880: The State of AI in Procurement: Who's Getting the ROI? W/ Nick Heinzmann

0:05Hi there, my name is Philip Ideson. I want to welcome you to the Art of Procurement podcast, the podcast that helps you, a forward-thinking procurement professional, position yourself and your team to proactively take advantage of the revolution that's taking place in procurement today. By interviewing industry trailblazers and sharing insights from our own experiences, my team and I pull back the curtain and shine a light on the strategies, tactics, and tools that procurement teams are using to elevate their impact. And today on the show, I'm delighted to welcome Nick Hinesman. Nick is a friend of the show.

0:39He's been on the podcast a number of times. He is head of research at Zip and is here today to talk about all the latest data and the findings in Zip's 2026 state of spend report. It's a great report. It goes into a lot of detail around what Zip is seeing in the marketplace, what people are telling Zip that they're seeing in the marketplace based on the surveys. And so I'm looking forward to getting into some of the bigger trends. But first and foremost, Nick, welcome back to the podcast. Yeah, thanks for having me, Phil. Always excited to talk with you about what's going on in procurement and also share some new and updated data today.

1:14Yeah, and I'm sure that because things are moving so fast, there's going to be a lot of interesting tidbits that we go through in the next 25, 30 minutes or so. And I wanted just to tee things up at the start by just asking if you could just give a high level picture of how you updated the research from last year. Last year was the first time that you did this research. And if there's anything that you kind of looked at and wanted to test the hypothesis around that was new over last year. Sure, sure. So we've been doing this report. This is our second annual version of it called the state of spend.

1:46And this year we want to look at the state of AI and spend specifically. Okay. So this year, we really wanted to continue a theme that we looked at in the first one, which was the bridge years. We saw a lot of people focusing on how do I get from a pre-AI state to one that is more AI native or redesigning their operating model for this new technology. And really what we wanted to focus on in 2026 is the state of AI for that. And so the report specifically wanted to narrow in on things like who is getting ROI and who's behind on that, what changes people are making to the way they organize their functions, the types of skills they're hiring for today.

2:23And then probably most important, how people are planning ahead for this rapidly shifting world. Because that was the main thing people saw last year was that they had accepted as almost inevitable, a huge wave of change coming for them in their function and across the business. But trying to figure out how exactly to, well, we said, build that bridge to get to the next part. And what we were looking at this year specifically as well, who has built that bridge? Maybe who's even crossed it? And then maybe who's also still stuck on the other side? Yeah. Okay. Interesting. So let's go into just a couple of data points that go through the survey first, but I think act as good jumping off points into a broader discussion.

3:01And first of all, there's a pretty provocative claim in the report that procurement has about 12 to 18 months to establish itself as a leader in enterprise AI adoption. I want to know from your perspective, First of all, what does it really mean to become a leader of enterprise AI adoption? And then why that window is such a short period of time that they believe that procurement has? When it comes down to leadership, what we're looking at is not just how procurement works with AI in its role or its function, but actually how does it lead the governance of AI across the enterprise? Procurement has this unique opportunity or almost imperative, really, to be governing AI through the vendor selection process, through risk reviews, through continuous performance monitoring, especially as tokens and costs become a huge issue in the enterprise.

3:51So procurement has this unique vantage point where it needs to be actually helping guide the adoption of AI throughout the enterprise. Something we asked about last year, but continued to see this year. And specifically with the window, there are challenges to that role in the enterprise. We see a lot of enterprises looking at IT, for example, as maybe the one that should be doing that. There's a lot of involvement from risk and compliance functions are legal. Finance, to be sure, is looking at the token bills. You know, the operations function is looking their COOs and saying, wow, if this continues out in a certain way, this is not going to get us the ROI to expect just purely based on the cost.

4:29And so procurement has the opportunity to be the nexus of all these different things happening. but there are also different and competing views on how this should be done or who should get that role so unless procurement itself is changing and becoming more ai forward and pushing adoption within its own function or across these processes that affect all these different functions it risks seeding that role to someone else in the business and arguably the window has to be 12 to 18 months because the world is continuing to change even faster. I mean, I'll just say from my own perspective, but a lot of the survey respondents are our own customers that we see here at Zip.

