A Contingent Approach to the Use of Artificial Intelligence in Business Decision Making

This thesis explores the integration of Artificial Intelligence (AI) into organizational decision-making through a contingent approach, addressing a critical gap in current scholarship. While the advantages of AI adoption in business contexts are widely discussed, limited attention has been given to the processes through which AI should be configured and aligned with human judgment. The central research question investigates how organizations can effectively integrate AI into decision-making processes by accounting for contextual, organizational, and technological contingencies. The study develops across three chapters. The first frames organizational decision making by revisiting classical models (Rousseau, Mintzberg, Quinn, and others), highlighting their strengths and limitations in uncertain and dynamic environments. The second chapter examines the main AI models (Constraint Satisfaction Problems, Machine Learning, Generative AI, and Agentic AI) evaluating their suitability to different decision contexts and emphasizing that no single model is universally optimal. The third chapter represents the core contribution of this work: a systematic literature review on the topic, combined with original analysis through the creation of seven distinct business scenarios, highlighting how each decision process is adapted to specific contingencies. The analysis demonstrates that the effectiveness of human-AI configurations depends on multiple contingencies, including the time horizon, the nature of available data, oversight requirements, and organizational capabilities. Strategic, unfamiliar, and high-oversight decisions benefit from human-dominated dynamic design processes, whereas repetitive and efficiency-driven tasks align with AI-dominated search processes. Hybrid and collaborative approaches prove most effective for complex strategic contexts with large datasets. The findings carry significant managerial implications. Competitive advantage does not arise from adopting AI per se, but from designing fit-for-purpose decision-making processes that align AI models, organizational structures, and human roles with specific contingencies.