Government call centers play a fundamental role in handling citizen inquiries and complaints. However, evaluating service quality, ensuring compliance with protocols, and detecting potential irregularities in interactions pose significant challenges due to the high volume of calls and the inherent inefficiencies of manual auditing, making it a costly, slow, and error-prone process. This paper presents an AI-driven solution for automating call audits in the “148” government helpline of Buenos Aires Province. The system leverages open-source artificial intelligence models, deployed within proprietary infrastructure, ensuring the privacy and security of sensitive data handled in these services. The project consists of two main stages: Automatic call transcription: We use OpenAI Whisper, an open-source Automatic Speech Recognition (ASR) model based on Transformer architectures, capable of processing multiple languages and handling variations in audio quality. Prior to inference, an audio preprocessing pipeline is applied, including silence removal, noise filtering, and volume normalization, to minimize transcription errors. Automated classification and auditing: Once transcribed, the calls are processed using a Large Language Model (LLM) optimized for classification tasks. A few-shot prompting approach is initially applied to identify relevant linguistic patterns. The model is then fine-tuned using LoRA (Low-Rank Adaptation) with a curated dataset of call examples, covering multiple categories and audit criteria such as violent or abusive language detection, proper request for personal identification (CUIT/DNI), and compliance with operational protocols in citizen-operator interactions. Automating this process significantly reduces the time and resources required for call auditing, providing objective insights into service quality, identifying areas for improvement, detecting behavioral patterns, and enhancing oversight of the “148” helpline. This ensures a better citizen experience and greater compliance with regulatory standards.