AAI_2025_Capstone_Chronicles_Combined

‭ResolveAI‬

‭Abstract‬

‭ResolveAI investigates the potential of artificial intelligence to transform IT support‬ ‭ticket resolution, addressing the challenge of inefficient help desk operations that can lead to‬ ‭delayed responses and increased costs (Brennen, 2025). In the current landscape of customer‬ ‭service automation, while many tools can classify tickets or generate responses, few systems‬ ‭integrate both functions to create a more seamless and intelligent workflow. This research aims‬ ‭to bridge that gap by developing a hybrid system that not only categorizes incoming support‬ ‭requests based on urgency, but also generates responses that are informed by relevant historical‬ ‭data and documentation. The approach involves comprehensive text preprocessing to transform‬ ‭raw customer requests into structured data that can be effectively analyzed. By incorporating‬ ‭both textual content and supplementary features, the model is able to learn contextual patterns‬ ‭that improve the accuracy of ticket prioritization. Initial experiments focused on a multi-class‬ ‭framework; however, challenges with data imbalance and overfitting prompted a shift to a binary‬ ‭classification model that more effectively distinguishes between tickets requiring low versus‬ ‭higher levels of attention. Complementing the classification component is a retrieval mechanism‬ ‭that accesses past tickets and documentation. Retrieved content is then fused with a large‬ ‭language model to produce context-aware responses. Ultimately, ResolveAI demonstrates that‬ ‭combining deep learning-based classification with retrieval-augmented response generation can‬ ‭offer a scalable and practical solution for modern IT support challenges, delivering faster and‬ ‭more accurate outcomes.‬

‭Keywords‬ ‭: Artificial intelligence, help desk automation, ticket classification, RAG‬

‭3‬

51