Paul Morrison
Hello and welcome to Retail and Consumer Pulse brought to you by WNS. In this podcast series, we'll dive into the wonderful world of retail and consumer goods. Our mission is to spotlight the latest strategies, trends and innovations in this evolving sector. And to help us do this, each episode features an industry expert or leader sharing insights on what it really takes to succeed in this hyper-competitive marketplace. My name is Paul Morrison, and I lead the WNS retail and consumer practice in Europe.
For today's session, AI in supply management, I am delighted to be joined by my colleague, Mike Tracy, based out of Chicago. Mike is a senior VP in the manufacturing, retail and consumer goods practice at WNS. He has over 30 years of experience in transformation and has worked with leading brands such as Albertsons, McCain Foods and Kellogg's.
Hey, Mike! Thanks for joining us today.
Michael Tracy
Thanks for having me, Paul. I look forward to a great discussion.
Paul Morrison
Absolutely! So it's timely. You've just authored a shiny new white paper titled - Rising to the Challenge about AI and Supply Risk Management.
We're going to leave a link to that on the show notes, and we'll dive into that in a moment. But perhaps, first up, you kick things off by telling us about your story and retail consumer and supply management.
Michael Tracy
You bet. Thanks again for the opportunity to speak. I have, as you said, almost over 30 years of experience in both the application software and the BPM markets, working with retailers and consumer product companies, and helping them transform and optimize their cost structures. And in addition to driving sustainable growth, I've worked with companies, the ones that you mentioned earlier, Kellogg’s and McCain Foods, Harley-Davidson, Albertsons, Walmart, just to name a few.
I have been in the space for many years and thoroughly enjoy working with the clients in those respective companies, solving some of their trickiest problems. So, I think as the supply chain has evolved significantly over the last 20 years, so has also the supply base. And I think that kind of gets to the root of our conversation today around supplier management, which is kind of more of a general umbrella, of which supplier risk and assessment and supply relationship management and supply chain optimization fall under.
Paul Morrison
And it’s a really timely conversation. Just a couple of weeks back, we had that really vivid demonstration of the fragility of supply chains and ecosystems with the global IT meltdown. What's your takeaway on that?.
Michael Tracy
Yeah, it won't be the last. So I mean, if you think it, even very recently, you think about the supply chains that have been repeatedly disrupted by unforeseen events. You have a kind of geopolitical event like the Houthi rebel attacks in the Suez Canal, right? You know, the drought in the Panama Canal or even unfortunately, the Baltimore Bridge disaster. You know, most recently with crowd strike, and I think what's interesting about crowd strike is that was a security update that was sent out to Microsoft users and on the surface you would say, well what did that have to do with supply chain? But if you consider the ripple effect, you know it was devastating, and there were systems down for multiple days - order processing, order management systems, transaction processing, payment processing systems were down.
You know, just commerce coming to a screeching halt, and I think, one of the best kind of examples or at least the loudest example was with Delta Airlines. They were down multiple days, and as a result, they believe that, you know conservatively $500 Million of revenue was lost because of the crowd strike update patch.
Paul Morrison
Yeah, that's a big number, and there's some definite keywords in there around infrastructure resilience, single points of failure, proactive risk identification. These are some of the things we may come back to as we dive into your paper. Well, let's do that. And let's look at the problem statement. I think the question is, what is the challenge with supplier selection and supply management? Why is it difficult, or to put it another way, if AI is the answer, what is the question? How do you approach that?
Michael Tracy
Yeah. I think if you kind of break it down, how can you accelerate time to decision with supplier analysis, supplier evaluation risk and influence and optimize those. Your supply chains - these were processes that historically took weeks and months to kind of conduct now taking hours or days. And if you recall a typical kind of procurement engagement, a lot of it's centered, initially around supplier, supply base and supply consolidation, too many suppliers and a lot of redundant suppliers.
And when you break it down, you find that after you were able to consolidate the supply base which took months. For example, you come to the conclusion very similarly across clients which is 80 percent of the spend is with 20 percent of your suppliers.
