The Decisive | Season 5 Ep.5 - Agentic AI Gives Data a Voice in Supply Chain Management

View Full Transcript

Kristen Hallam:

You're listening to the Decisive Podcast. Insights and analysis to empower confident decision making. Hello, I'm Kristen Hallam, your host for this episode of the Decisive Podcast. Today we're talking about pain points in supply chain management and the potential of agentic AI to address these issues. If your supply chain meetings have turned into daily triage, or if you're wondering why your head of procurement is guzzling coffee and popping antacids, this episode is for you. Making his triumphant return to The Decisive Podcast to discuss these issues with me is Brendan Neary, a director in our product management team. Brendan, it's great to have you here.

 

Brendan Neary:

Thanks, Kristen. Happy to jump on and be part of the podcast again, this time in a new light, doing a new role. So excited to talk about AI and our plans for AI.

 

Kristen Hallam:

Yeah. All right, well, let's get to it, Brendan. So managing supply chains has been a tricky business for years. No secret there. When you talk to supply chain and procurement leaders right now, what are the most common pain points that you're hearing about?

 

Brendan Neary:

Yeah, I mean, I think supply chain overall in the industry, depending upon who you talk to, is a very encompassing word. So it can be a wide variety of roles. And I mean, if you've ever been to a supply chain conference, you can see everyone from a warehouse manager to a delivery person. You see robots, you see computer tablets, you see literally automated drones coming across the oceans to deliver all the way down to final mile and things like that. So there is a wide encompassing variety of products that you would actually utilize for supply chain. In the space that we typically live in is a lot of the procurement range. And we historically have always been a data provider for that. So we've been providing data to help solve for solutions of pain points that our customers have.

I mean, we have a product pricing and purchasing that was born on the back of government contracting. And so when you look at a contract negotiation on a renewal for your current supplier, someone would say your contract next year is going to go up 10% because of X, Y and Z. And so a lot of our customers historically would use us to benchmark that and understand why their contract should be or should not be going up. And if they don't believe, based on our intelligence it should be going up, they would go ahead and negotiate that rate and then they would have a final outcome. So, I mean, there's ways that we've been in the space and I think that customers have been using us for that. I mean, we will do anything from contract negotiation. We will find alternative suppliers. So providing leads to understand people that you could alternatively source goods from.

We'll also tell you country risks associated with countries. We'll do credit risks associated to a lot of that. And I think since we've been in the data business, we've always solved for a lot of these issues. And I think now as we look towards the future of the ability of AI and how we can leverage AI to speed, to deliver, to create more intelligence, we're most excited about giving data a voice. So we're providing you a workflow of information that's providing you insight that's actionable and deliverable, that you can interact with.

It's going to be a new version of agentic UI, or user interface, but it'll be very different from the current experiences that our customers have now. But it's kind of starting to see the way that goes and how we can start to kind of do that. But in terms of what problems we deal with, a lot of times is just the ability for our customers to have choices. Right? It's a contract negotiation. What can they do? How do they actually benchmark it? How do you come back with intelligence to combat a problem that you might have? Or how do I find someone new? As opposed to just going to a trade show or spending time asking my friends, like, how do I validate, you know, these problems in a virtual way? And then now it's how can I access that with speed, Intelligence, accuracy, quality, all of which is what we are driving towards and why we're very capable to be in the AI business going forward. Probably the world's best company sorted for AI intelligence. Because the nature of our business is data and it’s cleansed, quality, structured data that you can provide insight from.

 

Kristen Hallam:

This is why we're here, because you've just listed so many things that we can unpack. Let's start by unpacking some of the telltale signs that businesses, our clients, are struggling with some of these issues and basically need either some different tools or maybe a different workflow to, to, to deal with that.