5:08The way that people were working six months ago before the rise of, you know, call it Cowork, call it ChatGPT Enterprise or any of these newer tools and harnesses, it's completely different. And a lot of these survey respondents even hinted at this in the way they responded is that they expect to be working differently even a year from now simply because of the pace of change. So as that window for change continues to adapt, procurement itself needs to be very much on top of keeping pace with that or it will only fall further and further behind. Yeah, I think when you look at how the step changing frontier models, you know, around the end of last year, that drove kind of that next, the next iteration of adoption, let's say, and And with every single frontier model update that comes, there's nothing to say the next one isn't going to be as transformational.

6:04And then everything changes again. And so often I don't like to put dates and things, and I feel like it's hyperbole. But how I think about it is you don't want to be slow and then find out you're getting left behind. And what's the cost of inaction? and it's better to be moving quickly and risk, you know, and there may be some risk associated with that than just wait and wait and wait and then you may have already missed the window. I mean, I'm interested in whether you think that speed becomes existential or can become existential for procurement or is that going a little bit too far into the, you know, the five alarms trying to make people to give people fears that are unfounded?

6:52I don't want to raise too many alarms. And to say that, you know, in a black and white way, procurement is going to be existing or not existing in a certain number of years because of AI, I think that'd be a little outlandish, to be honest. But I think there's an absolute risk of procurement's role and scope being diminished by AI, unless we embrace new ways of working. And there's lots of interesting evidence about the companies that do the best with AI in terms of adoption, in terms of scale and depth. And one of the most interesting things I found in the survey was that functions that are more often getting ROI out of AI and who are doing deeper and broader deployments are also democratizing the function.

7:39They're pushing their controls and access and starting points to the stakeholders to try to embed themselves and in a word more decentralized the function and i know that can be a controversial word and controversial term it's not about relieving or releasing control it's actually about extending those controls to surface areas where you don't have good visibility or good access today it's about a mindset shift when that's the big thing i always try to emphasize and this is that the way that some of these companies who are doing more faster getting better results, the way they think is actually quite different from what I would term like a laggard, or I guess in this report, the bystander group.

8:17And that mindset shift is a big part of how you address the opportunity in the next 12 to 18 months. Yeah, I think that's the next evolution of like the next natural evolution of the procurement operating model, because we've gone through decentralized, and it kind of brings a circle all the way around, but we've gone decentralized, decentralized, to center-led. It's now moving back to decentralized, but with all these controls in place. And so it has some of the characteristics of a centralized or a center-led model, but it's feeling, as you said, there's more kind of collaboration with the business units as opposed to a power and control, which is often what happens with a centralized, as you go from decentralized to centralized to perhaps center-led in the first place.

9:01And knowing the data, it was cited that only 17 % of organizations from respondents actually said they could demonstrate a clear ROI from AI. You talked about those that are deploying AI broadly were six times more likely to achieve measurable returns. Why is the depth of deployment, do you feel, or maybe it's even in the data, becomes such a strong predictor of whether an organization is going to get ROI from AI or not? In a word, it's about behavior. Depth actually changes the behavior. And it's not really about tooling or even like the skill set. But again, that mindset shift. So it's getting the ROI, which we call the builders.

9:4484 % of those people use AI multiple times a day versus just 40 % of the laggard group, which we call the bystanders. And so this immersion with and use of AI starts to educate you on what works, what doesn't work and everything in between. And so when you increase the intensity at which you use or let AI start working with you in the process or increasingly even sometimes own parts or in advanced cases, the whole process, you start to see actually correlations around how they answer other questions like cutting roles. And so the builders in this group that gets ROI again are most likely to be cutting operational and tactical roles.

10:25But they're also the most likely to be hiring for different types of roles, which are more around what we'd call a procurement engineer to use kind of the market parlance. These guys doing AI rollouts, redesigning processes, understanding how to do the AI evaluations on which models to be fit for a certain task. And they're hiring actually for these roles. So it's not just a cut and set it and forget it. No, it's actually about reshaping the function around what are skills and processes we need to make AI successful. So the depth changes the result because it requires actually refactoring how you work.

