Fixing Scope 3 with AI: Supplier Engagement, Data Accuracy, and Decarbonisation Levers

In this week’s episode of the Sustainable Supply Chain Podcast, I sat down with fellow Irishman Paul Byrnes, CEO of Mavarick AI, to explore how manufacturers can use AI and data to tackle the notoriously difficult challenge of Scope 3 emissions.

Paul brings a unique perspective, rooted in both deep data science and hands-on manufacturing experience, and he didn’t shy away from the hard truths: most companies still struggle with messy, unreliable data and limited supplier engagement. We unpack why primary data will soon become table stakes, why spend-based estimates can be 40% off the mark, and how engaging suppliers requires a simple but often overlooked question, what’s in it for them?

We also discussed where AI genuinely moves the needle:

• Boosting confidence in data accuracy by identifying gaps and “contaminated” entries
• Providing personalised training to help suppliers meet sustainability requests
• Uncovering and prioritising decarbonisation levers with clear ROI

Paul shared real-world examples, from medical devices to automotive, that show how targeted projects, rather than trying to tackle all 15 Scope 3 categories at once, deliver the best results. We also touched on the environmental footprint of AI itself, energy, water, rare materials, and how responsible computing and smaller, purpose-built models can reduce the impact.

For leaders wrestling with emissions strategy, Paul’s advice is simple: start by mapping your data landscape. Know where you’re rich, where you’re poor, and build from there.

This is a practical, candid conversation about making sustainability and profitability work hand-in-hand, and why efficiency wins are so often sustainability wins.

 If disruption hit tomorrow, would you know where your supply chain was most exposed? In 15 minutes my free scorecard helps you assess 27 resilience statements, calculate your score, and turn the result into three priorities and a 30 day action plan. You can download the scorecard free at tomraftery.com/scorecard.

