The Procurement Blueprint is brought to you by:

Free: The AI Bubble Report for Procurement Leaders

AI is changing software spend faster than most procurement teams can keep up with. This report looks at what that means for procurement, from 300% growth in corporate spending on AI-native tools in 2025 to the 66% of software spend companies are wasting on unused licences. It also unpacks why buyers are being pushed out of familiar pricing models and into something much harder to control.

The Editor’s Letter

HI THERE

Writing this newsletter every week normally just… ticks along in the background of my life. Easter disrupted that a bit.

The break itself was actually really nice. Now the two weeks at home with the kids, though… It felt like running a very small company with no structure and two very loud senior stakeholders with very strong views on absolutely everything. So sitting down to work today was lovely.

I’m not exactly easing back in either. I’m heading to Lisbon tomorrow for SAP Connect, speaking about some of the cool procurement agents we’re building alongside SAP. If you’re going to be there, come say hello! I’m also sneaking in seeing Celia (previous guest writer) for her birthday, which makes the trip feel slightly more exciting than just work travel.

Then on the 23rd I’ll be at ProcureTech Unpacked, doing a keynote on bias, hallucinations, and the legal side of AI in procurement. It’s one of those areas no one seems to really care about yet which is why I believe it to be so important and why I made this week’s main article about it.

And…..I also managed to get the last two spots for the event my readers here! (just email Joël and ask nicely! 🙂 )

What I like about the event is that all the vendors are there to explain their category rather than pitch their product. I wish I had a conference like this during my years as “Head of”!

You will be talking to people who are dealing with the same decisions and issues and you and not just wandering around collecting slides you’ll never open again.

The event is by application only and aimed at those who own the procurement digital roadmap, which is why I think it is going to be super exciting and successful so grab the tickets before they go! Today is the last day.

And as always, there’s something free and useful waiting for you at the bottom.

Procurement by Design

BEFORE WE MAKE THE DECISION

Someone in a small city US procurement team was drafting an RFP late in the afternoon whilst working through the usual pressure of timelines and stakeholder expectations.

There was a usual sense of urgency we all experience when a tender ‘must’ go out, after all, it was a five-year, $2.7 million worth deal so not petty change.

They opened ChatGPT, in an almost instinctive way, like opening Excel or Word used to be, pasted in parts of the requirement and asked it to refine the wording, improve the structure, make it sound more complete, more professional….Pretty normal stuff.

At some point, their request shifted a tiny bit.

Instead of just improving the text, this procurement professional asked the AI tool to shape the language in a way that would make it harder for one supplier to meet the criteria, but ensuring this would not stand out or trigger questions, just quietly steering the outcome.

This ask was a small adjustment that is almost unperceivable once it is inside a well-written document.

The output came back exactly as expected: clear, structured and credible.

And once it was reviewed, it moved forward through the process just like any other piece of procurement documentation.

The RFP was issued, suppliers responded, the evaluation took place, and a contract was awarded, with everything appearing coherent, justified and defensible.

The issue only surfaced later, when the outcome itself started to raise questions, because the selected supplier was significantly ($1 million!) more expensive than another bidder, something that is usually enough to make someone look more closely.

That closer look led back to the requirements, and the requirements led back to the way they had been written, and eventually to the use of ChatGPT as part of that drafting process. This brought up a full blown investigation.

This is a real life story, which you can read more about, but it is rather a scary one because there is nothing in that sequence of actions that that procurement person took that feels extraordinary these days. It was just a series of small, reasonable steps that collectively shaped a bad outcome.

And I find this is the part of AI in procurement that almost no one is talking about, even though it is already happening.

I am in this AI space a lot and I find it frustrating that still most of the conversation right now are focused on what AI can do, how it can accelerate sourcing cycles, automate evaluation, reduce manual work…. but there is little talk about the dangers.

AI is already influencing decisions without being explicitly recognised as such. It is daily shaping requirements, summarising supplier responses, influencing how options are framed before anyone has even started evaluating them…

Procurement has always relied on structure as a signal of credibility, so when something is clearly written, logically organised and neatly presented, it already carries weight, and AI is exceptionally good at producing exactly that.

Bias, for example, is very hard to see. Particularly when it is grounded in historical data that reflects previous decisions, preferred suppliers or established relationships, all of which are then reinforced rather than questioned.

If you ask an AI tool to generate a supplier shortlist, it will often return names that look familiar, and if you use it to support an evaluation, the ranking will usually feel logical. And this is because it is built on patterns that already exist, which makes it difficult to spot where those patterns might be limiting choice rather than expanding it.

