Introduction
Consumers of telecom services need to interact with the service provider for many issues including recharge, bill payment, plan change requests, among other things. Our client, one of the largest global telecommunications companies has to deal with tens of thousands of service calls regarding its products and services on daily basis. As a result, the client must manage a large labor force for its inbound customer inquiries. Many of the tasks are repetitive and consistency on quality of services and overall customer experience is also a major concern.

Challenges
Customers expect answers to their questions and call centers to be available around the clock. Telecom companies traditionally struggle with delivering satisfactory customer service without incurring a huge cost. And given that the majority of customer inquiries require the same, repetitive answers from a support agent, there are significant inefficiencies in the current model.

Our client opted for commercial Chatbots (more affectionately known as virtual assistants) to provide a solution to address some of these challenges, with the objective to help free up employees and scale the organization’s support model.
However, the client lacked in-house expertise on how to best leverage chatbots for its large number of inquiries and the complexity that results due to the array of product and service offerings and varying customer base. Realizing that a small percentage of reduction on human intervention can result in significant reduction on operating cost, the client engaged Ubertal to train and optimize its chatbot by leveraging Ubertal’s data science offering.

Solution
Ubertal was commissioned to review the current chatbot use cases and implementation and provide recommendations for an improved architecture that could handle complexity, improve overall customer experience and scale. We proposed new methods to detect when users express their intent and the nature of that intent. While previous work focused solely on responding based on keyword match, after Ubertal helped train a commercial chatbot, we were able to significantly improve the accuracy on service intent identification based on tens of millions of customer messages. Major deliverables from this engagement:

• Use cases identification
• ML and data science best practices
• Metrics automation
• Training data automation and optimization
• Customer retention/sentiment
• Product recommendation/up sell
• Chatbot deployment architecture redesign