Custom AI-powered intelligent agent increases sales by 17% and reduces workload for human agents
A telecommunications client wanted to automate their sales process through an intelligent agent to provide 24×7 service and help customers find product and service information easily. The intelligent agent had to be multichannel so customers could communicate with the telco clients via web, Slack, Telegram, or WhatsApp. Also, the system needed to be able to redirect the conversation to a human agent in case the customer was not happy or preferred to be helped by a human agent.
Cognition implemented an intelligent agent that is multichannel and multilingual, learns from conversation examples, and continuously monitors sentiment and intent.
They customized the solution in two steps: first, training the basic system, then monitoring and fine tuning. In the training step, existing or artificially created conversations were used to train the system initially. Later, a human agent monitored the performance and corrected dialogues with customers. Finally, the new dialogues were used to fine tune the system incrementally. In this way, customers could have a good user experience from the beginning that improved over time. By also monitoring sentiment, they could design the tool to transfer the call to a human agent if it detected that a customer was having a negative experience. The solution combined
text conversation with menus, graphics and images to make conversations very intuitive and efficient.
The system could handle over 60% of the conversations and transfer the rest to human agents, who received typified conversations with the collected information.