Application risk of AI technology in bank customer service center - note
One , Application status of artificial intelligence in customer service center
1, The impact of AI on financial industry
Establish financial big data system , Improve the ability of financial multimedia data processing and understanding . Innovative intelligent financial products and services , Develop new financial formats . Encourage financial industry to apply intelligent customer service , Intelligent monitoring and other technologies and equipment . Establishment of financial risk intelligent early warning and prevention and control system .
* For the financial sector , The application of AI mainly includes intelligent customer service , Intelligent outlets , Intelligent marketing , Intelligent wind control .
* outline : Development plan of new generation artificial intelligence
2, Application scenario of AI in customer service center
* （1） Intelligent customer service robot
* Using natural language understanding techniques , On the basis of large corpus , Developing context model based on scenario and business model , So as to realize the natural narration , Intelligent understanding of this purpose .
* Realize the system to automatically understand the customer's questions, answer and handle simple business , Such as balance inquiry .
* Mainly applied to websites , WeChat , Online banking , mobile phone APP The automatic question answering robot in the channel .
* （2） Intelligent voice navigation
* It mainly uses speech recognition technology and natural language understanding technology to understand the voice of customers , And according to the needs of customers, navigate to the corresponding nodes or guide customers to complete business processing .
* Mainly used in self-service voice service , Mobile banking APP And smart devices .
* Self service voice application
* Mainly through IVR The integration of self-service voice menu “ delayering ”, Improve customer satisfaction
* Help customers handle relevant business through interaction with customers , Consultation on implementation issues .
* Mobile banking application
* Mainly in mobile banking APP Upper integrated intelligent voice system , So as to realize the functions of navigation to mobile banking for customers , Handle relevant business for customers .
* （3） Intelligent marketing collection robot - Outbound robot
* Through the design of business scenarios , Realize automatic outbound customer identity verification , collection , Business notice , Satisfaction survey , Product marketing, etc .
* With the transformation of banks , The development of Internet loan business , Actively contact customers for care , Demand for marketing and collection will increase significantly , Outbound robot is a better choice to meet these increasing demands and control the labor cost at the same time .
* （4） Intelligent assistance
* Mainly used in customer service , Robot real-time monitoring seat and customer dialogue .
* When a customer asks a question , Robot understands customer's problems in real time , And give relevant answers and suggestions to the seats .
Robots can monitor the speech of seats in real time , When it was found that the seats were using forbidden words , When the service process is not compliant or the customer is dissatisfied , It can remind and intervene in real time , So as to promote the implementation of customer service standards , The role of improving customer satisfaction .
* （5） Intelligent quality inspection
* Speech recognition technology based on the realization of full recording of text transcribing , And data analysis and mining for the converted text .
* Find out if the seats use illegal words , Is there any nonconformity with the specification , Monitor the mood of the seats .
* Analyze the reason of customer call , Super long call , Repeat call , Reasons for long mute and other calls .
* Explore the causes of customer complaints .
* Forecast trends , Analysis of hot issues .
* Explore potential marketing opportunities .
Two , Risks in current applications
1, The risk of speech recognition error
* Key recognition error
* Wrong identification of transfer amount , Possible loss to customers , So as to bring operational risk and reputation risk to the bank .
2, The risk of robot's answer error
* There is a situation where the customer's problem cannot be identified or the customer's problem is answered by mistake .
* If there is a dispute between the customer and the robot due to the robot's wrong answer , Fear of customer dissatisfaction and complaints .
3, Risks brought by automatic outbound call service
Call out robot makes one round with customers through intelligent semantic understanding technology FAQ Or multiple rounds of dialogue and interaction , In the absence of human intervention , If the business model is not set up properly , Risk of business disputes due to customer understanding differences .
4, Risk of bank information disclosure
* Whether it's voice recognition or customer face recognition , A lot of data materials are needed for model training , These materials are from all kinds of customer data accumulated by the bank at ordinary times .
* This data contains a lot of customer privacy , In case of leakage , Great risk .
* Current model machine learning training for speech recognition , It is often necessary to rely on the supplier to take the materials out of the line , Left behind the hidden danger of information leakage .
5, Independent and controllable risks brought by third party products
* Most banks do not have the ability to develop their own AI algorithms , AI application is basically realized by outsourcing AI algorithm or integration of products and business system .
* These products are black boxes for banks , Unable to control independently .
6, The effect uncertainty risk brought by deep learning technology
* Lack of effective means to verify the effect of artificial intelligence , So as to bring uncertainty risk .
7, Risk of customer satisfaction decline
* While greatly saving manpower , Service experience decline caused by forced diversion , It also brings the risk of customer satisfaction decline .
Three , Coping strategies
1, Speech recognition error risk response
* Control the application scope of speech recognition service , Limit AI technology to low-risk businesses such as queries .
* For high-risk business , Need to add confirmation link .
* for example , Transfer business in mobile banking needs a confirmation page to be confirmed by the customer .
2, Risk response to robot's wrong answer
* Robots handle simple business , Complex business to be handled manually .
* Optimize knowledge , Let customers judge whether this knowledge is the answer to their own question from the reply .
* It's robots that remind customers to serve them .
3, Automatic outbound call business risk response
Increase investment in model training , Improve the ability of model trainer . After a business scenario is designed, you can try it first , Continuous optimization during trial . To be popularized after the model is mature , Adopt iterative way to optimize process continuously and quickly .
* For some important scenarios , For example, marketing scenario , You can filter customers first through robots , Transfer to labor when the customer has purchase intention , Provide professional services and marketing by manpower .
* The whole process recording can be used to record the outbound call process , For ambiguous business scenarios , Confirm the customer's intention by means of key confirmation, etc .
4, Risk response to bank information disclosure
* Establish the bank's own AI deep learning platform , Limit data to the bank , Try not to travel .
* It's true that the data of travel is not available , Do a good job in data desensitization , In addition, data tracking and data control should be done well , Ensure that data is not diverted for other purposes .
5, Independent and controllable risk response
* Artificial intelligence image speech recognition , Natural language understanding is the application of basic service level , Relatively single function . This kind of system is relatively stable in design , Business functions and personalized requirements are realized by business system .
* Banks should strengthen the construction of talent team in the application layer of artificial intelligence , Lay a foundation for the realization of autonomous control of the application layer of artificial intelligence .
6, How to deal with the uncertain risk of deep learning effect
* How to determine the effect of deep learning , Automatic test platform can be established pertinently , And pass the determined test set , Continuously test and compare the results to judge the effect of deep learning .
7, Risk response to customer satisfaction decline
* Change the way of adding robot knowledge horizontally to solve customer problems to vertical domain services , Sort out the business , classification , analysis .
* For simple tasks that robots can solve, robots can solve them , For complex problems that are difficult to be solved by robots, manual work is preferred .
* Formulate corresponding service standards , When the service ability of the robot for a certain business reaches the service standard requirements , And then give the business to the robot .
* Increase customer service evaluation for robots , To evaluate the service effect of robots , And getting feedback from customers . Further optimize the robot .