Why AI Won’t Rescue a Broken Procurement Process

AI in procurement is the topic everyone is talking about right now. Vendors promise automated sourcing, smart spend analytics and contracts reviewed in seconds. Some of it really works, and I use AI myself more and more. But in many organisations I see the same hope: that a new tool will fix problems that have been around for years. It rarely does.

A picture most procurement teams will recognise

Ask people who have worked in procurement for a while, and you’ll hear a very similar story, whatever the company or industry.

Clean data is the exception. Category management exists on paper, but it is rarely mature. Contracts are spread across different places, and nobody can say for sure who owns them: Legal, Procurement, or the business unit that signed them. In the ERP system, the same supplier is often set up two or three times under slightly different names. On top of that, the same suppliers are booked to different cost types, and those cost types are mapped to the wrong categories.

Put all of this together and you have a pile of chaos. That’s no sign of a bad team. It is simply how procurement functions grow over time, through reorganisations, mergers, new systems and people coming and going. Hardly any team is ever “finished”.

AI makes things faster. Including the mess.

Now imagine running an AI spend analysis on top of this. It will give you a clean dashboard and very confident numbers. They will just be based on the wrong data. If “Müller GmbH”, “Mueller” and “Müller Group” are three suppliers in your system, no algorithm will find your real spend with Müller.

The same goes for processes. If a purchase request goes through five approval loops, AI will route it through five loops a little quicker. Speed only helps when you’re heading in the right direction.

How do you hit 8% savings on a foggy baseline?

Many procurement teams get a clear target from management, for example 8% savings per year. That sounds simple. But how do you reach it if you don’t really know your baseline? How do you plan an e-auction if you’re not sure you have all the suppliers in a category, or the full spend?

You can’t. Or rather, you can report savings, but nobody will fully trust them, including you.

First comes a clear baseline

So before any tool, I always start with the basics:

• Clean categories. A category structure that makes sense for the business, broken down into sub-categories where it helps.

• Clean supplier data. Duplicates merged, suppliers mapped to the right categories and cost types.

• A focused supplier base. Ideally three to four suppliers per sub-category, depending on how far you break things down. More than that usually means you are giving away leverage.

• Clear ownership of contracts. Who manages them, where they live, and when they expire.

Once this is in place, you finally have spend figures you can rely on. Now you can look at each category and ask the real questions: where is the potential? Which lever fits best: a tender, an e-auction, a renegotiation, a demand management project?

Then think about tools

Only at this point does it make sense to talk about tools. And even then, I would ask a very practical question first: does the cost match the benefit? An e-auction tool, for example, is great when you have enough categories with comparable specifications and a healthy number of suppliers. If you only run two auctions a year, a licence for a big platform may not pay off. The tool has to fit the process you have built. A process bent to fit a tool rarely works.

Where AI in procurement really helps

With a clean basis, AI becomes a real help. These are the areas where I use it or would use it today:

• Contract reviews. Summarising long contracts, finding unusual clauses, comparing terms across suppliers. This saves hours.

• Supplier checks. Who is this company, who owns it, how stable is it, what do others say about it?

• Market research. Which suppliers exist for a category, and how does the market fit my spend and my requirements?

• Negotiation preparation. Collecting arguments, thinking through the supplier’s position, preparing scenarios.

• Tender documents. Having texts and specifications cross-checked for gaps, unclear wording or contradictions.

• 
  

In all of these cases, AI makes good work faster. It does not decide for me, and it does not replace knowing the business, the stakeholders and the suppliers.

A simple order that works

• Look honestly at how things run today.

• Build a clean baseline: categories, suppliers, spend, contracts.

• Simplify and fix the processes.

• Decide on your levers and projects.

• Then choose the tools, including AI, that support exactly this.

• 
  

It sounds less exciting than launching an AI pilot. But it’s the order that leads to results that last, and to savings figures everyone can trust.

AI is a great amplifier. It amplifies whatever is already there. So the most valuable AI project in your procurement team might start with a whiteboard, a coffee and an honest look at your own data.