Decoding the Silent Resignation: How AI Spots At-Risk Employees Before They Leave

Employee attrition poses a significant challenge for modern organizations, with the phenomenon of ”silent resignation”—where employees become disengaged long before formally resigning—exacerbating the problem. Late detection of at-risk employees leads to increased turnover costs, loss of institutional knowledge, and decreased team morale. This study presents an AI-driven solution to proactively identify employees who are likely looking for a new job. Using a public HR dataset, we developed and compared machine learning models, including Random Forest and XGBoost, to predict an employee’s intent to switch jobs based on demographic, experiential, and company-related features. Our results demonstrate that the XGBoost model achieved a strong predictive performance with a ROC-AUC score of ۰.۸۲, effectively identifying key indicators of disengagement. The most influential predictors were found to be the city’s development index, company size, and years of experience. This work contributes a robust framework for early attrition risk detection, enabling HR departments to implement targeted retention strategies before valuable talent is lost.

Keywords:

Employee attrition , silent resignation , predictive analytics , AI in HR , machine learning , XGBoost , feature importance