The Impact of Artificial Intelligence on Decision-Making Systems: A Systematic Review
Abstract
Background: Artificial intelligence (AI) technologies are increasingly integrated into decision-making systems across multiple domains, fundamentally transforming how organizations and professionals make critical decisions. Understanding the empirical evidence regarding AI’s impact of AI on decision quality, accuracy, and efficiency is essential for evidence-based implementation. Objective: This systematic literature review examines the impact of AI on decision-making systems across healthcare, business, finance, and public policy domains, following PRISMA guidelines. Methods: A comprehensive systematic search was conducted across multiple databases, yielding 2,244 papers for review. After duplicate removal and AI-powered re-ranking, 500 top-ranked papers underwent title/abstract screening (threshold ≥4.0), followed by full-text screening (threshold ≥4.5). Only peer-reviewed Q1, Q2, and Q3 journal articles were included in this review. Data extraction captured the AI technology types, application domains, methodological approaches, and decision-making outcomes. Results: The final analysis included 136 studies, all of which focused on healthcare applications. Machine learning (n=98), deep learning (n=89), and neural networks (n=64) were the predominant AI technologies. Studies have demonstrated substantial improvements in diagnostic accuracy (AUC 0.759-0.998), treatment planning precision, risk prediction, and clinical workflow efficiency. Deep learning models consistently outperformed traditional methods, with accuracy improvements of 7-10% across multiple clinical contexts. Conclusions: AI technologies significantly enhance the quality and efficiency of decision-making in healthcare systems. However, the evidence base is currently limited to healthcare, with minimal representation in the business, finance, and public policy domains. Future research should address the implementation challenges, algorithmic interpretability, and cross-domain applications.