11:01And if you don't refactor how you work, you can't get the ROI. But if you think you're just going to put a tool in there, it's going to start adding some automation, you're going to have negligible results. It feels because there's other research that I've seen and talked about on the show around how you use AI and what that means to jobs. And it sounds like what you're seeing from respondents in procurement very much matches what I've seen outside of procurement, which is, you know, you look at using AI as a tool for efficiency, which is really just a process play. That is what leads to probably the biggest risk to jobs.

11:38When you think about AI and being an efficiency plus and when the plus is, how can we do things we could never do before? How can we support organizational growth? That's actually where headcount is growing. And so those organizations, it ties to you deploy it deeper under more use cases that are both efficiency led, but also value driven. That actually leads you to come out the other side with something that is more valuable and is actually hiring more than if you just focus on it as being an efficiency play. Exactly. And we see this so many times in our own deployments with some of the more advanced AI organizations out there.

12:16is that there's not a hesitancy to apply AI, but there's a targeted approach for exactly what you just said. Some areas where there was no work being done simply because the volume of it was too high. Tailspin sourcing is a great example. Or there's a shift in focus where there are some people who might've been manually collecting the request intake process or doing the validation on those types of requests. A lot of that can be partially, maybe not fully because you want a human in the loop for some of those reviews, but partially automated. And you can take a lot of the work that was happening in procurement operations for interfacing with stakeholders and guiding people through an approval process, take that off their plate and not eliminate that person, but actually put them somewhere else where they can add more value and probably, to be honest, where they'd probably want to spend time anyway.

13:06A big thing we saw in the expectations last year and continued through this year was an expectation that supplier performance management, an often relatively ignored area in terms of both technology and process, would actually get an uplift in the amount of time spent by procurement teams. And now instead of looking at performance management as this nice to have post-contract, it's becoming a place where people want to spend time and have to be a core competency in developing and improving the relationship. Yeah. And it'd be more that it's not just checking some boxes and looking at contracts and attending a QBR.

13:39is generally about how can we maximize the value of the relationship that we have in place, which changes again from it being a box checking exercise to more of a strategic role. Exactly. And I think a lot of people even, they try to bucket these different types of activities as tactical and strategic and which ones should have AI and not AI. But even with supplier performance management, there's exactly what you said. It's not just a box ticking exercise for the contract, but there's information in the contract that's necessary. There's information in your purchase orders, your invoices, maybe stakeholder reviews.

14:15And those all need to be regularly gathered, analyzed, and then potentially, you know, a corrective action put on those. AI is really great at servicing and helping you find that information so you can communicate effectively as a supplier. But you should be doing the communication because a corrective action delivered via email by a bot is very easy to ignore. Yeah. I think one of the challenges is when I speak to procurement leaders at the moment is how to balance across their team, encouraging the usage of AI and encouraging, you know, growing AI literacy from a talent perspective. And therefore it's, it's using AI, using AI in your job versus, you know, to centralize governance and the use of approved tools.

14:58So I think in the data you had more than half of employees using what were classified as unsanctioned AI tools. How much of a problem are you seeing this becoming? And you want them to have access for learning, but you don't want them to be going off and doing things that are not really within the control layer from a security, but also from just these are the tools that we've decided we want to use. Well, from a security and governance perspective, it'd be hard to say it's not a problem, But I actually, again, think it's a mindset issue. And it very much reflects how more enlightened people talk about what we'd call Maverick or Rogue spent.

15:40Rogue AI use or ungoverned AI use is a signal of demand. It shows where people want to use a tool or where they're expecting benefit. it. And rather than just shut it down and say, don't do that, the best organizations are often looking at shadow AI use and saying, oh, this is a demand signal for where we need better automation. I think a lot of people are trying to raise their hand right now and say, hey, I know a part of this process that could be automated or even eliminated. What can we do about it? If the posture is, well, needs to go through committee people are going to be pushed to the shadow use because there's this feeling we saw this last year but it's only accelerating now this feeling of inevitability that if i don't learn how to use this tool if i don't figure out how to be more effective myself then i am at risk and again there's this complexity in the actual details there whether people are literally at risk because whether they use air or not is debatable but But the ability to perform more efficiently, to deliver better results, and then associated with AI use, there's a certain halo effect to that in a lot of enterprises.

