Chatbots: The End of Wrong Answers through Hybrid Agent + AI Solutions
Wrong answers when chatbots in customer service? There is hardly anyone who has not yet been confronted with this problem. Unfortunately, many conventional chatbot providers – including banks and insurers – still use the wrong answers for their daily life. However, sending automated, incorrect answers to the customer can be prevented by hybrid solutions.
by Maximilian Gerer, CTO e-bot7
In a hybrid solution, the system is first fed historical support requests to train a Convolutional Neural Network. The AI model learns how the best agents respond to customers quickly and accurately, and generates valuable metadata. Once a new message is received via live chat, email, social media or mobile messaging, the system will suggest the best answer to the agent. If the answer is above a certain “confidence level”, which is also called necessary security, it is automatically sent to the customer.
“The combination of agent and AI allows the system to learn with each request and become more efficient. This significantly increases the customer experience.”
Chatbots accept standard requests without customer’s notice
Using the hybrid “Agent + KI” system from e-bot7 ensures that no wrong answers are sent. The Confidence Level can be set by the company at will. Suppose the confidence level used is 98 percent. The topic “Chatbots and Artificial Intelligence to Increase Efficiency in Customer Service” will be explained in detail by e-bot7 founder and CEO Xaver Lehmann on April 11 at the conference “Financial Services of the Next Generation” (Frankfurt School).
If the system is now over 98 percent sure that it has the right answer to the customer’s question, answering the question will be automated by our AI technology. That the employee does not have to intervene anymore. In contrast, if the system is not 98 percent, but e.g. is only 85 percent secure, the employee will receive an AI suggestion directly in their existing CRM system as to how they might respond to the customer’s question. The agent then simply has to click on a submit button if he is satisfied with the response proposal and the message is sent. Since this does not require manual text input or cumbersome copy / paste processes of texts from databases or the like, the employee saves processing time while at the same time training the system.
If it is not the right answer, then the employee can simply adjust the proposal and then send it. The more agents use the system and the more customer requests the system receives, the faster the system learns and improves.
Maximilian Gerer works as a CTO and founder of e-bot7, a company that integrates artificial intelligence into customer service. He was a senior developer for 8 years and has a degree in physics from LMU.
For more information please click here.