Source: Harvard Business Review. Summary and analysis by S2V Advisory.
What happened
Harvard Business Review published an article by Heiner Himmelreich, Ilan Oshri, Paolo Scala and Anas Zaidani arguing that procurement is unusually well suited to agentic AI, meaning AI systems that can carry out multi-step tasks rather than simply answer questions. The authors point out that procurement work is structured, its outcomes are financially measurable, and it contains many judgment-heavy tasks that agents can support. As enterprises move agentic AI from experiments into production, they argue, procurement is a natural place to prove it out.
What it means for buyers
We agree procurement is fertile ground for agentic AI, with an important condition: the agent is only as good as the process and data underneath it. Procurement’s structure is an advantage only if that structure actually exists in your organization: clean supplier records, contracts linked to spend, a consistent category taxonomy and clear decision rights.
Before deploying agents, it’s worth answering three questions:
- Which decisions should the agent support, and which stay with people? Agents are well suited to preparing analyses, drafting RFx documents, monitoring renewals and flagging exceptions. Approving awards and declaring savings should remain human decisions.
- Can the agent’s inputs be trusted? If supplier and spend data can’t be joined reliably, the agent will produce confident answers built on weak foundations.
- How will value be measured? Agent productivity is not the same as realized value. Track results in invoices and operating performance.
Organizations that get the foundation right will move fastest. Our Advise service includes AI strategy and data-foundation roadmaps built on exactly these questions.