Procurement doesn't lack AI tools; it lacks clean data. Fragmented supplier records, inconsistent taxonomies, and gaps in master data mean AI models just scale existing confusion faster. Start small with one domain, fix the data first, then scale.
Technically, procurement is not primarily behind on AI, if you ask me. And it is mostly because of the operating discipline needed to use the AI reliably. And you take any powerful model, you build the models on, let's say, Azure Foundry, or you build automation models, but it cannot compensate for your fragmented supplier records or inconsistent taxonomies. Or even let's say the unclear process ownerships. These are not the things which AI is gonna measure it and define the verdict. Huge gaps in the master data, commodity mappings, specifications, critical spares planning. So AI can interpret the language and identify the patterns at a very large scale, which we cannot do at this point of time manually. But if your source if your source itself is incomplete, your taxonomy is ambiguous, then no one owns the exceptions when it comes as a part of the results. It would simply produce a faster version of existing uncertainty. So I would say it's— the barrier is more into the execution side where we need to fix the source, and it often starts with a small pool where you will decide about particular domain to start with.