 

Brendan Neary:

Absolutely. Our customers can utilize our information for either cost savings or for revenue growth opportunities. Right? It's a problem for businesses that lean out. Everybody seems to be doing more with less as we continue to go in the workforce across the world. So our jobs expand and our ability for us to do more, we need to have more at our ready fingertips and I think that's a lot for what our customers are asking of us is I'm taking on, I'm expanding a role. I need more subject matter expertise, I need to have more intelligence in order to effectively do my job efficiently because I need to do it quickly because I have so many other things that I need to take care of and I need to make sure that everything that I do is pointed and there's a reason for why we're driving towards a contract or driving towards a new vendor or driving towards changing our supply chain.

Supply chain is a very diverse place. I think that it also is very volatile. I think if you ask anybody in the supply chain space, they are probably walking around on fire most of the day. I think nobody's ever hey, supply chain is very calm and there's so many drivers that can change your supply chain. It could be weather, it could be government policy, it could be wars or conflicts. It's very crazy to think about how one thing can daisy chain across your total business. But supply chain gets affected so much that really being on the ball and having a backup plan for your backup plan is what you really need to do in order to stay diverse and to continually risk mitigate. And I think that's a lot of supply chain is mostly risk mitigation for having that backup.

Because of this, so much is always going to go wrong within a supply chain because they. I don't think you could ever describe a perfect world to anybody in supply chain because they would probably poke holes in it instantly.

 

Kristen Hallam:

Yeah, see that image I, I put out there early in the podcast, in the intro to the podcast about the triage meetings and the coffee guzzling and acid popping, et cetera.

 

Brendan Neary:

Exactly. I couldn't agree more. I mean I think our phones last year blew up more than any other time when it's just been a constant barrage of whether there's vessels broken down, there's tons of government policy that affects the total trade of businesses and what to do. I think your phone never stops ringing. We're hoping that in 2026, it's a lot of lessons learned and it's a lot of ability to kind of pivot and be more dynamic, knowing that you've had a previous year that was pretty volatile in the supply chain space. So we're pretty excited to kind of start to roll out and put out more in terms of creating some workflows and some abilities for our customers to have an intelligent design of product that can help them in a quicker, faster, more informed way.

 

Kristen Hallam:

Let's talk about that a little bit more. We have more data than ever to make smarter, more nimble decisions, to manage risks in the supply chains. What's the biggest challenge to deploying all that data to solve these problems?

 

Brendan Neary:

I think that's an age-old question. I mean coming from just my career in general, I started in a product space, I think in a singular product space, which is a lot of what most people have started in. Right? You buy something, you utilize it, we sell something, we sell it, we kind of know the use case. And then as at least my career here within S&P, as I've grown within an industry and talked to our customers and learned their personas and learned what data they subscribe to and how they use it and what they do, there's a lot of different data that you can leverage to create an informed decision, but you would never know that. I think one of the problems is not only the ability for us to interconnect it and provide that insight, but a lot of it too is just knowing the art of the possible for our customers. Knowing what our available data sets are is I think one of the greatest strengths of what we're trying to promote for 2026 has been we are not only in this space, but we live in this space.

And you can utilize these different data sources. There's elements of the data that overlap that will then daisy chain your data to be more and more intelligent. Right? To provide a workflow that helps. And that's kind of what we've been working on, is taking historically four different products that have sold out to their own use cases within different personas, across different industries, finding commonalities in different workflows and then creating an ability for our customers to interact with that and provide insight for their specific purpose. And that's kind of what we've been driving towards. We have a new beta AI agentic workflow that we've created that can help with some alternative sourcing. And our strategy behind that has been taking disparate data sets. So our macroeconomic trade data, our economics and risk data, our bill of lading or transactional shipment data, and our risk gauge data and creating a workflow for our customers to find strategic sourcing.

