

28 Oct 2026 · 30m · 16 chapters

Messer Americas’ implementation of agentic AI for guided buying in procurement, emphasizing that agents must be taught company-specific processes and that autonomy should be case-by-case.
Diana McCulloch, Director of Strategic Sourcing at Messer Americas (world’s largest privately held industrial gas company). Leads strategic sourcing with category managers and lead buyers; procurement is split into strategic sourcing, operational procurement, and procurement excellence.
Guided buying agents reduce user back-and-forth, enforce compliance, and improve process navigation. Biggest learning: it’s not “plug in and turn on”—documentation and process consistency must be built first. Adoption improves with peer validation and real daily-life examples. Trust grows via testing; human-in-the-loop is preferred for supplier/regulatory/commercial risk.
Wave 1 agents for guided buying and buying channels; Wave 2 planned for requisition reviews and PO conversion/confirmation. Short pilots with specific success measures to avoid hype.
Written by AI. May contain mistakes.
Discussion on the challenges and learning in the field of procurement today.
“And today on the show, I'm delighted to be joined by Diana McCulloch.”
Philip introduces Diana McCulloch and her role at Messer Americas.
“I'm the strategic sourcing director for Messer Americas.”
Diana explains the three distinct areas of procurement at Messer and their functions.
“And we have procurement excellence that is the area focused on processes and technology for procurement.”
Discussion on the importance of creating value and partnerships in procurement.
“We are a very lean organization for what we do.”
Diana shares her journey with Agentic AI and the implementation phases.
“So could you talk a little bit about your journey and where you are in your journey with Argentic AI?”
Insights on the challenges and learning from implementing guided buying agents.
“I think probably this was the, there's a lot of memory span that we have in the company.”
Diana describes her transformative moment when realizing AI's potential in procurement.
“And to understand that it is not about the technology.”
Diana discusses how to handle skepticism and the importance of education for successful AI integration.
“I'm not sure how that is going to look like in my daily life for my team.”
Learn about the importance of integrating technology with education in procurement.
“And we were involving IT from the get-go as we were exploring.”
Discover how peer validation enhances the usage of AI tools in organizations.
“But also, we saw that after implementing the first agent, at the beginning, it was curiosity what drove them into using them.”
Understand how data collection improves purchasing decisions and processes.
“And that is providing insights that we can then use for continuous improvement or savings initiatives or just to understand where we need to educate more.”
Explore the complexities of trusting AI agents in procurement tasks.
“Because definitely the company is changing all the time.”
Learn about the importance of realistic expectations when implementing new technology.
“I mean, you think that you have tested and everything is super good.”
Get practical advice on starting AI initiatives in procurement effectively.
“need to input to actually get a solution that works.”
Discuss the evolving role of AI in procurement and the balance between humans and technology.
“And that as well, because that is how we do it.”
The hosts discuss the importance of shared experiences and insights in procurement.
“And really, thank you for having these conversations.”

Get the gist, explore an idea or ask about something said.
On Art of Procurement, what are the key points of this episode?
Automatic transcript. May contain errors.
0:00Hi there, my name is Philip Eidson and welcome to the Agentic Procurement Diaries. It's a show where we explore what procurement could become in an age of Agentic AI. The Agentic Procurement Diaries is presented in partnership with Zycus, a leader in Agentic AI for autonomous procurement. If you are a CPO navigating this shift, this show is for you. The truth is that nobody has all the answers yet, not even the most forward-thinking leaders. We're all learning, experimenting, and figuring out what this technology could mean for our teams and our function. In each episode, you'll hear from CPOs about what they're experimenting with, what they're learning, the challenges they're encountering, and how they're preparing their organizations for what comes next.
0:42These are conversations about the journey told by the people taking it.
0:51And today on the show, I'm delighted to be joined by Diana McCulloch. Diana is the Director of Strategic Sourcing at Mesa Americas, the world's largest privately held industrial gas company. So first of all, Diana, welcome to the show. Thank you so much, Phil, for having me. It's a pleasure. Now, for listeners, just to give a little bit of context, could you talk a little bit about your role, what you're responsible for, and then perhaps how procurement just fits into the broader organization at Mesa? Sure. So I'm Diana McCollum. I'm the strategic sourcing director for Messer Americas. And as you mentioned, Messer is the world's largest private health gas company in the world.