Transcript

Copy

0:03

The easiest way to do it is actually at the design stage. And we hear this constantly, right? It's like, why are we retrofitting products that were never built to be sustainable, to be sustainable? And why aren't we focusing on our R&D and, and building for the future? Good morning, good afternoon, or good evening, wherever you are in the world. Welcome to episode 85 of the Sustainable Supply Chain Podcast, the only podcast focused exclusively on the intersection of sustainability and supply chain. I'm your host, Tom Raftery, and I'm delighted to have you here today. Before we get started, a quick reminder, you can now support the podcast and unlock the full back catalog by becoming a Sustainable Supply Chain Plus subscriber. For just five euros or dollars a month, less than the price of a fancy coffee, you'll get access to all 80 past episodes of the Sustainable Supply Chain Podcast and all 380 plus episodes of the Digital Supply Chain Podcast. The most recent four episodes are free for everyone, but the archive is exclusively for subscribers. Plus members get a shout out on the show and a direct line to me for suggesting guests topics or even shaping where we take the podcast next. You'll find the link in the show notes or at tinyurl.com/ssc pod. Now onto today's conversation, and you are in for a treat this week because you're getting not just one Irish accent, but two. I'm joined today by fellow Irishman, Paul Burns, CEO of Maverick ai, whose background spans data science, machine learning, and deep family roots in manufacturing. Paul and his team are helping manufacturers tackle one of the toughest challenges in supply chain today, scope three emissions. From cleaning up messy supplier data to making AI genuinely useful for decarbonisation, his perspective is both pragmatic and innovative. If you are interested in how AI and the data can unlock smarter, greener, and more profitable supply chains, this is an episode you won't want to miss. Paul, welcome to the podcast. Would you like to introduce yourself? Thanks Tom. Sure. So my name is Paul Byrnes and I'm the CEO of Mavarick AI. Prior to Mavarick, I worked in a number of data science positions in the Apple supply chain and in the IBM supply chain. I have a PhD in statistical machine learning, so I have a very data focused background. I am a second generation manufacturer, so my family is involved in the manufacturing and production of orthopedic implants. So manufacturing has always circled around me through my career. And Mavarick is the best of both worlds, essentially, it brings together deep manufacturing experience and cutting edge technology to help manufacturers solve very difficult tasks. And tell me, Paul, about Mavarick. What's the origin story? What you know, what made you sit up one day and say, I know I'll start Mavarick. Yeah, so, so Mavarick in its very simplest way, Tom. It's a, it's an AI and data company. And our goal specifically is to help manufacturers make better decisions in their supply chain with data. Given my, my links to manufacturing through two generations now, I've always been aware of the problems within the manufacturing space, but in many ways I was too busy solving issues in other areas of industries. The more I heard about issues within the manufacturing supply chain, such as data quality, supplier engagement, the ability to reduce costs and increase margins, the more it became evident that there was an opportunity for us at Mavarick to help companies within very complex supply chains to actually solve these tasks. Now, in terms of the, the origin story of Mavarick itself, the real catalyst that helped Mavarick start was that I was listening a lot to manufacturers telling me that they had specific problems. There was solutions in the market for the problems they had, but they felt that those solutions weren't listening to the actual users themselves, and they felt that they were being provided with solutions that didn't really fix their problems to hand. So we very much want to adopt a different approach at Mavarick, almost a co-design approach that has been shrouded in the foundations of listening to the users, making mistakes with the users, and ensuring that we're actually developing solutions for them that moves the needle for them internally. Okay, and was there a particular aha moment? The particular aha moment actually came by accident. We started off initially looking very much at how can we help manufacturers make more with less so looking at efficiencies and, and throughput within their plants. And we f ound that the majority of our large customer base were starting to ask us more about their supply chains. So asking us more, how could we help them understand the emissions in their supply chain? How can we help them understand, educate their supply base, and how can we help them understand how to could reduce costs in their supply base? And the aha moment for us, Tom really came is that when we looked at our technology and looked at the data sets we were engaging with, the vast majority of data sets that we were very, very well versed in were the data sets they were struggling with. So as a team, we made the decision, well, if more and more customers and more and more companies in our industry are asking us to solve a repeatable problem, then let's see if we can solve it. And it turns out we can. Okay. And before we get into AI and we, we will come to that in a while. What's the biggest data challenge manufacturers face, particularly when they're trying to decarbonize their supply chains? The biggest thing really is even a step before that, and it's understanding their supply chain. Even if you consider a medium sized company, maybe with a thousand suppliers, they're not truly aware of who their suppliers touch, where their suppliers source goods from and how those goods arrive on site. Now, if you do overcome that hurdle and these companies have a really accurate supply chain mapping, then it's data accuracy and it's the age old question, can I trust my data? A lot of these companies are struggling with almost a qualitative conversation amongst colleagues where my number, say this, your number say that. Do we meet in the middle? Now, it's very simply, these companies want a number that gives them a confidence over their, their data set and manufacturers, a lot of that data accuracy issues stem from the way data's being gathering. So, internally there's a, a mixture of manual data gathering. There's a mixture of historical data gathering. There's a mixture of automated data gathering, and then when they go beyond their front door to their supply base, it's often the case where suppliers are struggling