Federación Patronal Seguros S.A. is one of Argentina’s leading insurance companies, with over a century of experience and a strong nationwide presence in both personal and commercial insurance. With an extensive network of agents and a firm commitment to operational excellence, the company manages thousands of daily customer interactions—many of them through recorded phone calls related to claims, policies, and support. To ensure quality, compliance, and customer satisfaction, Federación Patronal continuously seeks innovative technologies to enhance monitoring and analysis in its frontline operations.

The Business Challenge and the Solution

With hundreds of hours of customer calls recorded daily, Federación Patronal faced the challenge of systematically extracting actionable insights at scale. Traditional manual auditing was slow, inconsistent, and covered only a fraction of interactions—limiting the company’s ability to enforce quality standards, identify coaching opportunities, and ensure compliance with regulations and service protocols. A scalable, automated solution capable of analyzing call content, detecting patterns, flagging anomalies, and benchmarking agent behavior against internal best practices became critical to improving customer experience (CX) and operational compliance.

To address this, DIRMOD partnered with Federación Patronal to deploy a GenAI-powered best practices analyzer for recorded calls. The solution uses Amazon Transcribe for accurate speech-to-text processing and Amazon Bedrock to apply large language models to call transcripts. It identifies key moments in conversations—such as empathy markers, resolution clarity, or policy misstatements—and evaluates them against a custom coaching rubric. A retrieval-augmented generation (RAG) pipeline enriches the model with company-specific guidelines, historical calls, and regulatory language.

The solution runs on a secure, serverless AWS architecture, leveraging AWS Lambda, Amazon S3, and Amazon DynamoDB, and delivers both call-level insights and agent-level scoring through a user-friendly dashboard for supervisors and QA analysts.