Combining NLP and Generative Models for Predicting Incident Category and Incident Routing in Incidents Management Systems

One of the most critical infrastructures of any customer-serving organization is its incident management and customer tracking system. Through this system, customers can report their issues by creating a ticket in the application which then gets assigned to the relevant support group, based on the nature of the customer's concern. The current system relies on manual review by human agents to dispatch tickets to appropriate SME’s that can then take the appropriate steps to resolve the problem. Such a process relies on the availability of human dispatchers and suffers from a high error rate. The proposed approach leverages Large Language Models (LLMs) and traditional machine learning algorithms to handle both structured and unstructured data in each ticket. First, generative models are used to pre-process the ticket data and classify the tickets into different categories. Then, these categories are vectorized and used as features for a classification model that assigns the tickets to the appropriate support entities. Moreover, the proposed approach uses advanced LLMs to build a vector database of similar tickets and their resolutions, which can be used to suggest solutions for new tickets. The proposed approach aims to improve the efficiency and accuracy of the ticket dispatching process and enhance customer satisfaction.