AI in Procurement: What’s Real, What’s Hype
Unpopular opinion from someone selling AI-native procurement: AI cannot negotiate, and anyone telling you otherwise is selling you a dashboard.
I set up the data analytics function for a leading multinational procurement consulting firm. I lived through the e-sourcing era at Ariba and Opera Solutions, when “the platform will transform your procurement” was the pitch of the decade. Some of it was real. A lot of it was hype.
AI is at that same moment now. So here is an honest scorecard from a practitioner, including the parts that work against our own pitch.
What AI does brilliantly today
- Spend categorisation. Classifying hundreds of thousands of messy invoice lines used to be months of analyst pain. AI does it in days, and handles the dirty data that broke the old rule-based tools.
- Supply market research. A category brief covering suppliers, cost drivers, price trends, and alternatives used to take a junior analyst a couple of weeks. AI produces a credible first draft in hours.
- Savings modelling and drafting. Scenario maths, business cases, deliverable drafts. Fast, tireless, consistent.
What AI genuinely cannot do
- Negotiate. A negotiation is a relationship under pressure. Reading the room, knowing when to walk, trading concessions across a multi-year relationship: that is human work and will stay human work.
- Build supplier trust. Suppliers extend favours, flexibility, and first allocation to people, not prompts.
- Own the decision. AI produces analysis. Someone with experience must decide which finding is real, which is a data artefact, and which saving will survive contact with operations. Judgement does not come pre-trained.
AI-native versus AI-assisted
The distinction that matters is AI-native versus AI-assisted. AI-assisted is a traditional team that occasionally asks a chatbot. The cost structure does not change; you still pay for the pyramid. AI-native means the legwork is done by AI by design, and you pay for experienced practitioners’ judgement on top. Same analytical depth. A fraction of the cost and timeline. That changes advisory economics structurally. It is why this firm is built AI-native from the ground up rather than bolting AI onto the old model.
So: real, but specific. AI removed the legwork. It did not remove the practitioner. Be suspicious of anyone claiming either extreme.