In most organizations, the conversation about AI in procurement starts with enthusiasm and ends with a veto. The CPO sees a tool that can research the market and prepare a specification in minutes; the CISO sees tender data flowing into a model the organization does not control. Both are right — and that is why so many AI pilots in procurement never reach production
In the new article, we show why neither a ban nor blanket permission resolves this conflict — and how architecture does: an AI gateway that separates two operating regimes. Market research goes to the best external models, procurement-process data stays in the local environment, and the boundary is enforced by technology, not by policy. Add to that an evidence package for every interaction — compliance with public procurement rules and the AI Act that can be shown to an auditor, not merely declared