, and security were also prioritized.
The data was handled according to security regulations, compliance requirements, and technical specifications throughout its lifecycle. Routine monitoring, warning, audit, and emergency handling mechanisms were established to ensure data security. Service switchover and recovery could be completed within minutes with the help of the active-active (DR) architecture. To assist contact center agents, the dialog data between agents and users was used to build an industry dictionary library.
Coupled with a new word discovery algorithm, this allowed for the automatic generation of core dialog content and critical information based on the post-trained foundation model of the communications industry. This helped agents by recommending communication corpuses in real time and standardizing and improving the efficiency of order creation. Agents were also assisted by a heuristic human-machine interaction software, reducing average call duration and the number of repeated calls.
Additionally, the technology offered auxiliary tools such as intent recognition, knowledge recommendation, and automatic case filling to improve service efficiency and enhance the user experience.
Implementing AI Technology
China Unicom's contact center not only serves telecommunication needs but also provides services and emergency relief in major events and incidents. During the flood in Zhengzhou, the center quickly responded and assisted millions of users with emergency help and consultation requests. Generative AI played a crucial role in providing immediate relief.
The center's foundation model is trained using a vast amount of communication data to improve voice and semantic recognition accuracy; enhancing its humanization level and the relevance of generated content. The implementation of AI has also reduced costs for the contact center, resulting in significant benefits for China Unicom.
Furthermore, the contact center provides services beyond customer service, including IoT, public cloud, online digital services, and terminal services. It serves enterprises in various industries, making it a valuable asset for China Unicom in the billion-dollar enterprise customer service market.
