Project decision-making through artificial intelligence: a mixed-methods study of public projects in the transition economies

Abstract

Artificial intelligence (AI) is increasingly embedded in organizational processes, challenging established assumptions about decision-making in project environments. Prior research has predominantly conceptualized AI either as a decision-support tool or as a substitute for human judgment, thereby overlooking the socio-technical complexity through which decisions are constituted. Thus, existing literature offers limited insight into decision-making in AI-enabled public sector contexts. This study aims to address this limitation by problematizing the theoretical foundations of project decision-making in AI-enabled public sector contexts. Drawing on a mixed-methods design, the study analyses 80 public-sector projects and 240 decision episodes across Kosovo, Albania, North Macedonia and Montenegro, complemented by 48 semi-structured interviews conducted across the four transition-economy countries. The findings suggest that AI does not function as a deterministic decision-maker but is enacted through ongoing interactions between human actors, institutional structures and algorithmic systems. Three interrelated mechanisms are identified: (i) hybrid rationality, reflecting the interplay between computational inference and contextual judgment; (ii) institutionally embedded distributed agency, capturing the asymmetric distribution of analytical influence, interpretive authority, decision authority and formal accountability across human and algorithmic actors; and (iii) algorithmic mediation, through which AI structures information flows and temporal dynamics of decision processes. This study contributes to project management and organizational decision-making research by developing augmented project decision-making (APDM) as an empirically grounded, mechanism-based integration of previously fragmented perspectives on AI-enabled decision-making. It explains the asymmetric distribution of analytical influence, authority and accountability across human and algorithmic actors, moving beyond human-centric, tool-based and technologically deterministic accounts of AI-enabled decision-making.