1:36We have operations in Asia, in Europe, in Americas. And we produce and deliver gases, industrial gases, medical, electronics, mixtures, et cetera. And for a very broad industry spectrum, ranging from healthcare, food, beverage, energy, utilities, you name it, it's very, very, very broad. And inside Messer Americas, we have three distinct areas in procurement. We have the strategic sourcing, That is the one that I lead, where I have category managers and lead buyers looking into the strategic spend of the company. Then we have the operational procurement that is in charge of the day-to-day operations and transactions related to procurement.
2:35And we have procurement excellence that is the area focused on processes and technology for procurement. And within these three areas, we look to the end-to-end process of procurement and making sure that we are delivering value to our stakeholders and go beyond, you know, just, you know, a price negotiation. Yeah. But really strength, that partnership with the business and bringing everything that we can. Yeah, so you're kind of, you're beyond the, we exist to help the business save money to the how do we create value of which cost is one part of many different areas where you support them. Yes, exactly.
3:27Yeah. How would you describe the culture of the team? So, Master's purpose is to help people drive progress. And we are an organization that is a have an obsession to serve, to get things done and to make things right, even in the face of adversity. And I think this culture definitely is very present in the procurement area. We are a very lean organization for what we do. and that makes us very hands-on, very entrepreneurial, very, you know, executional in some sort. We like ideas, but we like much better to have those ideas converted into solutions for the company. So that is basically the culture that we have.
4:21There's no ivory tower going on there then, it sounds like. It's about getting the work done, getting the job done. Yes. And we go beyond what is needed. As long as we know that is the right thing for a measure, we do it. Yeah. Do you feel that that's reflected in the relationship that you have with these stakeholders? 100%. Because they are seeing us not as a transactional thing or a step that they have to go through, but it's really a partnership that we are building. And very intentionally, we have been building in the past few years to ensure that we are not only bringing the savings, but also bringing the solution.
5:05Innovation is part of it as well. So could you talk a little bit about your journey and where you are in your journey with Argentic AI? Are you still kind of at the beginning stages, looking at what's possible? You've been piloting things, or are you a little bit further down the road right now? We're a little bit further. Actually, we are in implementation phase. So let's say that our journey started about two years ago and we were like looking at the market and there was a lot of noise and a lot of, you know, new fancy tools and things. And we were exploring and we were making like little pilots at the same time to make sure what really worked for Messer.
5:56But then at the beginning of this year, we started the implementation of the first agents and they are basically focused on guided buying and buying channels. So is to help our internal users to go through all the processes and requirements from procurement and go through documentation, policies, requirements, where I need to buy this. And that has helped a lot for the users to go directly where they should go. Also, not get all those questions and things from the users all the time. And also to bring compliance into what we are expecting the user to do. And we are already thinking in wave two, and actually very soon we're going to beginning implementation.
7:04In wave two, we are thinking on requisition reviews and how to conversion of requisition into purchase order and also PO confirmation. We have other things as well in the pipeline, let's say. We are thinking, okay, what else is out there and what makes sense for Messr? This is not like an end job and that's it or project, but it's something that as things progress, new solutions come into the market, we are willing to explore and see how they fit into our end-to-end procurement process in Messr. it's exciting to hear that you've started and that you're on kind of uh and i agree it's a it's a never-ending journey isn't it which is actually what's very interesting about these times is that there probably is no end point because you get further down the line and then you think well actually i could now do improvements on where i started and there's an improvement loop that comes i want to talk a little bit about the guide buying and um and kind of the use case what what were the pains that you are feeling that where that you identified you know what this is the place that we want to go and start?