And the ability now I think with AI that we're seeing is that you know what took months to consolidate supply bases now can take weeks. What took weeks, to be able to assess supplier risk, or even months in that case, can now take days and hours. And I think that's the question, right? What is the promise of AI and now, with a good 18 months head start in the business problem, I think that there are a number of solutions that we'll talk through here and in the next half hour.
Paul Morrison
Yeah. No, I like that. And I think you summarized it nicely. It seems then that the problem is that supplier management has been an aspiration for many organizations. But it's been out of reach because it's been too expensive, or incomplete, or not real-time, or low accuracy, poor insights and the promise the word you use there of AI is to make many of these higher expectations achievable for organizations. So, that's exciting. And I guess we should just very briefly pause on the point here around, you know, we're talking about AI. This is not the only podcast that's talking about AI at the moment. And I think for the purpose of this conversation, we're using a fairly broad definition of AI. But you know, that might include Gen AI, it might be machine learning, it might be deep learning, a combination of technologies that bring together pattern recognition, predictive capabilities, and that imitate analysis.
Michael Tracy
Right.
Paul Morrison
We're not going to get, we don't have time on this podcast to get too technical. But it's, would you agree it's a sort of broad range of AI technologies that we're talking about here?
Michael Tracy
Yeah, I absolutely agree. I mean, I think that with machine learning, large language markup models, multiple points of data sources, there are different techniques that are being used. Some that are embedded in third-party systems, some that are more project-based with third parties, some that are being done in-house, but from an AI perspective, there are multiple ways to not only look at the problem but also solve it from a technology perspective.
Paul Morrison
Absolutely. Well, as we sort of work through your analysis, I think the thing that really jumps out for me is that we talk about AI and supply management, and there are use cases that are visible scattered across the piece. There are many use cases of AI in production, making a real difference today for retail and CPG companies. So, you know it's really present tense and no longer future tense.
Michael Tracy
That's right.
Paul Morrison
You pick out five areas in the research of examples, maybe we could just step through those. Now the ones that jump out to you.
Michael Tracy
Well, just to preface that, I think if we were to have this conversation a year and a half ago, we wouldn't have all of those use cases. We'd still be at the beginning stages or, early innings of AI’s influence in supplier management. And I think if you go back we reference a Deloitte survey. 70 percent of CPOs indicated an increase in procurement-related risks and supply chain disruptions. 70 percent!
That same study identified AI-led supplier management as the foremost procurement strategy to unlock value. So, don't take it from us. It's prevalent across the industry, and certainly, it's beyond catching people's attention and people are now prioritizing as projects.
Paul Morrison
Absolutely. So first up, this may be the start of the life cycle. We've got market analysis and initial identification of suppliers.
Paul Morrison
Anything jumps out for you in that stage?.
Michael Tracy
Yeah. I think trends in new markets, the identification of new suppliers.
There's always something that is happening right behind the scenes. You know, there was a point where, again talking about cybersecurity and crowd strike. There was a point where that was all new. And most organizations didn't know much about that market. It was a burgeoning market. It was just beginning to start, and as a result, how do you get the information? Well, the good old-fashioned way as you would create a tab in your Excel file for that supplier. You'd have a number of basic questions and informational pieces that you were looking for. You'd go out onto a website. That's all changed. I mean the ability now with AI’s influence really is in speed and reducing the time to identify those suppliers.
You know, you have the ability now to go and scrape websites and pull down the relevant information and put it in a manner that's readable, or in some sort of display. And so, one of the things I found interesting in this was during COVID, which was kind of the grand daddy of all supply chain disruptions, was we had a lot of inbound inquiries around how can you help us right around safety supplies. And so, CPG and manufacturing companies, and even retailers too, were in need of real simple things – masks, sanitizers, soaps, things like that, paper towels.
And so, their MRO providers were out of stock on it just because of the overwhelming demand. Everybody was scrambling to try to find alternatives, suppliers, and in some cases, what we saw was there were a lot of local, call them mom and pop kind of providers. They had the inventory because they were not suppliers to the big companies that were out there.
There were companies that were going to Staples and Office Depot for some of these things as well. So, it's the identification of that available product was very slow, and at time, where speed was needed. And that's changed. AI is now able to influence that in a way of finding suppliers that you didn't know existed or were not on your radar before.