There is a well-known case where Amazon had to abandon its AI recruiting tool because it had learned from historical hiring data and started systematically downgrading CVs from women. The system was just reflecting past decisions while quietly embedding bias.

Hallucinations are even more deceptive, because they tend to present themselves as well-written, plausible outcomes. Easier to accept when time is limited and our work needs to move forward.

There has been many instances of lawyers submitting legal briefs containing citations generated by ChatGPT that turn out to be entirely fabricated, because the output was convincing enough to pass initial review.

In a procurement context, this can look like supplier summaries that fill in gaps with assumptions, market analyses that sound researched but rely on invented detail, or contract explanations that feel coherent while missing critical nuance, and once those outputs are written down, they begin to influence decisions whether or not they are accurate.

I feel like the legal liability behind all of this, on the other hand, has not evolved at the same pace as the technology, which creates a mismatch between how decisions are made and how responsibility is assigned.

A recent case involving an airline chatbot showed that when incorrect information is provided to a customer, the organisation remains accountable for that information, even when it was generated by an AI system, and the same principle applies in procurement.

Where any decision that affects suppliers or outcomes ultimately sits with the organisation, regardless of how much AI was involved in the process.

What this means in practice for procurement is that we are entering a phase where the quality of the output can no longer be used as a reliable proxy for the quality of the thinking behind it. This changes the way decisions need to be approached.

The tools themselves are not the issue, since they can be genuinely useful when applied with care, but the way they integrate into existing processes creates new points of risk that sit alongside the ones procurement has always managed.

There are a few practical changes that I suggest can make a meaningful difference, and none of them require you stopping the use of AI, only becoming more deliberate about how it is applied.

1. Treating AI outputs as first drafts rather than finished work creates space for judgement, particularly in areas that directly influence supplier selection or evaluation, where the cost of being wrong is higher.

2. Separating the act of generating content from the act of evaluating it helps avoid circular thinking, especially when the same tool is used for both, which can create a false sense of validation.

3. Paying closer attention when something feels specially polished or persuasive can act as a useful signal.

4. Making the use of AI visible within the process, rather than allowing it to blend into the background, helps maintain accountability and keeps the focus on decision-making rather than output generation.

The crude reality is that AI is already part of procurement decisions, whether it is acknowledged explicitly or not.

The Bellingham case stands out because it makes that dynamic visible, showing how a series of small, reasonable actions can shape an outcome in ways that only become apparent after the fact, which is exactly the kind of scenario we as procurement professionals are expected to manage, even when the tools involved are new.

Freebie(s) of the week

MY ULTIMATE PROCUREMENT POLICY

You’ve probably seen a hundred procurement policies in your life.

Most of them sit somewhere between “no one reads this” and “this will slow everything down”.

So I wrote the one I actually wish I’d had.

This is my Enterprise Procurement Policy, built over years of writing them from scratch, fixing broken ones, and quietly borrowing the bits that actually worked across different organisations.

It covers the structure, the rules, the grey areas people usually avoid, and the parts that make the difference between a policy that gets ignored and one that actually holds up when it matters.

I don’t make this one free often as it is one of my bestsellers so it is available for the next 24 hours only.

After that, it goes back behind the paywall where it normally lives.

Download My ULTIMATE procurement Policy here.

Do you want access to other great templates from previous newsletters? Have a look at the full store below:

That’s it for this week.

That’s it for this week.

I’m off to Lisbon tomorrow, so if you happen to be there, I’d genuinely love to meet a few of you in person. It still feels slightly surreal how many of you read this every week and yet we’ve never actually crossed paths.

I’ll also be in Madrid at the end of May for the AERCE Congreso, which should be a good one.

Next week, we’re bringing in a new guest writer with a public sector background, which I’m particularly looking forward to. It’s one of those areas that most people in private sector don’t fully understand until they’re in it.

On a completely different note, I’ve watched more Bluey and Peppa Pig over the past two weeks than I care to admit, and there’s a line from Bluey that stuck with me:

It’s said to a six-year-old, but it works just as well when you’re looking at everyone else’s roadmap, everyone else’s transformation, everyone else’s version of what “good” looks like.

Procurement especially has a habit of copying what looks successful somewhere else, without always asking whether it actually fits.

Worth keeping in mind before you sign anything expensive.

See you next week.

Until next time

Procurement worth reading.