The rising prevalence of work-related stress, particularly in remote settings, requires innovative and accessible monitoring solutions. In response, this paper details the development and validation of a mobile application that uses machine learning to detect work-related stress through acoustic biomarkers from the user’s voice. The system relies on a Decision Tree model to analyze key vocal features, including Mel-Frequency Cepstral Coefficients (MFCCs), pitch, and Root Mean Square (RMS) energy, for stress classification. Model robustness and generalizability were rigorously assessed using Leave-One-Out cross-validation. The final model achieved a 75% in accuracy, precision, and recall. These findings validate the efficacy of using acoustic biomarkers as non-invasive indicators of stress. This research contributes to a functional prototype, offering a significant step toward practical, low-cost tools for periodic mental health monitoring scheduled by leaders in the context of modern work environments.