AAI_2025_Capstone_Chronicles_Combined

‭ResolveAI‬

‭inaccurate predictions, such as misclassifying low-priority tickets as high priority or vice versa,‬ ‭could lead to suboptimal ticket routing, impacting resource allocation and resolution times.‬

‭Performance Impact‬

‭Our project was not focused on answering a research question, but rather demonstrating‬ ‭how a system like ResolveAI can operate effectively in a real-world production environment.‬ ‭Specifically, our goal was to alleviate the burden on limited technical support engineers by‬ ‭intelligently routing incoming support tickets based on priority. The performance of our‬ ‭bidirectional LSTM based priority classification model is critical to the success of the system as‬ ‭a whole. Accurate prioritization determines whether a ticket should be escalated to a human‬ ‭agent or redirected to an automated LLM service using retrieval-augmented generation.‬ ‭While our benchmark metrics were not the strongest, the model still demonstrated the‬ ‭potential for meaningful routing decisions. Inaccurate predictions, such as classifying low‬ ‭priority tickets as high priority, could lead to a misallocation of limited human resources, while‬ ‭the false negatives could delay the resolution of issues that are truly urgent.‬

‭21‬

69