Leveraging Large Language Models for Intelligent Customer Service Channel Navigation

Customer-centric organizations offer multiple channels for customer service, allowing customers to reach out as needed. A common challenge, however, is that customers often default to the general support hotline, even when other channels may be more appropriate for their requests. The rise of Large Language Models (LLMs) presents an opportunity to enhance customer experience by leveraging their reasoning and decision-making capabilities for optimized channel selection. In this study, we explore the potential of LLMs to route customers to the most suitable service channel within a large European insurance company. Our approach involves collecting data on customer preferences for different channels across various scenarios, as well as their underlying reasoning. Through a series of experiments, we empirically assess the ability of LLMs to classify customer preferences and explain their reasoning. Our findings provide valuable insights into the feasibility of utilizing LLMs as decision-support tools for customer service channel routing.