So if I were to type into the chatbot like steel for instance, or rice, or name any product that you would want to potentially buy as a procurement officer, it will tell you the top countries that export it by value over time. It will then tell you the actual risks associated based on our data. So it will tell you what are the geopolitical risks, the cybersecurity risks associated with that country and give you a risk score, a risk outlook forecasted score about that. And then from there we'll provide you the top companies that are associated to those countries that are exporting that product as a lead generation. And then from there we'll also provide you a credit risk scoring based off of our risk gauge information. So we'll actually provide you like, hey, is this, is this someone that you should actually look towards based on some of the risks associated with their financial health? So taking into account multiple data sets in a workflow, meanwhile leveraging an LLM to provide you insight as to what that data is telling you. So I think part of that is us allowing you the ability to access our data in a workflow. And the second part is adding an agentic voice to read out our data and provide you insight in a consumable manner.

Right. I mean instead of looking at a table or a graph nowadays you can get it in a, in a word format that's readable, that's understandable, that's succinct and very easy to kind of leverage and for our consumers to understand and then copy-and-pasting it into a PowerPoint deck. It's already consumable for a customer base. I think the days of us being, like I said, that data company versus that intelligence company, it's, you're really taking into account that agentic language and we're giving data a voice that our customers can consume and glean out information way easier than historically having to pull out numbers, run analysis, look at them and try to write it out themselves. It's just a speed to deliver that intelligence is instant and you're getting a lot of that information. And from an accuracy standpoint I think it's just we are in such a space where we are dealing with numbers and data and it's a lot easier for us to benchmark our accuracy and ensure that our model's working out correctly, as opposed to if you were to look at something like an open source AI, which is a little bit more encompassing of the Internet and Internet driven traffic and it has a lot more chance of hallucination, where our system isn't really going to be hallucinating that much just by the nature of what it's actually pulling at, which is just raw numbers and tables.

 

Kristen Hallam:

Now, there's been a lot of talk in the market about the promise of AI and supply chain management. And the picture that you've painted just now, Brendan, is a rather vivid, positive picture of how AI can be deployed in this way. What sentiment are you seeing in the market when you talk to companies about AI and using it for supply chain management? Are they eager for AI solutions or is the industry maybe a little bit more cautious about adoption at this stage?

 

Brendan Neary:

I mean, I think AI across the board is a—I wouldn't say a polarizing topic, but I would say it's definitely varying person to person, company to company. I do think that there are a lot more startups that are eager to use AI. I think a lot of the larger organizations are a little bit more reserved in how they're trying to roll out. They're really starting to put their feelers out to understand where the market's going, to understand how AI can be implemented. Whereas then if you have a small startup with three people and you're just trying to get the most speed to market, I think you can leverage AI and you can do that in a very fast and efficient space. You're also less worried about having a quality issue where you have a brand to worry about, where I think there's, there's a lot of large institutions that have some level of risk aversion when it comes to negative impacts of AI.

So I think there has been a lot of that. But I think we're getting to a space where AI is generally accepted as it's already happening. The genie's out of the bottle. You don't put it back in. We have to adapt. Are we adapting to the benefits of the business? And I think that's kind of what we're working towards. When I mentioned our agentic model and our workflow, we are not at the point where we're just putting out a total chatbot that all encompasses all of our data and triangulates all of this information that you can just take into and you're just asking it questions and solving all of your problems. Like that's not what we're doing right now.

We're making it very focused. We're using very small models. Small and, and we're taking a look at our data piece by piece while we're kind of interconnecting everything together. So the supply chain space is a hundred percent ready for AI. You have automation, you have workforce. Like in terms of the things that could move goods within your warehousing space, that's already done. I mean, that stuff already exists today. It's already going to be ever growing in terms of autonomous vehicles and what we've seen there.

So that type of AI exists and will continue to expand. And I think that our customers are starting to see the value of taking in some AI and cleaning out a lot of information. And I think really from our perspective, it's providing that intelligence in a consumable manner for our customers. Right now, it's just speed to deliver, it’s accuracy. We can interconnect a lot of information and intelligence to provide a lot more insight. They're buying five products, they're getting 10 products, and they would want to know how to piece those things together more insightfully for themselves. And I think like where we're starting to kind of work towards is where do you Venn diagram your workflows of all of your different products to kind of create a singular workflow specific to that persona and how can they leverage the different insight for what they're trying to do? Knowing that we cover customers in procurement, we cover customers in transportation and logistics and financial institutions. So we have a workforce across personas, we have workforces across industries that we have to solve for.