8:19I think probably this was the, there's a lot of memory span that we have in the company. And we also identify that we not necessarily had all our processes as documented as we thought. And a lot of the knowledge was staying with our team and how you do things in case A, case B, case C. So actually one big surprise, let's when we started on the implementation was the amount of documentation and processes and things that you need to teach the agent because the technology is great, but the technology is not, you have to adjust it to the way your company is working. And that takes a lot. It's not just a plug it in and turn it on and it does the work for you.
9:34Exactly. That was one of the biggest learnings that we have had. And to understand that it is not about the technology. First, you need to put your house in place and organize it. so you can actually teach an agent how you are expected to perform. There's nothing like recognizing all the inconsistencies of how you do things when you're actually trying to teach an agent to go and do it. I've certainly seen that from our business as well and implementing some of these agents is that the time is in the what's the process going to look like as much as the implementation of the agent itself. Exactly.
10:19It is. it is was there a moment where you looked at and perhaps this is as you're implementing guided the guided buyer agent that that kind of was the aha moment that wow I kind of see what the art of the possible is here which is then you know leading to well next we want to do this process and next we want to do that process that was kind of the moment of okay where was this all my life you know i i have it i have a very like like like in my head it probably i will never forget it and in the beginning when we started to explore and the different vendors all i mean the demos usually they look very nice and everything but usually they were they were um For me, there was sometimes a difficulty to connect on, okay, this is how it looks like in your screen.
11:20I'm not sure how that is going to look like in my daily life for my team. But I remember when we saw the demo from the provider that we ended up partnering with, where I saw how in a very conversational way, basically a user would say what they wanted to buy and that would trigger a whole procurement process from the beginning to the end with policies, guiding the buying channel, the suggestion, advices and everything. And I was like blown away. And I was like, that was the moment for us, definitely, when we started to think about AI differently, not about how AI can help to improve the efficiency of a person, and more about how AI is going to revolutionize and change the whole procurement process.
12:27um it's exciting to hear because um it changes i think this is where a lot of energy is coming from right now is it changes your perspective on what procurement can actually be and what it can do and uh and i speak a lot about on the podcast about the idea of we grew up of procurement being these four walls that we were contained within and it was how can we optimize within those four walls and now those four walls have been knocked down because we can do so much more or we have the potential to build procurement to be so much more than it used to be. I'm sure that when you were starting to implement, you got a lot of skepticism as well and pushback.
13:02Could you talk a little bit about, you know, things that you had to defend or objections that you had from stakeholders or even colleagues about you want to do what? I think probably the approach that we have had, on one side, we have been very, very fortunate, I have to say. When we started this journey, we were clear about one thing and was we want to be as, as, as, as, as to start exploring as soon as possible. We don't want to wait for all the noise to be settled and to have the answers for all the things that we don't know yet. You're comfortable with the unknowns. Exactly, the unknowns. So we had a big support from our leadership team and from our CIO.
14:04So that was huge. And we were involving IT from the get-go as we were exploring. So we could have the conversation not only of what was possible and what we could do for our processes, but also the conversations about guardrails and data privacy and governance and the whole thing. So we were fortunate to have a lot of support on that side. Now, we also recognize that the technology could not come alone and that we needed to provide education also, not only for the procurement team, but also for the whole organization. Because if you want them to adopt and use this technology, you need to go to have these two things go hand in hand.
14:59So we have done a lot of education in general about the foundational concepts about AI. That has helped. But also, we saw that after implementing the first agent, at the beginning, it was curiosity what drove them into using them. But something we discovered throughout all these months is that what drives better usage is like information coming from a peer, like a peer saying someone else. Yeah, validation, like social proof. It is. This is great, et cetera. So we have been learning a lot about change management, especially around AI and how to influence people to use the tools more. and it's not a communication, it's not an email that they receive that is not going to take us anywhere.
16:10It's more about showing them real cases that touch them and what they do in a daily life. Interesting. Have you found that through the data you can now capture through this process that you've uncovered interesting things that you didn't know about how people buy or how they think about the the act of buying so i think in in general this is this is actually one of those areas that we think that we need to work on now and also into the future yeah data management data strategy and everything, the more, and it's not just AI, but we notice that the more we go into digitalization, more tools, better quality of what we put in our purchase orders and all that, that is also coming with all these agents, is bringing a lot of visibility into more details in what we are buying.
