We built a custom AI-powered customer service representative for a major telco in order to allow customers to search plans, hundreds of support articles, and FAQs effortlessly with human conversations.
Problem Statement
Telco customers should be able to ask questions to a virtual customer service representative (CSR), because self-service customer support is often frustrating for customers. When given a question from a customer, the virtual CSR should search the knowledge center for a matching question and respond with the answer exactly as it appears in the FAQ page.
Architecture
The system leverages a large language model (LLM) in combination with a knowledge base of company docs, plans, and FAQs in order to assist the user with accurate responses.
Key Features
- Auto-updating knowledge base: We developed a React web application that allows users to specify HTML and PDF documents to source information from in order to build the knowledge base. In a future version, we will make it so that users can schedule updates to the knowledge base.
Leave a Reply