Paul Morrison
Yeah, absolutely. It's all about speed. And I work with an application called Forestreet recently, which is all around crunching the markets for some unusual categories and really finding the long tail of emerging suppliers before they were highly prominent and obvious, finding hundreds and thousands of suppliers that basically would have been previously invisible to a procurement function.
That's just one example. Yeah, it's totally new functionality. So, it's exciting. Can I move us on to the stage beyond that sort of initial market analysis, identification to more about the supply evaluation a bit further down the line. What are your thoughts there?
Michael Tracy
I think in a lot of ways around identifying and mark analysis and supplier identification that we supplier evaluation is being accelerated by speed. And what AI can do - AI again large language markup models handle large volumes of structured and unstructured data.
The ability to go and do high speed web scraping, you know of supplier profiles of their compliance data, their certifications, their financial and performance metrics. I mean all these areas.
Again, it would take weeks, where now it can take hours. And so, I think there was a really neat example we used here with Koch Industries, and they use an AI-powered tool to optimize its supply base in real granular data at the stop at the SKU level. And I would say that unlike traditional kind of procurement methods that focused more on high-level purchasing categories and aggregate spending, what Koch is doing with this tool is essentially ingesting these big comprehensive data sets that includes the supplier information, purchase orders, invoices.
And it offers detailed insights into those qualified suppliers and it identifies backup options within existing supplier networks. So going back to the example around, you know masks and sanitizers and safety-oriented, this tool would have done a wonderful job of that four years ago.
Paul Morrison
That's a great real example, and I think you cite another example, Unilever, I think has got something similar in terms of doing that breadth of analysis. So that's interesting as well. So, let's move on down the flow then to the idea of risk assessment and monitoring this idea of you know once you're in a supplier relationship providing that sort of proactivity that proactive assessment, continuous assessment, what's your take here?
Michael Tracy
That's right. And again, going back to the speed, you know that AI now is tasked with in terms of handling large language models, big, big data sets. The disadvantage cascades now into the risk assessment and the monitoring and what had been snapshots and very episodic data that had been collected on you know on a supplier risk is now able to be conducted continuously.
And that is the big gain or the big win in risk assessment and you're able to incorporate what recent incidents, right crowd strike incident, operational intricacies, financial stability, things like that. And we have a couple of case studies here.
One of ours, we used with a leading frozen foods. You know a company that was doing AI-driven continuous risk intelligence using our Smart Risk platform. Smart Risk comes to us via our Smart Cube acquisition back in 2022, but it's advanced technology that essentially streamlines the supplier network and identifies risks, and in this particular case, allowed this organization to proactively adapt to market demand. So, that's one way of taking a look at it. Walmart - we use as an example here. I was really impressed when I read about this. It uses AI, well I should say, it would probably be getting back into the SRM piece but, I would say from the risk assessment piece, the example we're using there is one of our own.
Paul Morrison
That's great. Let's look at that SRM piece with Walmart. What excites you about that one?
Michael Tracy
I'll get to the example in a moment, but I think with SRM, there's been a large evolution in the space. There's a lot of tools, third party that help manage that part of the process, and what a lot of SRM is really a lot of questions and answers, things going back and forth between the supplier, in the organization, the manufacturer of the CPG company. And so where you saw more of kind of FAQ-based inquiries, a set of standard questions and a set of stock answers, right?
The models now write me algorithms. They're essentially continuously learning. So that question on top of other questions that come in become a fully automated chat session with suppliers to streamline the communication. And in the Walmart situation, which I thought was really unique, they're essentially taking chatbots now to another level, to automate negotiations with tail spend suppliers at which is more of their small dollar spend, but it accounts for about 20 percent of its expenditures. So, it's impactful, but that approach that they're using – the AI-driven approach – streamlines the procurement process making it more efficient by handling those negotiations that would have historically always been in human form, whether there would have been phone calls teams or e-mail.
Paul Morrison
Hugely laborious across hundreds of thousands of suppliers, or probably never even happening at all, so it's a really good case, and I think there've been similar awards and cases for, I think, Siemens and T-Mobile, if I remember correctly, with the same sort of focus on tail end. And I think from there it can go to other parts of the portfolio as well and achieving really strong ROI on those automation. So that is great.
