AnBot covers all conversation log information of Ping An. In a data-driven mode, it can construct the knowledge base quickly with the smart recommendation schemes, thus reducing the online operation cost. Its knowledge base operation efficiency is 5 times higher than that of the last generation robot.
AnBot has many core abilities such as inferring other information from one fact, smart error correction, long-sentence understanding, and multiple rounds of Q&A. At present, the accuracy of knowledge base retrieval based on deep semantic understanding has reached 96%. And in multiple-round dialogues, the entity recognition accuracy is over 95%.
As the brain of the smart customer service of Ping An, AnBot can provide customers with a full set of smart customer service solutions. The non-inductive two-way switching between robot and the customer service rep, semantic understanding, emotion recognition and other services can be performed, with customer service efficiency doubled and the risk of complaints effectively controlled.
Used for bilateral or multilateral problems with unclear business attributes consulted by users. The central control robot, supported by powerful algorithms, can quickly clarify users’ intentions through questioning, thus realizing the one-stop online service experience for users.
Identify customer needs via extensive entity recognition, switching freely between tasks and Q&A.
Via a unified service portal, the central control robot identifies the customers' intentions, and assigns the Q&A tasks to the corresponding department.
The unified service portal identifies customer needs and distributes them to corresponding business segments such as life insurance, property insurance, banking and securities through smart Q&A.
If a robot is not smart enough, it often fails to accurately recognize customers' problems, resulting in a poor customer experience and waste of efforts. AnBot, with the most advanced NLP algorithm, is capable of deep semantic understanding with high Q&A accuracy.
In response to scenarios where context semantics cannot be associated, man-machine dialogue is rigid and user expressions must adhere to the robots rules, AnBot is capable of context semantic analysis and supports multi-round Q&As, thus answering questions smoothly and naturally.
For the weakness that users cannot be actively guided during online information collection and processing, smart questioning for service information can be achieved while entity extraction is performed, to connect the core business system for business processing.
For situations where robots cannot follow the changeable thinking of humans and cannot answer questions during processing, AnBot can switch between tasks and Q&A freely, such as answering questions whilst processing business instructions.
Robots are unable to learn, so they need manual knowledge maintenance, which results in cumbersome efforts, frequent missed instructions and high knowledge base maintenance costs. Deep self-learning, together with manual assisting confirmation, can significantly reduce labor costs.
When one service portal undertakes multiple business lines with a robot maintained separately for each, multiple robots cannot cooperate. The central control robot can identify the customers' intentions and assigns Q&A tasks to the robot in the corresponding business field.
Ping An Bank