In March, CPOstrategy attended Ivalua Now 2025, a gathering of over 800 procurement leaders in Paris. As is the case with virtually all procurement events right now, two topics dominated the discourse: disruption and artificial intelligence (AI). Whether caused by tariffs, market fluctuation, shifting regulations, the skills shortage, or climate-crisis-driven destabilisation in vulnerable regions of the world, disruption is the risk at the top of every CPO’s agenda — according to a recent Gartner report. Ivalua CMO Alec Saric, who opened the event, began by acknowledging the myriad challenges facing the procurement sector in 2025, reflecting that “on top of everything else that’s been going on, it seems like we now have to contend with weekly changes to trade policies.” The increasingly uncertain sourcing and procurement landscape, Saric continued, is forcing the industry “to assess the impact on our organisations, reassess our supply strategies, and it’s all happening so fast.”
Ivalua’s newly appointed CEO, Franck Lheureux, echoed the sentiment. “Procurement has never been [this] critical to your organisations,” he said. “I think about the word, and I’m not shy about using it, is chaos.”
Whether to sidestep said disruption or to unlock new opportunities for procurement to create value, generative AI feels like the hopeful counterpoint to discussions of disruption — a way for organisations to mitigate their most pressing pain points and seek new opportunities to elevate the procurement function beyond its traditionally reactive, functional role. Ivalua itself is throwing itself into a fully-committed exploration of GenAI; its founder and previous CEO, David Khuat-Duy, recently handed off the CEO role to long-time Ivalua exec Lheureux, assuming the newly created role of Chief AI Officer in order to focus on unlocking the technology’s potential.
The AI hype gap in procurement
However, bridging the gap between the things generative AI’s creators claim the technology can (or, very soon, will be able to) do and what it’s actually capable of, is a matter that the whole sector seems to be grappling with. And not always with a great deal of success, despite considerable financial and human capital investment.
Saric acknowledged that “I know many of you have already started using AI to some extent and we can be honest: so far, the impact has hardly been transformational.”
However, he added that, despite use cases being slower to appear than expected, he doesn’t believe that AI is “just another hype technology” like blockchain or the metaverse. Nevertheless, he stressed that “the changes that are taking place, both in terms of their speed and magnitude, are really unbelievable.”
At Ivalua Now 2025, CPOstrategy spoke to procurement sector leaders from some of the biggest organisations working in the discipline, from consultants to software vendors, to explore the degree to which they’re adopting GenAI. And, more importantly, whether it’s living up to the hype.
George Nico, Director, Optis Consulting
To what degree is Optis investing in GenAI, and how is it measuring up to your expectations for the technology?
“Just like most organisations right now, we’re still trying to get a lay of the land in terms of what’s working and what isn’t. I can’t necessarily say exactly how much we’re investing, but we’re starting to break off a part of our business to focus on GenAI, specifically.
“As for how valuable GenAI is proving to us or to our clients? I think I’m going to give you the same answer as a lot of other people: It’s not there yet.
“Sure, it’s very exciting. Some of the most interesting work that we’re doing now is helping us prepare for when GenAI actually is valuable — we’re starting to think about the use cases work with our partners like Ivalua, who are doing good work building their architecture for AI. But, right now, it’s really just experimenting to figure out what works and what doesn’t.”
Are there any areas where you’re seeing GenAI work particularly well?
“Simple things like note-taking, having it record our calls and being able to answer questions based on those conversations.”
So, mostly efficiency stuff at the moment?
“At the moment, yeah.”
The main goal for this technology eventually is for GenAI’s outputs to become trustworthy and valuable, right? What kind of timeline are you seeing for potentially getting to that point? Or do you think there’s a disconnect between the narratives around that timeline and what’s actually happening?
“Yeah, I think people and companies are quite enthusiastic about the timeline between now and then — as they should be. I think the technology is moving quite quickly and, by the time AI is coding itself, it’s going to get even faster.
“Especially in terms of what Ivalua has been doing, we already have some clients that have deployed some valuable use cases for themselves, but as it continues to grow over the next year or so, and especially as Ivalua releases their V 10, I think that’s where it’ll really hit the ground running.”
Fraser Woodhouse, Digital Procurement Lead, Deloitte
How is the value GenAI is creating in procurement right now measuring up against the expectations for that technology? Are we at a stage where things are maybe a little overhyped?
“I think that’s probably incorrect. The hype is real and for good reason. I’d say a lot of procurement’s ability to access the latest GenAI innovations is lacking for a variety of reasons. Sometimes, they don’t have the tech that implements the newest features, or there are too many controls and risks around data security. There are many reasons why people can’t access those latest GenAI features, but when it does come, it is going to completely revolutionise what we do as a discipline.
“Right now, what you’re seeing in most tools on the market are single-point solutions for specific use cases. There are many tools that do one thing very well.
“I think where we’re heading for the moon is the progress we’re seeing towards unscripted, multi-step, AI agent-driven, contextually intelligent toolsets that operate autonomously. That’s coming not just to procurement, but everywhere, and it’s going to change how we think about the entire department.”
How long is that runway in your opinion, between now and the moon?
“Between now and the moon? I think there will always be a “moon,” no matter what it is. As soon as one thing comes out, the next thing becomes the “moon.” If you’d asked me a couple of years ago, I would have thought we’d be further along by now.
“You saw those first-generation use cases flooding the market a couple of years ago, and it’s taken longer than expected to really connect them. Many organisations still aren’t seeing the value. A recent survey showed that almost 90% of CPOs have piloted something related to GenAI, but only roughly 40% are starting to see the value.”
