Procurement in the AI era

Procurement is being handed a problem it was not designed for. Traditional supplier evaluation assumes a product that is roughly the same next quarter as it is today. AI vendors ship a different product every six weeks. The thing you evaluated is not the thing you are running.

This changes several contract fundamentals. Model update rights, deprecation notice periods, data residency guarantees, exit and portability provisions, liability for agent-initiated actions — these are now first-order negotiation items, not boilerplate at the back. Procurement teams that treat AI contracts like SaaS contracts from 2019 are leaving significant risk on the table.

The second shift is that the supplier base is consolidating in some layers (foundation models) and exploding in others (application and agent vendors). Due diligence on a two-person AI startup selling a mission-critical workflow is a different exercise from due diligence on an enterprise software vendor. Many procurement functions are not staffed for it.

The firms getting this right are pulling procurement into the AI strategy conversation early, not treating it as a back-end rubber stamp. The best governance signal in most companies right now is whether the Chief Procurement Officer has an AI point of view.

If they do not, that is the first hire.