16:52So to that point, though, the shadow AI usage, when you look at it by function, you'd expect if IT is going to be there saying, I don't want more tools, IT is actually the one with the highest percentage of shadow AI use. They're the ones trying to go around their own processes to make themselves more efficient and test the adoption. It's not to say, oh, just let the tokens flow and spend on anything. That's not what I'm trying to say. But you have to be willing to live in the messy new world we're in, which is not yet fully formed. And you see the ROI benefits, but you see also, I would predict, better retention in your team and also just a better satisfaction in how they're able to work because if they feel like they're restrained in how they can get work done, they're either going to try to go around that or go find somewhere that allows them to work the way they want to.

17:48And I think that's increasingly how people are starting to look at AI. It's not necessarily just a threat, but if you have to, again, going back to the earlier point, the right mindset, if you're not able to use it, then you're not able to work the way that you feel like you're expected to work these days. Yeah, I feel like you have to be able to, allow your team the ability to experiment, you know, one for their own personal development, but also that's what drives adoption. That's what drives interest. That's what starts to connect with what the art of the possible is for the use of AI across all your different processes.

18:26And it's, like you say, it's a little bit messy at the moment, but I think it's absolutely necessary because, you know, you learn from that and then you can apply it as the models change and the capabilities of the tools around them change, then at least you have that grounded kind of understanding to build on as opposed to feeling like you're getting further and further behind. Yeah. And case in point, when we asked about what skills that people are hiring for right now, the number one thing people are looking at when they're taking on new people is the ability to evaluate AI output. It's slop detection, basically, but also So it's the ability to be fluent with this, really.

19:04I've heard scary examples where people are producing some sort of sourcing review. They've done an event, they've selected a vendor, and they're producing a summary, but they go to their boss, and the boss starts asking a few questions. And they can't get past the questions because they generated the summary, but they didn't live and breathe the summary the same way as if they'd written it up themselves. And so that AI output evaluation skill becomes critical in the future world because the volume of stuff we can produce is huge. And when so much of your workforce is actually using AI multiple times a day, you have to look at it as not just about generating.

19:45That's step one. The bottleneck actually becomes how do we verify what's useful and what's good and then actually apply it. If you get to how people use it, it's not just about the frequency, but most people, almost 90 % say AI is a net positive, but after spending time correcting for it. But that last part is really important. You can't just take the output and say, oh, I'm 100 % faster. You have to add in the part where you go through, you understand, you critique, you go back and forth. And so you can tell that the people who are better at getting ROI out of AI are actually also doing the evaluation work.

20:22And if you're not training people how to do that or hiring for that skill, you're missing out on the ROI to go back to the earlier point, really, because it's a skill set. It's a mindset. It's not just a tool. I think that's critically important because anybody now can go in and prompt a tool and get a lot of a big response and go and copy and paste that and present it as their own. and so you have that on one end of the spectrum on the other end and this is you know I spend so much time with these tools but I'm going backwards and forwards I'm asking questions I'm critiquing things I'm asking for you know I'm putting together panels of experts within or telling the the LLM to put together panels of experts who have different perspectives and I'm looking at those and thinking which ones do I want to deep dive in which then ends with something which may be AI generated that I'm using internally to help me process a decision and some decision making.

21:21But there's so much more thought that's gone into that process and what the output is versus here's one prompt and I'm going to send it and share it. And I think that we've done a lot of research amongst procurements and a lot of the people who engage with us as a community. And I think the number was 80 % believe that they can detect something that is AI written. And a lot of that percentage to have a negative opinion when they believe that it is AI written. But it's because they look at it as not being the output of some thoughtful process. It's just the output of a question and an LLM response.

