This paper describes a predictive model trained to predict failures of a computer numerical control machine. The model’s objective is to support the shift from corrective maintenance operations to a preventive strategy where the model alerts operators of potential failures. The model is trained using computer numerical control data and alarm signals to learn situations that can lead to failure. When a possible issue is detected, the operators received a notification message while performing other routine tasks (hourly product quality checks). The aim is to improve process quality, reduce downtime and minimize the defective parts. At the same time, our approach empowers operators with real-time data and predictive insights, enabling them to autonomously perform maintenance tasks and make informed decisions to prevent equipment failures.