The Pull #010 - The AI Mirage in Procurement

Everyone Wants AI. No One Wants to Do the Dirty Work.

If I had a penny for every article telling us how beneficial AI is for procurement this year alone, I’d be buying myself a new car.

There’s a certain high that comes with hearing phrases like “predictive sourcing,” “negotiation bots,” and “smart analytics.”

We all want it — dear God, we want it more than anything.

We dream of the day we can feed our spend cubes into a shiny model that tells us exactly where we’re wasting money and where our assumptions are wrong.

But no one talks about the people behind this tremendous effort.

How many identical entities does your supplier have in the system?
You know the ones — Thassos, T.h.a.s.s.o.s, Thassos Inc— all pointing to the same registered company.

How many identical products are sitting in your catalog, spelled three different ways?

How well is your ERP actually integrated with the risk tool, the vendor overview, the market intel platform?

And who’s doing the work to connect them all?

Let’s be real:

Companies want to look AI-ready, but they don’t want to invest in the unglamorous part — the human effort to make the data usable.

Procurement is already drowning: renewals piling up, contracts half-reviewed, Finance demanding forecasts, IT changing cost centers again.

And then someone says, “Leverage GenAI.”

With what, exactly?
The same headcount, the same bad taxonomy, and a dozen data sources that can’t even agree on supplier IDs?

AI doesn’t shorten the road.
It just makes the finish line look closer while you’re still running in mud.

In theory, it takes “a year to prep the data.”

McKinsey’s 2024 report even said it: companies see a 20–40% increase in validation workload during the first year of AI integration — efficiency comes later, once governance matures.

In practice? That year resets every quarter.
Business units rename things.

New tools appear.

The AI starts hallucinating patterns that don’t exist.

“Did you know you spent €1.2M on office bananas last year?”
No, it was laptops. Thank you for your insight, Skynet.

Yes, AI can classify spend, spot anomalies, or flag supplier risk in seconds.

But validation — the human confirmation — now takes longer than ever.

You have to verify the logic, audit the results, explain why Supplier A was flagged as high-risk, and in regulated industries you have to document every decision like it’s a potential court case.

So yes, the analysis is instant. The governance? Eternal.

AI demands cleaner, structured, labeled data — something legacy procurement systems simply don’t have.

A major company tried to deploy AI-driven spend analysis last year.
They discovered 40% of supplier names were duplicates or misspelled: IBM Global Services, I.B.M., I.B.M. Romania SRL...
Before the model could even run, the team had to rebuild the taxonomy — it delayed the rollout by three months.
The AI could analyze millions of lines instantly.

But humans had to make those lines usable first.

Many teams now follow a “human-in-the-loop” rule: no AI output is accepted without traceable rationale.

That means logging the AI’s reasoning path, verifying data sources, and writing summaries for auditors.

It’s governance on steroids — and it slows everything down, even when the model’s speed is dazzling.

And yet leadership still wants dashboards.

So we push through — knowing results are “directionally right” and “good enough for PowerPoint.”

The Bottom Line

If there’s one thing I’ve learned, it’s this:

AI in procurement doesn’t fail because of tech.

It fails because of under-resourced humans expected to fix decades of data chaos while still delivering savings.

Maybe the real transformation isn’t “AI adoption.”

Maybe it’s giving procurement teams the time and support to make AI actually work.

Stay rested, stay nerdy.

Until next time,

Zvi

From the Feed

Worth your time this season:

#1 – Savings at award can be meaningless: Pharma Edition 

Competitive bids aren’t always the cure.

You can celebrate savings on day one, only to watch them evaporate six months later if you don’t lock down post-award costs and governance.

How do you stop savings from being clawed back after the ink is dry? 

Check out the full post here.

 #2 – Quarterly Business Reviews 

A QBR without business stakeholders is just vendor marketing theater.

The only way to get them to show up is to anchor the session in their pain points, not just procurement’s trackers.

How do you get your business stakeholders to actually show up at QBRs? 

Check out the full post here.

#3 – The myth of “plenty of time”

That “we’ll deal with it later” mindset kills renewals.

Without a clear renewal calendar, procurement gets dragged in only once the fire is already burning. That’s why I built the Renewal Tracker.

Download it here.

ANCHOR TOOLS