Adoption and Automation Risk of AI Tools in Public and Private Enterprises: A Machine Learning-Based Analysis

Abstract

Artificial intelligence (AI) is transforming enterprises globally, introducing new efficiencies while raising concerns about workforce automation. This study investigates AI tool adoption and potential automation risks in public and private sectors using data collected from 477 respondents in North Macedonia. Pseudo-labels for automation risk levels were generated via k-means clustering, and two machine learning models, Random Forest and XGBoost, were applied to predict high-risk roles and industries. Random Forest exhibited superior stability and performance across scaled configurations. Feature importance analysis identified sector type, industry classification, and AI adoption level as key predictors of automation susceptibility. Results indicate that private sector jobs and industries with high routine task proportions are most vulnerable, whereas public sector and creative roles remain relatively insulated. The findings provide actionable insights for policymakers to design targeted reskilling programs and anticipate labor market shifts. Future work will scale the analysis to the Western Balkans region and incorporate expert annotations to refine risk predictions.