And that's really what we're excited about, is we've spent so many years providing insights to our customers and working with them, we feel like we have a good understanding of really what they're trying to accomplish. And now we can effectively enter the market and help them do that better with AI. And I think that's why we're excited about it is it's helping us interconnect our data more with our customer. And that's ideally what we want to do to drive. What does success look like for us?

 

Kristen Hallam:

All right, well, we've been talking about agentic AI here and late last year, S&P announced a partnership with IBM around agentic AI. And first up on the list is creating some AI powered tools to tackle some of these challenges that we've been talking about today. But maybe before we get into the partnership, maybe we should back up. And for those who may not be familiar with agentic AI, could you just give us a brief description of what that is?

 

Brendan Neary:

Yeah, I mean, absolutely. Typically you're using AI as just the ability to have automated intelligence in some degrees. When we think about agentic AI, it's more of an autonomous system that is leveraging language models, typically large language models. Right? So you're taking kind of an autonomous system that can basically reason itself. It can achieve whatever action or activity that it needs to do based off of a human supervision. Right? So it's the idea that it can work by itself, leveraging information.

And, I mean, when you think about AI, a lot of times when I talk to people, it's the ChatGPT of the world or the open source AI, anthropic, the large language models and the ability to have this autonomous system that you can interact with, which you can chat at and it will create decisions and decipher a lot of its information from there. So in agentic AI it’s mostly a chat feature to provide, you know, an insight based off of all of its information that it's gleaning out and tasks that it's doing. So it's the ability for that to be implemented in our data. But yeah, I mean, I think a traditional AI versus agentic, you're typically looking at a lot more probing of the AI where you're actually like prompting it and you're kind of forcing it to do something and you're having like a human interact completely with it. And then it's a programmed capability versus an agentic model eventually could be a little bit more autonomous to be running on itself and doing based off of tasks given to it. So like providing the ability for it to have a script, go out into the world, call someone, have a discussion, take down information, provide it back to you, is a little bit more into the agentic model of beginning to kind of look at that task, but also has some capability to understand the language that it needs in order to provide it. It's not just your typical script of reading out loud. It can kind of free flow a little bit from the script and provide more human-like context based off of that.

 

Kristen Hallam:

Cool, cool. So now could you tell us a little bit more about that partnership that I mentioned, the S&P IBM partnership, Brandon, and why you're so excited about it because. No, you're excited about it?

 

Brendan Neary:

Absolutely. I mean, I'm definitely excited to be working with IBM. IBM historically being a infrastructure, a software, a technology business and S&P being that data-intelligence business, it blends itself together to be a great partnership. I'm also excited because we get to jump to the front of the line for AI and start to work our models and work towards a greater partnership, working with the IBM team and talking to them and co-developing things together, also bringing things to market ourselves, potentially even going into market with them into their marketplace. So it's a partnership by which we can kind of create things together, driven by leadership. And it's almost like a universe that you're getting into. It's not necessarily like your typical agentic AI.

You're not just looking at one LLM, you're looking at a total orchestrator. So an orchestrator allows you to plug in different contexts for you to kind of create this total insight, which is if you think about the agentic model and you think about having an LLM, when you start to create large models, you increase the chances of hallucination. But if you kind of create more small language models, you're keeping it and you're narrowing that creativity or that quality of that information being designated to these smaller agents. Therefore, by default, you would have less hallucinations and an increased quality. So the idea is coming into a marketplace where you have an infrastructure like an IBM, they have a kind of like a bring your own license for different aspects. I'm using their orchestrator and their ability to kind of maneuver the data as well as governance and quality of the data.

