A Generative AI and NLP-Based Research Prototype for Call Resolutions and Automation: The Case of a Healthcare Contact Center

Call monitoring at contact centers is still very inefficient and subjective. Call resolution involves post-hoc analysis leading to latency and delay in the identification of urgent issues. We propose a comprehensive novel generative AI (Gen-AI) system, integrated with our custom-built sentiment Analysis and transformer-based topic modeling components, that automates call summarization and topic labeling using a large language model (LLM) and other transformer-based model architecture. The system enhances operations by converting call audio to text transcript, identifying call dispositions, generating structured summaries in a SOAP (Subjective, Objective, Assessment, Plan) format, using in-house custom LLM to classify calls as satisfied, neutral, or dissatisfied, and assigning thematic labels using the integrated transformer-based models for detailed categorization. In addition to the conceptual building blocks of the proposed system, we also built and described a functional software prototype that demonstrates these capabilities through a user-friendly UI that allows for the searching and filtering of calls based on various criteria. This Gen-AI approach can potentially reduce the administrative workload for agents, improve documentation accuracy, and provide actionable insights for leadership, thereby enhancing auditing processes and member satisfaction. Our proposed next-gen Gen-AI-based system has the potential to transform contact centers by improving operational efficiency and service quality, with further validation planned to assess its effectiveness compared to current practices.