There's a dangerous narrative running through the AI conversation in procurement right now. It's subtle enough that most people don't notice they've accepted it. The narrative goes roughly like this: if AI can technically execute a task, the human in that role is a placeholder — and removing them is the natural next step.
Senior leaders have embraced software metrics as proxies for transformation success. Process maturity scores. Click-reduction statistics. Touchless transaction rates. These numbers are real and they're worth tracking. But they've become dominant in a way that obscures something important: a frictionless process is not the same thing as a successful business outcome. They're related. They're not equivalent.
The global supply chain is not a mathematical system. It's built on trust, on context, and on relationships that took years to establish and can fracture in an afternoon. An AI agent can monitor delivery dates and issue automated penalty notices. It cannot call a supplier during a geopolitical disruption to negotiate priority allocation based on a ten-year relationship. When supply chains break — and they do — software doesn't save you. People do.
What the efficiency obsession misses
Technology is very good at optimising for speed. It can process thousands of transactions in seconds, surface anomalies in data sets that no human would find in a reasonable timeframe, and execute rule-based workflows without fatigue or inconsistency. These are genuine capabilities and they matter.
What technology does badly — and will continue to do badly for longer than vendor conversations tend to suggest — is navigate the relational and contextual complexity that underlies most significant procurement decisions.
Knowing that a technically compliant recommendation is commercially wrong because of a supplier relationship dynamic that doesn't appear in any data field. Understanding that a stakeholder's verbal agreement in a meeting is not the same as their organisational commitment. Deciding when the model's output should be questioned rather than acted on. These are not edge cases. They're the moments that determine whether a category strategy lands or stalls, whether a supplier relationship survives a difficult period or doesn't.
If your digital transformation strategy treats people as cost centres to be minimised through automation, you are building an organisation that will perform exceptionally well under normal conditions and fracture the moment something genuinely unexpected happens. Market disruptions are not scheduled.
The First and Second Curve
The distinction I keep coming back to is what I think of as the two curves of procurement value.
The First Curve is the traditional domain of transactional procurement. Data entry, purchase order processing, basic compliance, invoice matching, routine supplier onboarding. It's repetitive, rule-based work that consumes enormous team capacity and adds relatively limited strategic value once it's running correctly. This is the domain where AI automation belongs — and where it performs reliably. Strip the First Curve bare. Automate it aggressively. That's not capitulation; it's resource allocation.
The Second Curve is where the value actually lives. Complex relationship management. Geopolitical risk evaluation. Category strategy that accounts for commercial, political, and organisational constraints simultaneously. The ability to read a situation that doesn't appear in any data field and make a call. These are not tasks that get automated. As AI handles more of the First Curve, they become more — not less — important.
The mismanagement I see most often isn't teams that have automated too much. It's teams where talented people are still spending the majority of their time on First Curve work — not because the automation isn't available, but because nobody made a deliberate decision to move them off it. People who could be operating on the Second Curve are being used as expensive data processors. That's a capability waste problem dressed up as a technology question.
What the transition actually requires
Deciding that your team should operate on the Second Curve is the easy part. Getting them there requires something more deliberate than removing transactional work from their plates and hoping they fill the space strategically.
The transition is a leadership project. It requires investing in commercial context — helping people understand not just what the business is buying, but why, and what the organisation is trying to achieve. It requires building real stakeholder relationships rather than stakeholder management processes. And it requires developing the kind of critical relationship with data that lets someone look at an AI output and know when to trust it versus when to push back.
That last capability matters more than most transformation conversations acknowledge. AI systems make confident errors. They misclassify spend, misread supplier risk, and apply rules to situations those rules weren't designed for. The human who can catch that — who understands the context well enough to recognise when the output doesn't smell right — is not a redundancy. They're the quality layer that makes the whole system trustworthy.
Teach your teams to treat AI outputs as hypotheses, not verdicts. That's not scepticism about the technology. It's how the technology actually works best.
The supply chain reality
The argument for human primacy in procurement isn't sentimental. It's grounded in how supply chains behave under stress.
During the disruptions of the last few years — whether pandemic-era shortages, geopolitical supply shocks, or energy market volatility — the organisations that maintained supply continuity were, almost without exception, the ones with deep supplier relationships. Not the ones with the most sophisticated monitoring dashboards. The ones where a procurement professional could pick up the phone and have a conversation that an automated system couldn't initiate.
That relationship infrastructure is built over years of consistent, honest engagement. It doesn't appear in your supplier master data. It doesn't show up in a risk score. And it cannot be stood up in a crisis from scratch — only drawn on if it was built before the crisis arrived.
Invest heavily in your digital architecture. The efficiency gains are real and the competitive pressure is genuine. But bet your future on the people who understand your supply base, your stakeholders, and your organisation well enough to navigate what the system can't see coming.
When it matters most, that's the capability that won't let you down.