21:56And being able to bridge that gap is such a big difference from a skill set perspective. And I felt it as well in the seriousness of the outputs and the influence that those outputs will have on kind of decision making. Now, I want to just kind of, as you look at the totality of the responses that you got and what you were kind of expecting when you went into the survey, was there anything that you looked at and just thought, wow, that jumped a lot in the last 12 months or that was something that I wasn't expecting or I wasn't expecting at the level of which the data came in. This is an interesting one.

22:41I think the planning horizons were very interesting to me. We had a whole section about how we're adapting here. And specifically, we were asking about how far out do you feel that you're able to plan right now? Universally, it's closer to like six months to a year. The feeling that people are aware of what's going to happen in the next year or they can trust that their plans will survive contact with reality. I expected a shorter planning horizon, but I didn't quite expect this feeling of just pure uncertainty. And I think the most interesting part to me was also, again, looking at this split.

23:24People who are more in the AI world have a shorter horizon. They just expect constant churn and change. Whereas the bystander group is a little more you know they think that things will be more as usual almost they can almost kind of live in their bubble yeah um and that's that's the part that i i it gives me a little worry going back to some of the earlier like 12 to 18 month window questions is that everyone sees the uncertainty you know you take tariffs you take geopolitical stuff but um when you get more involved in like the redesigning of your pro your functions there's people start to see oh i have a lot more work ahead of me um and they i think they're just expecting well in six months we're gonna have to go back to this um it becomes a more but if i take from like the software well a more agile process it was very popular to talk about this like 10 years ago to apply it to process design too but people are expecting to have to continue to iterate more often and one year planning in a formal cycle i mean we do that for for budgets at the enterprise corporate level but in reality i think a lot of people are looking at their functions and saying we need to go at the you know quarter half year level and just expect we're gonna go at this point and then we're gonna to re-plan for the world because we just don't fully expect to know based on every factor why it's going to be like that that's probably i mean that's just not the way that we're conditioned to think is it an act and the speed that we're conditioned to think and act it's set a one-year plan or set a three-year plan and just act quarter by quarter towards that plan um not having such kind of short term it's not necessarily short-term thinking but it's recognition that one things are moving so quickly and two things can be so much different in a quarter or in six months time but not being and i think this is important not having that inaction you know paralyzed by inaction because you're afraid of what's going to happen in the next three months and that feels like that's almost the biggest biggest hurdle to overcome is recognizing that change is going to happen but not not making that um an objection to doing anything and the mindset it goes into how we look at transformations in general.

25:45If you were five, 10 years ago, you would expect most types of transformations to be like a three to five year journey. And you'd have a huge roadmap with a partner and you'd go sequence by sequence in a kind of crawl, walk, run. The three to five year mindset though assumes that you will get to sequence these things in a nice, early way. And the data doesn't really say that. It actually says you have to move a lot of things simultaneously. You have to reshape your stack. And one of the more interesting stats in here to your earlier question too is 68 % expect AI native vendors like some of the frontier model providers to actually disrupt how the software stack works.

26:26It's a big finding. It's a big expectation that your traditional set of vendors, you don't even need to name names, just like the traditional set you've been working with the last 20, 30 years in some cases, face an existential threat from a new class of software. Don't get to finish some of your three to five-year plans while some of these things change so rapidly. And so the difference becomes then that instead of being super optimistic and certain about a given future, you have fewer people who are very confident about what the finish line looks like for them. And then you have people expecting, you know, I need to change every six months and I'm just going to have to iterate through that with whatever's available.

27:11There's obviously some foundation planning people are putting in, but they're not looking at these like three to five year, you know, sequential journeys that look nice on a slide anymore. I think that's uncomfortable for a lot of us, but learning to live in that world is the challenge if you are a leader today. Now, I think sometimes because we're around AI so much, you know, I'm sure you're around AI all the time. I'm around AI all the time. You know, sometimes you wonder, are we, and then you see people who are further ahead than you are. You're like, oh, I've got to catch those up. Yeah. How much of a, I don't want to say a bubble, you know, is it, is it the, the, you know, the 5 % who are really embracing and adopting AI and pushing through their organization or is that number expanding so it's no longer just the early adopters but it's becoming you know more of uh the the the 30 the 40 of procurement organizations rather than the 5 like where are we on the adoption and those are really um fully embracing the opportunities and the recognizance of having to drive transformation because of ai in their business?

