Procurement has never adopted a technology this quickly. It has also rarely had less confidence in the data underneath it.
Two findings worth holding together
Research published by The Hackett Group puts deploying AI-enabled technology among procurement’s top priorities, with more than half of procurement organizations reporting some deployment of agentic AI. In an Efficio survey of CPOs and CFOs at large European companies, fewer than one in five said they had full visibility into their indirect spend.
A CPO benchmark study from ProcureAbility, covering 160 senior procurement leaders, found the same tension from another angle. Half named poor data quality as the leading barrier to scaling AI, and only about one in ten reported scaling AI across multiple processes.
Read together, they describe one problem. Organizations are deploying AI on data they cannot see clearly.
Why pilots stall
The pattern is consistent. A pilot is scoped to classify spend and surface savings. The demo impresses. Then someone checks the output and finds the same supplier under four names, a third of invoices with no usable category, contracts disconnected from the invoices they govern, and cost centers restructured two years ago and never remapped.
The model did exactly what it was asked with the inputs it was given. The inputs were the problem. At that point the pilot either expands quietly into an unbudgeted data cleanup, or stops at the demo.
Four questions to answer before funding AI
- Can we join spend to suppliers to contracts? Without a reliable key across those three, analysis is guesswork in a confident voice.
- Do the people closest to the data trust it? Ask the analysts and buyers who handle it daily. They know exactly where matching breaks.
- Is there a repeatable way to refresh it? A one-time extract produces a one-time answer. Scheduled reports or secure file delivery are often enough; an API is not automatically better.
- Who owns the decision? Let the model generate hypotheses. Keep the savings call, the award and the accountability with people.
One idea to act on
Before your next AI conversation, run a one-page readiness check on a single category. Can you tie last year’s spend to suppliers, contracts and renewal dates without manual cleanup? If yes, that category is ready for AI-supported analysis now. If no, you have just scoped the real first project.
Read more: