Purpose This study aims to examine how public procurement can govern the adoption of artificial intelligence (AI) so that transparency, accountability, fairness and privacy/security are not only stated in policy but delivered in practice. Drawing on public value theory (PVT), we conceptualize procurement as the mechanism that connects the authorizing environment (statewide policy), operational capacity (capabilities across procurement and information technology (IT) and observable public-value outcomes. Design/methodology/approach The study uses a qualitative, multi-source design. Data include ten semi-structured interviews with state procurement officers, Chief Information Officer and IT/security leaders, platform intermediaries and AI vendors across multiple US states, supported by a systematic review of statewide AI policies. Coding followed the PVT structure, examining (1) how policies create obligations, (2) the capabilities required to implement those obligations and (3) how procurement-related AI use cases generate evidence of public-value outcomes. Findings The analysis shows that statewide AI policies influence practice when they are written as procurable obligations that procurement can translate into specifications, evaluation criteria, contract terms and monitoring requirements. Six capability bundles shape whether these obligations can be enacted effectively. Several emerging procurement use cases (e.g. automated document processing, vendor question and answer chatbots, anomaly and fraud detection, spend analytics and supplier matching) produce auditable traces of public value when supported by those capabilities. Their effectiveness relies on a stable partnership between procurement and IT across sourcing and contract management activities. Originality/value For procurement research, the paper positions AI not only as a tool used within procurement but as an object governed through procurement, extending digitalization research with a public policy perspective and applying PVT to an area that has received limited attention. For practitioners, it provides a framework that translates policy into enforceable procurement requirements, identifies the capabilities needed to operationalize them and links concrete AI use cases to observable public-value outcomes.
Patrucco et al. (2026) studied this question.