Procurement by Design

THE TRUST CONSTRAIN

“If it feels like things are moving fast, brace yourself. This is the slowest things will ever move ever again.”

That’s what my old Jonathan Ross said, currently Chief Software Architect at NVIDIA and Groq founder.

I met Jonathan back when we were both in our early 20s and he was still at NYU. Back then I already knew he had a brilliant mind and was obsessed with technology. Soon after that, he landed his first role at Google, where I got to visit him a few times and the rest is history. I admire the incredible career he has built and I always knew he was meant for greatness. So I will take his comments at face value.

From his perspective, the story is physical and measurable and in terms of compute capability, he’s 100% right. Chips are becoming more specialised for AI workloads. Inference is getting faster and more efficient (inference is where the commercial value actually resides). If it’s cheaper and quicker to run models at scale, businesses use them more. When businesses use them more, infrastructure scales to meet demand. That’s how the AI acceleration compounds.

But I don’t experience AI as a “curve.”

In our procurement world, acceleration means we are being asked to embed systems that consume real resources, cost real money, and influence real decisions before we have even had the time to look at reliability, governance, or measurable economic value.

There are four things from this whole AI wave that genuinely affect us.

1. Cost doesn’t stay flat.

AI doesn’t work like a SaaS licence. It’s compute, which scales with usage.

If capability accelerates, adoption accelerates. And that means spend volatility increases.

We shouldn’t think of this as “buying 200 seats and knowing the cost for three years”. It is a more like buying cloud “in the olden days” where the more people use it, the more it costs.

2. Reliability is still probabilistic

I’ve seen various models hallucinate confidently as late as last week (and even this morning!). I still see AI create humans with 4 legs and 6 fingers.

As the technology gets better and cheaper, there’s more pressure to automate more things with it. The problem is the tools are still not fully reliable. They can be brilliant one minute and confidently wrong the next. So we’re being pushed to automate faster than the tools are ready to be trusted without oversight.

And that decision sits within procurement.

3. Environmental impact is not abstract

AI workloads run in data centres that consume significant electricity and, in many regions, water for cooling. Leaders are now publicly defending energy use because the scrutiny is rising.

At the India AI Impact Summit, Sam Altman defended AI’s energy usage by comparing it to the food humans consume over decades of life, arguing that “training a human” also requires enormous energy input. The reaction online was probably not what he expected.

So this is as much of an ethics debate as it is a commercial one.

If we embed large-scale AI usage without understanding its footprint, there are huge moral and governance consequences.

Procurement already evaluates Scope 3 emissions for suppliers in other categories. So why would AI infrastructure be exempt?

4. The economic proof is still thin

A recent National Bureau of Economic Research survey found that although around 70% of firms report using AI, more than 80% say it has had no measurable impact on productivity or employment so far. Goldman Sachs’ chief economist has also said AI investment contributed “basically zero” to US GDP growth last year.

Maybe that will change, we do not yet know.

But right now there’s a gap between the narrative and the measurable macro impact.

And when budgets tighten, procurement will be asked whether we the AI that was implemented funded durable capability with a proven ROI.

Public trust matters here too, but not in a philosophical way.

• If employees distrust tools, adoption drops.

• If regulators distrust vendors, compliance tightens

• If vendors soften safety commitments under competitive pressure, that’s a vendor risk signal

We’ve already seen high-profile AI companies quietly adjust mission statements and responsible de-scaling safety language over the past year. That may be commercially rational but to us in procurement it is very relevant from a risk perspective.

So Jonathan’s statement is about speed in his world but our job is about control under that speed.

If this is the slowest it will ever feel, then we have to get sharper now!

Here’s what that I humbly believe it actually means for us in practice:

• Model AI as expanding infrastructure spend. Assume usage growth. Stress-test cost scenarios before signing multi-year commitments.

• Decide which tasks can tolerate errors and which absolutely cannot. If the output has low business impact, like early research or rough analysis, you can live with the occasional wrong answer. But if the output of the tool you are looking at affects money, legal terms, or regulatory exposure, you can’t.

• Include AI within your existing cloud and Scope 3 reporting lens. If you’re already tracking emissions from digital services, don’t treat AI as exempt. Large-scale AI adoption increases infrastructure usage. Ask where workloads run, whether region choice is flexible, and ensure sustainability reporting links back to the underlying cloud provider.

• Treat policy and mission changes as early warning signs. When a vendor changes safety language or restructures governance, it can mean their risk appetite is evolving. If they move faster than their safeguards, you inherit that exposure once their system is embedded in yours.

• Tie expansion to measurable impact. No scaling without evidence of real productivity or risk reduction in your own environment.

Uncontrolled acceleration can be avoided. And in procurement we are one of the few functions positioned to make this happen.

Freebie(s) of the week

THE PROCUREMENT INFLUENCE PLAYBOOK

When I stepped into my first head of indirect procurement role, I wasn’t prepared for how different the influence dynamic was going feel.

In theory, nothing dramatic had changed from when I was a simple consultant at Accenture. I was still negotiating, still analysing spend, still accountable for commercial outcomes. In practice, the budgets sat firmly with Marketing and IT. They owned the strategy and the suppliers so I was just there to “support”.

I remember sitting in a meeting with a Marketing Director where the agency had already presented creative concepts before we had even reviewed the commercial structure. The relationship was what in sales jargon you would consider “warm”. The campaign was really exciting so by the time pricing came into focus, the emotional commitment from the Director’s side was already there. Walking in with process language would have made me the villain.

IT brought a different kind of pressure. Technical teams would spend weeks evaluating different platforms, then loop me in once their preferred (often the sole!) vendor was clear. At that stage, I wasn’t shaping their decision at all but I could only strengthen it without undermining their expertise. If I pushed too hard on validation, I risked looking obstructive.

What I struggled with most was how to frame my position in those scenarios. How to protect margin without sounding territorial. How to introduce challenge without triggering resistance. How to influence when authority wasn’t explicit.

That learning curve is the reason this playbook exists. It captures the conversations that I have noted down throughout the years that have helped me shape commercial outcomes and the language that made a difference for me over time.

If you’re working in stakeholder-heavy environments where you influence more than you control, I think you may find it useful.

You can download the free PDF here (available for 72 hours).

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’ll leave you with a thought that feels increasingly relevant the faster this space moves:

AI capability may be accelerating exactly as predicted.

The question is whether our governance, judgement and discipline are accelerating with it.

Because technology scaling is impressive but maturity scaling is harder.

See you next week.

Until next time

Procurement worth reading.