Michael Tracy
That's right.
Paul Morrison
It may be called the humble chatbot, but it's starting to achieve some really, really big impacts. So, that's exciting stuff.
Michael Tracy
That's right.
Paul Morrison
And our final chunk of the life cycle or sort of lens to look through, maybe it's a broader view of the supply chain optimization.
This is a big one. How do you see this space?
Michael Tracy
Yeah, I mean this is probably the home run area where AI is influencing this in a real positive way. I mean the algorithms can now look ahead to everything from lead times and transport costs and demand fluctuations by simulating different scenarios such as cost modeling scenarios, things like that, and that has been accelerated significantly over the last few years by using AI. One example I know that we use here is what Home Depot did kind of during the pandemic.
They really took this to a different level. As it related to inventory and getting its products out of places like China, out of stuck in the ports on container ships. One of the things that they did is rather than lease them the analytics that they had done, led them to acquire the container and the ship themselves so that they would own that part of the supply chain from end-to-end. And I thought that was a really interesting example.
One might say a drastic example, but if you looked at what lead times were on inventory during the pandemic, people couldn't buy cars because chips weren't available for months, not days and weeks. So highly disruptive to the supply chain, obviously was COVID. But, Home Depot, I think, had some real genius answers to that.
Paul Morrison
That's great. We've really rattled through five different areas. I think that for me, the takeaway is we talk about AI and in the supplier management space.
There are loads of examples, and it's really flourishing and accelerating. So, exciting place. This leads to the next question that you tackle in your analysis, which is really how to succeed? Where do you start?
And there've been plenty of organizations that have had these visions and had attempted to make the first steps, but success isn't guaranteed.
You set out four areas, one establishing a solid data foundation, investing in AI training, finding the right solutions, and then focusing on the human expertise side as well. Do you want to sort of call out just a few of the key must haves or success factors?
Michael Tracy
Yeah, you bet. I mean, I think on establishing a solid data foundation for AI is probably the most critical element of these four building blocks. I think it goes without saying that data foundation layer is really going to be the key based on the data sources that you bring into these models, the size of the data sources.
You know, whether you're talking about business intelligence or you're talking about AI - the old adage I think is still applicable, which is kind of garbage in garbage out as far as the data layers. So having scalable data lake, strong governance, I think, are essential building blocks to the model's accuracy. And I think one of the things that we've seen this far again early innings, but what we've seen this far is those investments really lack in the mid-size to small companies. It's still not a priority. I think it's very aspirational.
So they're still using a lot of Excel spreadsheets to manage inventory and forecast, and they're bringing data into small models. You know where they're able to manipulate it. But I think this is an area where third parties can really help accelerate and establish a data and analytics CoE, right from which AI applications can be built.
Paul Morrison
So, the Mondelez or the Kraft Heinz of the world have been investing in this data infrastructure for 20 years and they are putting it to good use now with AI. But other organizations haven't always had that investment. So, there's a bit of a catch up is what I hear you saying..
Michael Tracy
That's right.
Paul Morrison
I guess the other thought when we talk about data here is that it's a hugely variable space and complex. So, the nature of the data foundation is going to vary depending on the particular use case, the part of the life cycle that we're going to be focusing on. It's going to be very focused on the type of AI that's being deployed.
And it's going to vary, depending on how it integrates with your other systems. So, it's a simple bucket that contains a massive wealth of complexity. So, we need another podcast for this one.
Michael Tracy
Absolutely. Yeah, I mean you think about, do you have geopolitical risks and exposure in areas?
You know, in the Middle East or do you within your supply chain, do you have climate-related exposure in LATAM or in Asia PAC? I mean and those are data sources that you're going to be bringing into these models. So yeah, it's definitely not one-size-fits-all. There's a lot of complexity with it.
Paul Morrison
Yeah. Exactly. I think that's a great point.