17:19And that is providing insights that we can then use for continuous improvement or savings initiatives or just to understand where we need to educate more. Right. Because definitely the company is changing all the time. People are changing, rotating and whatnot. And then the knowledge that you have built so far, You need to retrain and go back and make sure that people are following what you are expecting. So we are actively working already on that. And for sure, there's going to be more to come as more agents are coming in and better data is coming out. Right. As you start to bring in more agents, do you feel that there are areas where you would trust an agent to act without a human in the loop?
18:15Or is that a step too far at the moment? I think for us, it's not a yes, no question. Yeah. I think it depends. So when we look at this in different angles, one definitely is how much information, training, etc. we have been able to give to the agent, test a lot the agent to understand the behavior and if it's doing what it's supposed to be doing, that is one thing, but also the risk. What is the task that is expected from the agent? And what kind of risk we would have? So in the example of the buying channel, It's something internal, it's something that we put a lot of effort into the information that we put in and also on the maintenance of the information as well.
19:19And we tested it, we know it is good, et cetera. So we are good with not having a human in the loop. But when we are talking about maybe commercial exposure, supplier relationship, regulatory risk in some cases, those for us probably are something to consider a human in the loop. And at the end, we are not going for, or our objective is not to have full autonomy on everything, but to set the right type of autonomy depending on the agent that we have. Yeah. So definitely a case by case basis, but you don't necessarily, it's interesting, it's not a hard yes or no, but it's really dependent upon what the agent is doing and how confident you are.
20:15And it's, I think you build up trust over time. Yes. And how confident you are in its ability to do what it should do and not do what it shouldn't do. And it's a never ending as well. I mean, you think that you have tested and everything is super good. And then sometimes it gets you an answer that you are like, Are you sure? No, I hear that. When you look at your experience so far, is there something that has exceeded your expectations when you started? And on the flip side, perhaps something that hasn't quite delivered what you hoped it would through, you know, at least the early stages? Yeah.
20:57So in terms of exceeding expectations, I think has been more like as expected, to be quite honest. I think most of the suppliers out there, they do a very good job about upselling what they have. So the expectations are already too high. And when you get into things, I think probably we have more examples on the other side that maybe your expectations were higher than what you were actually getting. Yeah, the hype didn't match the reality. And because of that, for us, it has been very important to do short pilots with very specific measures of success so we can understand if that solution is a good fit for a measure or not, and if not, to not scale it.
21:57And we can get out and move on. And it doesn't take very long at all to do those pilots. It's like in your personal life, going and creating a free account with it and playing around for 10 minutes gives you an idea of what you can see as the answer possible or not. And pilots can be very similar. You can either understand this has got a lot of potential. It might not have the data it needs right now. It might not have the workflows it needs, but I can see what it can do and I can see the power of it versus this probably is not what I thought it was going to be. And when you open the box of the technology, let's say, and when you understand, again, coming back to our comment about the training and the information that is needed, when you open the box and see really how much you need to input to actually get a solution that works.
22:54that a lot of the time change your perspective in terms of this is going to work or not. Yeah. Is there any advice just from your experience so far that you'd like to leave listeners with as we start to wrap up today? Sure. I think my advice would be start sooner rather than later. And it's not about being perfect, but the more you explore, the more you see solutions and do your pilots and try it out, you are going to build that
23:31that knowledge inside your company and really make better decisions after that. So for us, that would be one. And second of all, the short pilots, that is crucial. And the third thing I would say is do not underestimate the amount of information that you need to have documented so you can train your agents yeah i think that's something we alluded to before is often missed and um um you need to train an agent just as much as it's a uh yeah i think of it as an intern that's coming into the business who um you know perhaps one that is uh you know has many, many, many, many, many times the memory and recall ability that we do and has perhaps got some procurement training, but still need that contextualization.