They have a very strong program for kind of looking at quality assurance and algorithms to ensure that you have a quality response coming out of your agents. I mean, ideally we would provide a holistic solution for our customers that they would be gleaning out information and getting a total response. But for instance, you know, if we were building a tool about understanding your vendor and looking at vendor risks, we have all sorts of data that we can throw at vendors and a vendor risk based on credit risk based on geolocation risk, and we can build out an agentic model of looking at the risks and providing you a risk score associated to that. But one thing we wouldn't do is we wouldn't actually go into an open source version of an LLM and an open source model where we would start to pull in information outside of our universe. Right, but you could take another model and then attach it to our model and run it together simultaneously and add that value to there. So I think that's kind of an interesting concept is you can kind of go to the table and take our data and take other data and build your own models and products.

And I think that's the way that the industry seems to be going with all the disparate data sources and data solutions. Finding one that's going to rule them all in the, in the Lord of the Rings classic nomenclature for the future of AI, I don't think it will be there, but I think the ability for you to have, connecting all of your different current, whether it be databases, mainframes, CRMs, SaaS products, whatever it is, and then plugging them into a universe that's taking out all the information of your own personal data that you have for your company, whether it's your vendors, whether it's your payroll, whether it's your employees and you want to benchmark them to an insurance plan, you want to benchmark your vendors against a risk portfolio. You want to make sure you manage your stock of all the different products that you have. And you want to have indicators of when you need to make new orders and who those orders go to. That type of decision and build could exist within a model for your AI. And you could use all of that insight to kind of create and link all of those agents together to do that on your behalf and kind of create your own company infrastructure. So I do like the idea of, you know, you can use a third-party solution that could create that.

And I think that's why we've gone that way so far, is we want to put data in the hands of our customers, not only for us, but also for them to be in a universe that they can attach their own models together. And I think that we pair very well together in a common goal of doing that.

 

Kristen Hallam:

Awesome. Now, I had a question about TPM. First of all, I want to remind our listeners, in case they missed it, that we had a January 17 episode with Peter Tirschwell talking about TPM and what to expect at the world's largest container shipping conference this year. So, yeah, just wondering what you're excited about for TPM, Brendan, and what will you be doing there?

 

Brendan Neary:

Absolutely. I think first and foremost, my favorite part about TPM is seeing the customers that we've had for years. Probably my most memorable times at TPM have been with our customers, whether we're attending a session together, getting lunch or getting dinner with them. In terms of what I plan to be doing, like I said, we have our demo station. We're going to have an IBM discussion. I'll talk a little bit about what I'm excited about and we'll be at the booth, so if you have questions, you can come by and ask.

And we'll have a version of it that we can demo on screen in a beta version for some of the agentic workflow that we have and how it works, which I think will be cool. And then we'll also just be talking to our customers and like I said, just really trying to glean out what's next for us and what we think our customers are ready for, what they would like to see as well as while we're on our own in the R&D labs, sitting behind the screens, trying to figure out what we think would be cool. And then we're just going to start kind of building and betaing out to our customers and see what works, what's consumable, what they like, and we're excited about that.

 

Kristen Hallam:

All right, well, it sounds like you'll have a lot more to say about that, about AI and supply chain management in the coming months. And Brendan, I could talk to you all day, but it is time to wrap up our conversation. Any final thoughts for our listeners?

 

Brendan Neary:

It's been a pleasure being on the podcast again. I was very fortunate enough to get the call up again, so I feel very accomplished. Hopefully we'll have a further follow up about kind of Maybe there's a 2027 episode of what we've done in 2026 and where we're going into the future state. So I'm excited for the next year of R&D and development and we'll see where we go from there. Thanks again.

 

Kristen Hallam:

Yeah, sounds great. Well, thank you so much Brendan for taking the time to share your insights with us. And thanks to you, our listeners, for tuning in. Click in the link of the description of this episode if you'd like to learn more about TPM or register for the conference. Also, we'll include some links to our thought leadership. Until next time, thank you for listening to the Decisive Podcast from S&P Global. Please subscribe and join us for next week's episode. Until then, stay curious and stay informed.