AAI_2025_Capstone_Chronicles_Combined

‭ResolveAI‬

‭Introduction‬

‭Problem Formulation‬

‭We are addressing the challenge of streamlining and optimizing the IT help‬ ‭desk/customer support ticket resolution processes by leveraging a hybrid AI-powered workflow.‬ ‭We plan on using a combination of tools and technologies in order to enhance automation and‬ ‭accuracy in ticket handling. The importance of this solution comes from its ability to alleviate the‬ ‭backlog of unresolved support tickets while simultaneously addressing the redundancy that often‬ ‭arises within large volumes of customer/user inquiries. The solution would free up help desk‬ ‭technicians to focus on more complex problems while also reducing operational costs, enabling‬ ‭faster resolution times and improving efficiency (Capacity,2024).‬ ‭The primary end users include external stakeholders (customers/users) who submit‬ ‭support tickets and internal support/ help desk teams responsible for managing and resolving‬ ‭these tickets. Our hypothesis is that a hybrid AI system, combining deep learning-based ticket‬ ‭classification with retrieval-augmented generation for context-informed response, will‬ ‭significantly enhance both the accuracy of ticket prioritization and the relevance of automated‬ ‭responses, ultimately reducing resolution times and increasing overall efficiency.‬ ‭We will be using the Multilingual Customer Support Tickets dataset from Kaggle. This‬ ‭dataset provides real-world customer support requests, which makes it a strong proxy for a live‬ ‭system (Tobias,2025). In the event that there is a demand for additional data somewhere in the‬ ‭process, we may leverage InstructLab to generate high-quality synthetic support tickets and‬ ‭responses. In a production environment, live customer support tickets would be ingested from‬ ‭platforms such as ServiceNow, Jira Service Desk or other platforms.‬

‭4‬

52