28:24It's an interesting split because I would say there are, it's rare for me to encounter anyone who hasn't recognized that this is something they need to tackle. Yeah. I think where the split is, is again, the mindset around how quickly they need to change and then what is necessary to start changing. I think there are really some organizations who are leaning back on some of the classic objections. Oh, I need my data to be good. Oh, I need to have the right skills in place. I'll take the data one, for example. The data one is a real problem. I'm not trying to say, oh, it's fake. It's a real problem.

29:03But what people describe as I have a data problem is often really three or four problems. And those all need to be unpacked and dealt with in different ways. There's a master data governance problem. There's a your where is the data stored and how is it governed problem. And I think when you look across the spectrum of AI use cases, it's a little difficult to make sense of them all. And we hear a lot of these terms like, oh, you need great data in order for the AI to work, where there's actually some use cases that are not necessarily data dependent in order to get them off the ground. But there are some use cases that actually help improve go forward data cleanliness by helping prevent bad data entering the system in the first place.

29:53Again, there's human loop, there's governance around this. And then there's some problems that are huge problems that need to be solved with a larger project. But going back to what we were just talking about, you can't sequence these things like logically over three to five years and expect them to happen the right way. You need to start working on them piece by piece and being able to swap out some of these problems and tackle bigger ones over time. When you look at this builder group, they're much more likely to go forward with a use case pilot it but then push through the actual issues whereas those who are in the laggard bystander group they're they dabble they look at a pilot they say it didn't work maybe it worked a little bit but then they don't push past that um the dabbling is is really the actual issue it's not committing and pushing through some of those issues because you get a lot of information when you work through some of these issues and instead of letting those issues be a blocker letting them be a guide post into what to do next is kind of the unlock and that's where a lot of people need to focus to get past that initial hesitation if you haven't crossed the bridge into working in this new way well nick i want to wrap up there because there's there's a lot more we can obviously say about the report I also want listeners to get the opportunity to get their hands on the data themselves and see how they can apply that, benchmark themselves, their own organizations, see what some of the leaders are doing.

31:22Where would be the best way for them to go and get that? You can find more about the State of AI and Spend Report by going to zip.com slash stateofai. You can also check out a lot of what we've been writing about in some upcoming webinars we're doing. We're going to do state of AI and spend through a webinar format. We'll be on the road talking about this. And if you haven't already, you can also register for Zip Forward in New York State this year in September, where we will have a whole session going through the data and talking about some of the insights and recommendations we're making for those in attendance.

31:58Great. Well, I'll be there as well. So I'm looking forward to being at Zip Forward and doing some facilitation while I'm in town. So for anybody who wants to go and check out the report, I will include links to the State of Spend and State of AI Spend report in our show notes for today's episode. Those will be at artofprocurement.com slash podcast, where you'll find the show notes for all of the podcasts on Art of Procurement. That's artofprocurement.com slash podcast. Until next time, Nick, I want to thank you so much for coming on the show, sharing some of the data that you've been gathering through State of Spend and look forward to talking to you again soon.

32:33Awesome. Looking forward to it. thanks as always phil thanks nick if this episode struck a chord with you please do send it to somebody we grow here at art of procurement through word of mouth and that would be really appreciated you can also support us by giving us a thumbs up a star rating or a review wherever you listen to your podcasts since 2015 we've built the world's largest free resource for procurement professionals looking to elevate their impact. Our resources span podcasts like this, videos, blog posts, papers, and events. To join us on the inside and to ensure you never miss an episode, a webinar, an event, or a post, please do subscribe to our weekly newsletter This Week in Procurement.

33:18You can do that at artofprocurement.com slash subscribe. That's artofprocurement.com slash subscribe. Thank you for joining us today and I look forward to seeing you again soon. Thank you.