One of the success factors you talk about is sort of finding the right solutions, which I think is a really interesting dimension. And I think, built into this, is the variation of the classic make or buy question - where do you buy AI, how do you buy it, how do you bring it in? And I guess from my perspective, I see organizations sometimes. AI will be baked into the ERP, you know their I value as other Coupa or Ariba(s) of the world and it'll be baked into the recent upgrade or release.
It may be part of the point solution or ProcureTech-focused solution or it may be something that's more home-brewed and crafted through combination of open-source or low-code/no-code platforms. So, there's a range of ways that you're going to be buying or bringing AI into your supply management function. I think what strikes me, I don't know, if you agree with this.
Probably, it'll be increasingly common to buy it in, not to make it. I don't know. Would you agree with that or am I going off piece there?
Michael Tracy
Yeah, I think the providers right in the ProcureTech space today are ahead of the builders, those that are building their own kind of solutions. And even if you're building your own even third-party help advisory or consulting help is certainly something to look at. We're working with an organization today, a manufacturer today. They're looking at steel and a couple of other commodities which you know they can track via subscription. But, they're looking for a different angle. And they’re wanting to incorporate more kind of data inputs that the subscription tools don't incorporate today or don't consider and which is one of the reasons that they're talking to us about that. And that gets back to the smart risk offering that we have. So, I think that the third parties right now are certainly ahead in this versus the builders, but I even think if you're building that, there's a role for a third party in there for advisory and consulting work.
Paul Morrison
Absolutely. And I think, the last of your recommendation is about the human dimension. There's a big debate here, I guess, in all AI discussions around how you balance high tech and high touch, and what's the role of the human here, and what do you see as the way forward here in this dimension?
Michael Tracy
Yeah, I think the misnomer is AI is a black box, and it’s going to run on its own and make decisions on its own. And there's a lot of doomsday predictions that I think get over inflated about what AI could do.
And AI cannot operate today in isolation, right? It requires human intervention, oversight and it's another area. I frankly think that the real competitive CPG and manufacturers are and will continue to invest heavily in, but whether it's their own people or they hire a third party, it still is going to require that human intervention, those skills, whether they're at the data management layer side, the data lakeside or write the interpretation of the results.
The display of those results, right? Who consumes those results, so it really is not the black box I think that people have been hearing about AI.
Paul Morrison
Interesting! So I takeaway from that - we've got many applications for AI, they're going to be powerful and enriching, manual work will be reduced.
Particularly in around some of that sort of data gathering and more the transactional spade work as it were, it's going to be creating a lot of rich real-time insights. But these insights will need to be assessed and put to work and acted upon. In a way, you might say that this is going to raise the bar then for professionals in that supply management space. Be careful what you wish for because expectations of them are going to maybe be higher than ever. So, we'll watch this space.
Michael Tracy
That's right.
Paul Morrison
Any we have to wrap, unfortunately, we've blasted through the best part of an hour, Mike. So that's great. Any final takeaways from your side, any final thoughts on this piece?
Michael Tracy
Yeah, I would say that as kind of getting the most value out of AI and supplier management. I mean, I think number one it’d be good to make sure you have a good outline and a good plan. And I think you know where do you want to impact? We kind of call out five areas.
It doesn't mean you have to tackle all five of them. So, we talked about market analysis and supplier identification, supplier evaluation, risk assessment, supplier relationship management and ultimately supply chain optimization. Those are all large areas. They could be daunting to tackle all at once but I think having a plan and an outline is key. And I think the discipline and rigor to become successful and optimizing your supplier management process.
I think we feel very strongly that speed to insights and continuous learning are within your reach and want to wish the best of luck to those that are out there prioritizing these projects.
Paul Morrison
Brilliant. It's a wrap. Thanks, Mike. We have to close it there. I've really enjoyed the rapid run through. Thank you for the discussion and I look forward to crossing paths soon.
Michael Tracy
Likewise, Paul. Yes, take care.
Paul Morrison
Brilliant. And I look forward to your next market research paper. So, we'll come back around the mics in a few months, I'm sure. And thanks to our listeners for tuning in to the Retail and Consumer Pulse. If you've enjoyed the show, please do like us and follow us, and we look forward to having you on the podcast again. Thank you very much.
Michael Tracy
Thank you.