Traditional service assurance chatbots primarily work using keyword searches on article knowledge bases. This method introduces inefficiencies with respect to the semantic relevancy of retrieved information, which leads to dissatisfied customers and increased help desk call transfers. We’ll explore how retrieval augmented generation (RAG)-powered chatbots can go beyond simple lexical article retrieval to engage customers with dynamic conversations based on data from multiple knowledge sources like customer history and enterprise data. We’ll present insights and learnings from our service assurance chatbot pilot. We’ll discuss our RAG design and how it’s used to understand customer needs and drive improved business outcomes.