Communicating workforce risk in the AI Era: Predictive Modeling of Individual Layoff Vulnerability and Implications for Corporate and Crisis Communication

Abstract

Rising concerns about worker displacement and job risks due to the accelerating integration of artificial intelligence (AI) into the workplace are reshaping the nature of work. The nature of work is changing as artificial intelligence (AI) rapidly permeates the world of work and is being integrated into all aspects of it. This paper explores the potential of building a multivariate predictive model for determining worker susceptibility to AI-driven job displacement and identifies occupational and AI-exposure risk factors for employee layoff risk. The study employs descriptive statistics, chi-square tests, one-way ANOVA, Pearson correlation analysis, multinomial logistic regression, Random Forest, and XGBoost with data of 20,000 workers in eight industry sectors. The accuracy, macro F1 score, and AUC-ROC are used to measure the performance of the model, and the generalizability is evaluated by 5-fold stratified cross-validation. The highest accuracy is obtained with XGBoost at 94.95%, the highest macro F1-score with XGBoost is 0.95, and the highest weighted AUC-ROC is 0.9948 with XGBoost. The routine task percentage (r = 0.78) and the percentage of tasks automated (r = 0.75) are the strongest positive predictors, while the creativity requirement is quite strongly negative (r = −0.75). For AI adoption, there is a significant relationship with layoff risk (χ² = 7,935.68, p < 0.001), where 80.8% of workers in the high-AI-adoption environment experience high layoff risk. The findings point to potential vulnerability factors such as routine work and automation exposure and protective factors such as creativity and seniority. The study offers actionable insights into reskilling and workforce planning, HRM, and responsible AI use.