The shift underway in procurement is not about adding another AI tool to the stack. It is about a function moving from using AI to being powered by it. That was the takeaway procurement writer Tom Mills brought back from a day at Amazon Business Exchange in London, and the framing he offered has proved unusually durable: not AI replacing procurement teams, but AI taking the weight off them so they can focus on the parts of the job that actually need a person.
Five ideas anchored the argument. AI does not remove the need for great teams, it raises the bar, because two teams given identical technology still produce different outcomes. AI is starting to act rather than answer, taking the next step instead of responding to a question. The boring foundations matter more than ever: data, then integration, then optimisation, in that order, and you do not get to skip the first part. Procurement’s biggest AI opportunity might be time itself. And the ambition should stretch past efficiency toward something more commercial, more strategic, and more human where being human matters.
The question posed to readers was which of the five matters most. The answers split cleanly into two camps, and the more interesting argument happened inside the winning one.
Time Won. Then Came the Caveat.
The time argument drew the most support, and the most immediate qualification.
Strahinja Jovanovic, who builds supply chain operations in eCommerce, put the caveat first. “Freeing up time does not automatically create better procurement. Teams also need to be deliberate about where that time goes. Otherwise the administrative work disappears and gets replaced by a different set of low value activities.”
That is the failure mode nobody plans for. Automation removes a category of work and the vacuum fills itself, usually with meetings, reporting, and a different flavour of administration. The capacity gain is real and entirely invisible on any measure that matters.
Mark Strange, a transformation leader, took the point further and landed on the sharpest reframing in the discussion. “The opportunity is not really the time AI gives back. It’s what we choose to do with that time. If we automate transactional work and simply fill the space with more activity, we have improved efficiency but changed very little.”
His conclusion inverted the original question. “Perhaps the question is not just what could procurement become? It’s what do we want our people to become better at? Because AI can create the capacity. Leadership still has to decide what that capacity is for.”
That distinction matters operationally. Capacity is a technology outcome. Direction is a management decision, and the second does not follow automatically from the first.
Pierre Mitchell, a longtime procurement technology analyst, grounded the argument in economics. “Economic cost equals opportunity cost. The biggest cost in procurement is the opportunity cost of resources, in a function that delivers 5x to 10x ROI, wasting time on low impact tasks versus investment of their time into high impact activities that build way greater capabilities.”
Framing recovered hours as an investment decision rather than a productivity gain changes how they get allocated. A CFO would never let a 5x-return function reinvest capital without a thesis. Time gets treated with far less rigour.
Hazel Tinsley, a procurement and third-party risk leader, named where it should go. “More time for judgement, relationships and understanding the wider enterprise, rather than simply doing the same things faster.”
The Unglamorous Prerequisite
The second camp argued that none of this happens without the foundations, and their case was harder to romanticise.
Georgia Giannia, a senior IT procurement professional, delivered the line that best captures the problem. “Everyone wants the clever use case, nobody wants to reconcile the vendor master, but you cannot optimise spend you cannot see, and that is where most of the value has always been hiding.”
That last clause deserves weight. The value was never trapped behind a missing AI capability. It was trapped behind a vendor master nobody wanted to clean. AI does not unlock it. It simply makes the failure to unlock it more expensive.
Frederick Magana, a CIPS Fellow, described what the work actually involves. “AI adoption requires a lot of heavy lifting with data cleansing, approval workflow alignments, procurement process mapping. Unfortunately people usually see the end product.”
Nuha Luqman, working in energy supply chain, stated the dependency plainly. “Without clean data and integrated systems, the promise of AI collapses. Get the basics right, and the opportunity becomes transformational.”
The two camps are not actually in conflict. The time argument describes the destination. The foundations argument describes the toll booth. Functions that skip the second and campaign for the first end up with expensive automation running on data nobody trusts.
AI as a Governed Teammate
A separate thread addressed what happens when AI stops answering and starts acting.
Esther Adorable Emoedume, who builds procurement AI, drew the line most precisely. “AI should become a governed teammate, not an unsupervised decision maker. Procurement professionals should retain accountability. While AI does the information work, the human makes the consequential decision.”
That formulation, information work versus consequential decision, is a cleaner operating boundary than most AI governance frameworks manage in several pages. It also implies something uncomfortable: the boundary only holds if someone maintains it deliberately. Agentic systems drift toward taking more steps because that is what they are built to do.
Rebecca May Shah, a commercial lawyer working in AI governance, confirmed the pattern outside procurement. “These closely align with the changes in the Legal world. AI will help us, and act as a starting point, but it cannot replace human judgment or relationships.”
Hisham Serry, a supply chain consultant, noted a measurement practice worth copying. “Some companies have started to track the time saved by Copilot Cowork and are beginning to utilize that additional time for other more procurement thinking activities.”
Tracking recovered hours is the only way to know whether capacity is being redeployed or quietly reabsorbed. Very few functions measure it at all.
The Widening Gap
The most consequential prediction came from Rebecca Bellairs, co-founder of a marketing pitch platform. “The already strong teams are going to catapult into previously unknown performance levels with the appropriate use of AI in next few years. And the cost and cash obsessed tactical set ups are going to really struggle. The tools are just applying a magnifying glass and multiplication to what you’re already doing.”
Her warning about skill investment cut against the prevailing training rush. “Potentially too many people focusing only on AI fluency and not on upskilling in their own subject matter expertise. Both are essential.”
Abhishk Sinha, a procurement leader formerly at Maersk, compressed the same idea into a sentence. “AI will expose which procurement teams were strategic all along.”
That is the uncomfortable corollary of the raises-the-bar argument. Identical technology producing different outcomes is not just an optimistic note about culture. It means the distribution widens, and functions currently mistaking activity for strategy will find that harder to hide.
Takeaways for Procurement Leaders
Three lessons run through the discussion. First, decide what the capacity is for before you create it. Automation without a redeployment thesis produces efficiency and changes nothing. Leadership owns that decision, not the technology.
Second, fix the vendor master before the use case. You cannot optimise spend you cannot see, and the value has always been hiding behind exactly the work nobody volunteers for.
Third, track the hours you recover. If nobody measures where saved time goes, it does not go anywhere useful. It gets absorbed.
If AI gave your team back a day a week starting Monday, do you know what you would have them do with it?
Continue the discussion on Chain.NET, the global supply chain community, at www.chain.net. Ask questions, join events, and access exclusive resources.