The digital transformation of human resource management (HRM) is reshaping the way organizations analyze workforce dynamics and make strategic decisions. Employee attrition, in particular, remains a pressing challenge, generating costs in recruitment, training, and lost institutional knowledge. In this study, we benchmark three machine learning models (Logistic Regression, Random Forest, and XGBoost) on a public HR dataset to evaluate their ability to predict employee turnover. After a rigorous preprocessing pipeline, including feature selection, handling of categorical data, and class imbalance adjustment, our results show that XGBoost consistently achieves the best predictive performance, with accuracy above 97.2% and strong recall for minority classes. Importantly, the use of explainable AI techniques highlights job satisfaction, tenure, and compensation as the most influential factors driving attrition, confirming insights reported in recent HR analytics literature. These findings not only validate the effectiveness of advanced AI models but also demonstrate how digital HR practices can support evidence-based governance, proactive workforce planning, and organizational resilience. The study contributes to bridging the gap between predictive accuracy and interpretability, offering practical implications for both private firms and public enterprises undergoing structural transformation.