From Manual to Agentic: A Maturity Model for Artificial Intelligence Adoption in Public-Sector Procurement

A series of technologies has transformed public-sector procurement, ranging from e-procurement and robotic process automation to predictive analytics, generative copilots and now, the most recent, agentic artificial intelligence that can plan and execute multi-step sourcing, supplier-qualification, and contract workflows independently. The debate around adoption often works in binary form, like AI is “either or,” obscuring the fact that there are different requirements for those data, governance and oversight to enable the use of the new capability. It elaborates the Agentic Procurement Maturity Model (APMM) that is a five-level reference model for the journey that public-sector procurement takes from manual processes to governed agentic orchestration, and identifies at each level the organizational prerequisites and dominant risks. The model is a design artefact created in accordance to a design-science and maturity-model procedure that involves identifying six criteria from the literature, creating the artefact following these criteria, and applying the artefact to a synthetic public-sector case and testing the artefact against the criteria. The APMM outlines maturity levels along 7 dimensions; from Manual to Autonomous. Applied to the case, it shows an unbalanced picture that sees an ambition for Level 3 analytics standing on Level 2 data, access, and assurance foundations; and it states that the transition from decision support to delegated action is where accountability, access control and validation are no longer governed by traditional point-in-time controls. The article doesn’t discuss which model to purchase, but rather AI readiness in terms of data, access, and assurance maturity, providing practitioners with a diagnostic and offering a roadmap and researchers a means to find the unsolved problems of access governance, algorithmic fairness, and continuous assurance.