How would you say we close that gap?
“There are a few ways to access GenAI right now. One option is agentic AI, where you can build something yourself. Honestly, I think this is where we’re seeing the biggest return on investment—when organisations have the funds and skill set internally to build something, or when they engage the right partner to co-develop a solution tailored to their specific use cases. That’s one way.
“Another option is buying a specialist tool, a point solution. The challenge there is integrating it within your processes. It may do one thing very well, but it can be clunky if you have to jump between tools or deal with the technical debt from all those activities. Alternatively, you can “sweat” your existing tech stack. But honestly, a lot of the existing suite providers have been slower to innovate compared to some of the newer solutions on the market.”
“So one answer to your question is to wait for that existing tech stack to catch up. However, that’s not appealing to many. If you want the latest and greatest, you’ll need to make some interesting choices. Speak to someone who can help guide you through those decisions, or you’ll be asking the same question next year when the latest GenTech solutions come out.”
So, in part, you’re suggesting that people resist the FOMO?
“No, I don’t think you should resist it. There are many ways to access time-saving features now. For example, Deloitte is shortly releasing a GenAI marketplace, which gives you low-barrier access to generative AI use cases. Some of the tools you hear about have marketplaces where you can pick high-impact use cases without the burden of cumbersome integrations. That’s often what prevents organizations from pursuing these technology journeys. So, there are ways to access the latest now, while also preparing your foundation for when the “moon” comes.”
Oh, that’s really interesting. Can you tell me a little more about the marketplace, maybe providing an example of a more plug-and-play AI solution?
“As I mentioned earlier, there are different ways people can access GenAI. One way is through an enterprise AI tool, like a copilot or something similar to ChatGPT. A lot of what procurement organisations are doing now is copying and pasting data into these tools. It works, but it’s not perfect—it still saves time, but to really get value, people need to learn how to write effective prompts, which can be challenging for many.
“It would be great if you didn’t have to think about prompts in the same way, and if you had a tool with a language model specifically trained for procurement use cases. This would allow you to easily input what you need, get the results, and move on. It would be like a pay-per-use solution—simple, efficient, and targeted for procurement.”
Alec Saric, CMO, Ivalua
Earlier you spoke about GenAI’s potential to take existing data and platforms to offer new recommendations and pull new insights from different bits of data. How close are we to getting users and customers to a point where they trust those recommendations?
“Right now, there’s still absolutely a human in the loop. We’re not at the point where things are completely autonomous without any human involvement. That’s mainly because of trust, and in some cases, the capabilities aren’t fully there for certain processes. But for the most part, it’s a trust issue. Right now, AI is typically used to provide a draft or a recommendation, which can save a lot of time. However, organisations are still not fully trusting those outputs to be used without human review.”
It feels similar to how driverless cars evolved over the last five years. You can get it 90% of the way there, but a human is still needed in the process. Is the ultimate goal to reach a fully autonomous situation, where you can trust AI recommendations and let it “take the wheel,” so to speak? That’s the final goal, right?
“Yeah, that’s an interesting question. From a technology perspective, the aim is to make the AI so reliable that it could be fully trusted. However, we’re not recommending that strategic activities be completely handed off to a robot.
“For really strategic decisions, it will always make sense for humans to be involved. But for more tactical items, like sourcing a one-off purchase from a tail spend category, it can be a time-consuming task for procurement. It doesn’t have a major impact or reputational risk. The worst-case scenario is that an employee doesn’t get the item exactly when they expect it, because the supplier is unreliable.
“For these more transactional and tactical activities, the goal is to remove the human from the process as much as possible, so people can focus on strategy, relationships, and making final decisions for more significant matters. From our perspective, that’s the direction we want to go in: freeing up people to do what they do best—focus on relationships and strategy—while technology handles the rest.”
What was the expectation and hope that AI would be delivering by now, and how far away do you think we are from that stage, based on the current rate of progression?
“I think the gap we’re seeing is the level of use within organisations. There may have been an expectation that, by now, we’d have 40 use cases doing things across the organisation. But it was never expected to be fully autonomous at this point—that has become more of a recent development with generative AI. What was expected, though, was that more people within the business would be using it.
“The reality is that consumers adopt technology much faster than businesses. Businesses are more risk-averse; they have policies, regulations, and the cost of a mistake can be far greater for them than for an individual making a poor decision, like ordering the wrong item.
“What we’re also seeing is that there’s an assumption that you can just take AI “out of the box” and apply it directly to your organisation. That works in some cases, like summarizing a contract—it’s a fairly universal task. But in many cases, it requires some refinement. For example, one organization was pulling a category intelligence report with recommendations for sourcing strategies, and it was about aluminum. The organisation is a steel producer, so they have very specific market indices they use for tracking aluminum prices. The model gave a good recommendation, but it referred to a different third-party source than the one they use.
“These kinds of adjustments are necessary, and what we’ve found is that AI tends to work better when it’s rolled out for specific use cases on a pilot basis. You can evaluate and refine the tool based on those use cases and then roll it out more broadly.
“Of course, you can have general AI tools to find information on the web that can be made available to users, but I think that refinement step is important in many cases. This will be even more crucial as we approach a more autonomous state.
“If we’re going to automate complex, multi-step processes—especially ones where humans aren’t involved—organisations will need to be sure they understand exactly what processes are being followed and how decisions are being made.”
