OPTIMIZATION OF CUSTOMER REQUEST PROCESSING USING NEURAL AGENTS AND RAG ARCHITECTURE

Abstract

This article examines ways to improve the efficiency of customer support systems in telecommunications and IT companies through the integrated application of neural AI agents and Retrieval-Augmented Generation (RAG) architecture. Traditional automated support systems (rule-based chatbots) often place a significant burden on first-line support during large-scale incidents, such as failures of base stations or communication backbone networks. The study proposes a method for the automated processing, analysis, and clustering of incoming customer requests in real time. The proposed multi-agent system not only responds to standard inquiries by leveraging a local knowledge base through RAG but is also capable of identifying hidden relationships among incoming messages. When correlated requests are detected, the neural AI agent automatically creates a unified root incident ticket (Problem Ticket), assigns a dynamic priority level, links duplicate customer requests to the root incident, and sends initial notifications to affected users. The implementation of this architecture significantly reduced the average time required for the initial localization of system failures and substantially decreased the duplicate workload for support operators.