LLM-Enhanced Cloud PBX-IP for Real-Time Emergency Call Triage: A Preliminary Evaluation

Response time in emergency services is critical for public safety and life preservation. Traditional human-based triage systems may suffer from latency, scalability limitations, and susceptibility to human error. This work proposes and evaluates a cloud-based PBX-IP system integrated with Large Language Models (LLMs) for automated emergency call triage. The architecture combines Voice over IP (VoIP), Speech-to-Text (STT), and Retrieval-Augmented Generation (RAG) to enable real-time classification and routing of calls. Experimental results show an accuracy of 83.40% and an average response time of 2.4 seconds, demonstrating the feasibility of the proposed approach.