AI in procurement has generated no shortage of ambitious claims this year, autonomous agents negotiating contracts, fully automated sourcing decisions, procurement functions running themselves. The reality on the ground looks different, and arguably more useful.
Recent industry research points to a wide gap between intention and deployment. A large majority of procurement organisations are now piloting or actively scaling some form of AI, a sharp rise from just a few years ago. But when it comes to running AI at real scale across the function, adoption drops sharply, often into single digits. Procurement, in fact, shows up as one of the slower-moving functions for AI adoption compared with the rest of the business, not because the opportunity isn’t there, but because the groundwork often isn’t.
The most commonly cited barrier isn’t budget or IT policy, it’s a knowledge gap: teams that don’t yet know how to apply AI to their specific workflows, or don’t trust the output enough to act on it without checking everything by hand. Siloed working compounds the problem. AI applied to a fragmented, disconnected process just automates the fragmentation faster.
Where AI is actually delivering value right now tends to be narrower and less glamorous than the headlines suggest: spend analytics that surface patterns a person would take days to find manually, contract review that flags missing clauses or unusual terms before they become a problem, faster and more consistent supplier onboarding, anomaly detection that catches an invoice or a rate card drifting from what was agreed.
What these use cases have in common is a connected data foundation underneath them. AI applied on top of clean, unified supplier, contract and transaction data finds real patterns. AI applied on top of five disconnected spreadsheets mostly finds the gaps between them.
At TSM, we implement Ivalua’s Source-to-Pay platform, which brings supplier information, sourcing, contracts, risk and transactions into one connected operating model, the foundation that makes AI in procurement actually useful rather than another tool layered on top of the same fragmentation. If your organisation is further along in AI planning than AI deployment, that gap usually says more about the data underneath than the technology itself.
