Artificial Intelligence in Public Procurement: Aligning Operational Capacity, Legitimacy, and Public Value
TL;DR
The study identifies three tensions: (i) contested performance gains versus social and environmental costs; (ii) data access for model performance versus stewardship duties for public and supplier data; and (iii) opacity versus accountability and contestability requirements.
Abstract
Public procurement is adopting artificial intelligence (AI) more slowly than private‐sector procurement, despite similar operational opportunities. We analyze this gap through Moore's Strategic Triangle (ST), focusing on public value, operational capacity, and legitimacy and support. Based on 38 expert interviews and a follow‐up resonance questionnaire, the study identifies three tensions: (i) contested performance gains versus social and environmental costs; (ii) data access for model performance versus stewardship duties for public and supplier data; and (iii) opacity versus accountability and contestability requirements. We elaborate the ST for AI‐enabled procurement by showing that AI changes how the three existing conditions must be held together. The tensions surface as practical questions about whether gains can be evidenced as public value, who controls procurement data and vendor learning, and whether AI‐supported recommendations can be explained and challenged. The strongest mitigation levers sit in agencies' contracting and decision practices, including work design, data rights and portability, audit trails, and supplier challenge mechanisms.