24:28And that is, that contextualization is what takes the time and needs to be, and you can't just do it, set it and forget it either. So it needs to be continually, you need to be continually teaching it as things change. And you hope that you can ultimately build in these self kind of learning loops so it kind of trains itself based on the outputs of what happened but it still needs a human to be overseeing that process yeah and remember these our companies are are not built from one day to another so we are talking about the years and years of knowledge that has been built and and just ingrained in our culture and the way of doing things so and so putting all that into an agent is when you start to think, oh my God, I have to teach this as well.
25:21And that as well, because that is how we do it. It's not a generic process. It's how Messer in our case does it. Yeah. When you start out thinking it's a three-step process. And by the time you get to step two, you realize I've got 10 more steps. I want to go and teach it now. The last question, and this is probably the toughest one given the pace of change at the moment, but it's just something that when you look forward, let's say three years, time is there something about procurement you think will be true in three years that isn't actually obvious to us all today that is a very hard one and i hope i i can have a crystal ball here yeah right and with the speed of things i i think three years is is is is way ahead i don't know exactly how it's going to look like what i do know is that definitely um there's gonna be more and more these agent plus human interaction.
26:17We don't see us only agents taking over. I don't see it. What we are looking or expecting from the future is that all that repetitive, not added value operations that we have done for forever and ever to be taken by agents and leave those more subtle judge, I mean, the judgment, the context, those things that are humanly in nature, relationship with vendors, that kind of things for me are still going to be human. and important to have a human there uh at the end is is gonna is gonna be how how to land into the perfect combo between humans and ai where we're going to have the best deliverable possible that is going to be much better than any any result that we can get from any of the parts alone yeah well it's and i said it's the hardest question for a reason because i think that things are moving so fast that that's that's one of the reasons why uh you know we we we started this specific podcast was i think that we're all on this journey together and it's a little bit of a journey into the unknown um and learning from each other helps us kind of weigh up the risks the opportunities and inspires us to get started and i think that the journey that you're on at mesa is definitely inspired um as in understanding where we can start and some of the benefits and And I truly appreciate your time today and sharing those with us and spending time with me and our audience.
28:05So thank you very much, Diana. Thank you so much for having me. It was a pleasure. And really, thank you for having these conversations. Because for all of us, as you mentioned, especially at the beginning, all these topics were kind of like a leap of faith. and hearing from others as well on what they are doing and what has been their journey. It helps everyone to understand we're not alone in this and get ideas as well, how we can do it in our own companies. Well, I appreciate you sharing your perspective, Diana, and we'll have to check in with you to see how you're kind of progressing as you get some of those new agents off the ground.
28:49Thank you so much. Thank you.
28:58Diana McCollough: 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.
29:35Diana McCollough: You 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.
Dear Diary… I think my advice would be: start sooner rather than later.
Most procurement teams expect the hard part of agentic AI to be the technology. Messer Americas found the hard part was everything they had never written down.
When the team began teaching its first guided buying agent how to behave, they discovered how much of their process lived in people's heads: how you handle case A, case B, and case C. The agent could only be as good as the house it was trained on.
Diana McCollough is Director of Strategic Sourcing at Messer Americas, part of the world's largest privately held industrial gas company. She leads the category managers and lead buyers responsible for the company's strategic spend.
In this episode of The Agentic Procurement Diaries podcast, Philip Ideson and Diana discuss:
- Why maverick spend and undocumented processes made guided buying the first place to deploy agents
- How Messer decides where an agent can run with no human in the loop, weighing testing depth against commercial, supplier, and regulatory risk
- Why peer stories drive adoption far more than emails or announcements
- How short pilots with specific measures of success stop the team from scaling tools that don't live up to the demo
Links:
-Diana McCollough on LinkedIn: https://www.linkedin.com/in/diana-mccollough-b4874823/
-Zycus on LinkedIn: https://www.linkedin.com/company/zycus/
-AOP Provider Directory | Zycus: https://artofprocurement.com/provider-directory/zycusinc
-Subscribe to the AOP Newsletter: https://resources.artofprocurement.com/art-of-procurement-podcast-subscribe
-Subscribe to Art of Procurement on YouTube: https://www.youtube.com/@ArtofProcurement



