Transforming Support Efficiency in Exploration: A Combined GenAI Models Solution for Improved Application Support

Abstract This GenAI-powered and domain-specific solution leverages LLMs to streamline and enhance support for exploration software applications, by delivering rapid resolution and a dramatically improved user experience. Considering a use case of approximately 20 web applications to support exploration processes and a user base of approximately 5,000 users, traditional manual support methods struggle to keep pace. Users often experience support delays, with resolution times averaging over 5 working hours, highlighting the need for more agile and intelligent support solutions. This solution is built on the robust Microsoft Multilingual-MiniLM-L12-H384 and Meta llama 3.1 pre- trained models. The process initiates by segmenting legacy support documentation and historical cases into smaller, manageable "chunks," which are then transformed into embeddings that preserve their semantic meaning and context. These embeddings are stored in a vector database, enabling rapid retrieval. When a user submits a query, it is embedded and processed through the semantic search model, retrieving the most relevant contextual chunks from the vector database. This retrieved context is feed into LLM to enhance the model's comprehension to deliver highly accurate responses. Based on our analysis, this solution is anticipated to reduce user inquiries substantially, cutting them from 4,000 to 2,500—a 37.5% decrease. A significant portion of these inquiries relate to recurring questions that, in the previous support model, required manual responses from the support team. By automating these routine interactions, the solution helps optimize support team capacity and enhances responsiveness to more complex user needs. Furthermore, it has boosted the development team productivity, resulting in 18% reduction in time spent on support tasks, thus allowing developers to focus more on innovation and system enhancements. These results underscore the solution's transformative impact on operational efficiency and user satisfaction. This solution will be integrated across all exploration software systems, serving as the primary support agent. Furthermore, automated ticket generation will be introduced to create and route support request through an integrated tracking and monitoring system whenever the system is unable to resolve user inquiries. Additionally, the system employs a proactive approach, detecting and addressing issues as they occur during user interactions, effectively transforming it into a responsive first-line support mechanism.