Procurement spent two decades buying platforms to simplify itself. Now AI has arrived and much of the industry is doing the same thing again, adding copilots, prompts and agents on top of processes everyone already knows are cumbersome. That is useful automation. Whether it amounts to transformation is a different matter.
Dr Salisu Uba, founder of NatQuest and a commercial and procurement transformation leader working across EMEA, argued the profession is asking the wrong question entirely. Not “how can AI make procurement faster,” but “if we designed procurement today, with AI available from the beginning, what would we stop doing altogether?”
The examples he raised are uncomfortable because none of them is hypothetical. Would we still produce annual category strategies when market intelligence can be continuous? Would sourcing always begin with a requisition when demand can increasingly be anticipated? Would professionals spend hours comparing documents when machines do it in seconds? Would we buy separate tools for sourcing, contracts, suppliers and risk if intelligence connected those decisions?
His framing of the work ahead was three verbs. Automate what should remain. Remove what should not. Reinvent what procurement can become. The next significant procurement company, he suggested, may not build a better procurement platform at all. It may build a different way of making commercial decisions.
The responses split between those who found the question liberating and one who explained, in detail, why the market will not answer it.
Why Vendors Will Keep Selling Task Agents
The most substantive challenge came from Kris Kosmala, who works on digital and AI transformation in shipping operations, and it is worth reading closely because it explains the behaviour rather than just criticising it.
“Software companies wanting or breaking into this information technology space will be pitching task oriented AI agents, because they see it as the fastest way to monetize their development effort. They also know those agents have no moats to speak of. They are not inherently different from each other, their value is comparable to competitors, and the solution is easily copiable. So cashing in as rapidly as possible is the only way to make return for the original investment.”
That is the structural reason the reinvention conversation struggles to start. A vendor who redesigns the commercial decision model has a long sales cycle, a hard integration story and an uncertain return. A vendor who ships a contract-comparison agent can sell it next quarter.
His point about workflow risk is equally practical. “The legacy solutions are designed from scratch for that end-to-end followed by another end-to-end. Rip and replace of one workflow destabilizes the well honed process.”
Then he named the part the industry avoids. Buyers will save money on these projects, he argued, but not by freeing people for higher-value work, “the stand-in AI fantasy pitch,” but by freeing those people completely.
His follow-up went further, and it challenges the premise of transformation itself. “Thanks to ERP we track the economics of procurement well. Those can only change if somehow we rethink why a business needs to procure anything. And that’s rather not on the horizon. In fact, the AI agents cement and reinforce the procurement economics as we know them today. What they allow is to think that suddenly a company like Gap can go from 500 people in sourcing to 50. And maybe they do, and they will definitely change the economics of this space, but not the business model execution.”
That distinction deserves to sit in every transformation deck. Changing the economics of a function and changing what the function does are not the same achievement, and the first one is far easier to sell as the second.
The Integration Tax
Edward Buntoro, a technology product owner with experience across SAP Ariba, S/4HANA, Anaplan and ServiceNow, located the problem one layer below the AI debate.
“The fact that organizations have fragmented technologies and do more on technology alignment instead of maximizing the technology functions, mostly driven by cost benefit and project owner. We should see it more on the integration and value instead of cost, and NOT all processes need to be automated.”
His example of what should stay human is well chosen. “Vendor evaluation is required more judgement despite of technology input.”
That is a useful boundary. Document comparison is mechanical and should go. Deciding which supplier your organisation can actually work with carries commercial, technical and ethical weight that does not reduce to a score.
Uba agreed and named the trap. “Technology fragmentation often means organisations spend more time connecting systems than improving commercial outcomes. Integration matters, but it must serve a clear purpose beyond cost savings.”
He also drew the line on what AI should be doing. “AI should support professional judgement, not replace it, particularly where commercial, technical and ethical considerations are involved. The objective should be better commercial decisions, not automation for its own sake.”
Faster Is the Wrong Benchmark
Chinmay Agarwal, co-founder and CEO of Negotiations.Ai and formerly at Kearney, pushed back on how AI tools are being positioned in his own category.
“It is not about just being faster, it is also about being smarter and more informed going into a negotiation.”
That is a small correction with large implications. Speed is the metric vendors can demonstrate in a demo. Preparation quality is the metric that changes the outcome, and it is almost impossible to show in a fifteen-minute sales call. The industry defaults to the measurable one.
Umar Shaikh, working across facilities and vendor management, drew the same conclusion about process. “AI shouldn’t just make existing procurement processes faster. It should challenge whether some of those processes are still necessary. The real value comes from redesigning the decision-making process, not simply automating the old one.”
Fabio Marcucci, founder of a B2B trade infrastructure business, isolated the sentence that does the work. “The question about what we would stop doing is the most interesting part here. AI shouldn’t just make existing procurement processes faster if some of those processes don’t need to exist in the first place.”
The Question Procurement Should Actually Run
There is an exercise buried in this discussion that any procurement leader could run without buying anything. Take your current process map. Mark each step as one of three things: this creates commercial value, this exists to feed a system, or this exists because somebody once asked for it.
The third category is usually larger than anyone expects, and no amount of AI applied to it produces anything but faster waste. The second category is where integration spend goes to die. Only the first deserves a copilot.
Uba’s reply to Shaikh set the ambition. “If AI can sense demand earlier, continuously read markets, predict risk and support decisions in real time, then we should be questioning why some procurement processes exist in their current form at all.”
Takeaways for Procurement Leaders
Three lessons run through the discussion. First, run the stop-doing audit before the automation business case. Annual category strategies, requisition-triggered sourcing and manual document comparison are candidates for removal, not acceleration.
Second, be honest about where the savings come from. If the ROI model rests on headcount, say so internally rather than selling the board on freed-up strategic capacity that never materialises.
Third, protect judgement explicitly. Vendor evaluation, ethical screening and relationship calls carry weight that does not reduce to a score, and the tools being sold today are not built to carry it.
Which procurement process in your organisation would you not create if you were starting today with AI already in the room?
GSCC members access events, replays, reports, learning tracks, AI tools and the prompt library. See the full benefits at https://gscc.co/benefits and the event calendar at https://www.chain.net/c/events/.
Continue the discussion on Chain.NET, the global supply chain community, at https://www.chain.net/c/ask.