for resources. They might not be well versed from a knowledge perspective or have the skills internally around decarbonisation, so when they're getting asked for data from their customers, they're not really sure what they're getting asked for. So it already much does boil down to data accuracy. In terms of one of the biggest challenges for the, the supply chain. And how do you fix the supplier data issue? Fixing the supplier data issue. I, I think this is something that before we engage with any company where we're very confident it's gonna come up because we haven't really met a company that's nailed it yet. What we're seeing in industry, Tom in general, is a lot of the, the larger OEMs are taking a bit of a Pareto approach where they're looking at the smallest number of suppliers that make up the largest amount of their emissions or their cost base and working with them very closely. The first step to that really is doing a supplier engagement plan, and that's something at Mavarick that we've tried to automate quite heavily to ensure that there's a win in it for the supplier. I've worked on the supplier side. And I was drowning in requests from different systems, different OEMs looking for data in different types. And for me, I felt like I was a reporting function for my customers, which in many way I was, but I didn't feel like I was getting much from my data back. So in order to really nail the supplier, primary data is what we'd call it, acquisition problem, supplier engagement, the biggest hurdle there is what's in it for the supplier? And if there's something in it for the supplier. Then they work with you very well. Now, that could be very simply that they leave the table with a very accurate, validated data set that they can use in future tenders or contracts with other customers. But it all comes back to people and engagements, and that's before we even get towards data automation or utilising AI to help them capture this data very easily. And how do they capture the data? Because I know in a lot of cases, companies are having difficulty capturing that primary data and instead are just reporting industry averages. Yeah, that, that's very true. So we're seeing in the pharmaceutical sector, the majority of companies were early movers towards this, and they, they've adopted a spend based methodology. Studies have shown that up to 40% of an error rate can occur when they're actually measuring from a spend based perspective. So the shift towards a primary perspective very much has to be a cultural change within the organisation. Now, a lot of areas they're starting, Tom, is when they're thinking about supply chain data, manufacturers heads typically go to, I need to engage my suppliers, but they forget that a lot of data sets that are core to their business functions internally have very rich information that they can actually utilise before even engaging one supplier. So data sets such as the bill of materials, the BOM, that's an area that we focus on very early with manufacturers and showing them that if they can get their part mappings correct, if they can increase their supplier accuracy, if they can ensure that weights of materials on invoices are accurate, then they're actually going to make a lot of inroads when it comes to fixing the primary data challenge internally. When it comes to fixing it externally, we've made a lot of mistakes with companies, and companies, made a lot of mistakes trying to address this. Where we've landed on Tom, and this has actually been a very iterative process, working with manufacturers closely with their suppliers, is how can we ask suppliers for data that they already have? So for example, when we're talking about decarbonisation, instead of asking them for a product carbon footprint, that might be common knowledge to you or me, but for someone that doesn't sit within sustainability doesn't make a lot of sense. Can we ask 'em for the raw inputs of data that they have to hand? Can we ask 'em for energy usage, materials used, packaging used, and let technology transform that into a decarbonisation data set? And that's something that's working really well for us at the moment. But again, it does come back to what we were discussed prior. They have to be aware of the win. And if the supplier's aware of the win, then the primary data challenge gets a bit more simpler. But is still a big challenge for 'em to solve Yeah, because I can imagine a lot of suppliers are thinking, this is my proprietary data. You know, why should I waste my time and give up my proprietary data to you? Exactly. That's it. And some OEMs have suggested to us in the past, look, can we integrate with supplier systems? Now if you mention, can we integrate with your system to supplier? Straight away as, as rightly so, their guard goes up because potentially you've visibility into their margins and you've visibility into their IP, and it's a very uncomfortable conversation to have. The only way that OEMs are trying to incentivise at the moment is from a contract perspective. They're trying to bake it into contracts, but the conversation that's coming back, even in the automotive sector, we hear it every week with the tier ones and the OEMs is. Am I gonna be paid for this? And the OEMs are actually putting down to the Euro and cent, how much is this data request going to cost me? So that there is a hurdle to come there. Does it make financial sense for the suppliers to actually go to the effort of putting resources against it, but this is where technology can come in and this is where it can help ease the burden so that everyone wins when they leave the table. And obviously AI is often pitched as the magic bullet, but where does it genuinely move the needle on things like Scope 3 emissions. Yeah. Prior to really engaging with sustainability teams within the OEMs, a lot of, let's say sustainability professionals were coming to us talking about forecasting. But when we really dug into it, being able to forecast your Scope 3 emissions aside from your annual report didn't really make a lot of sense and didn't have a lot of impact. There was three main areas that that we're finding are really having an impact and driving value. And the first is what we mentioned around data accuracy. How can we provide them with a quantifiable method to have confidence in their data? Now we're utilising a lot of supervised machine learning and unsupervised machine learning to fill data gaps. If there's a lot of missing data, can we actually infer what that data value should be from existing data sets and systems? On top of that, what we do is we try and identify as early as possible, what's a contaminated data point, so the technology can automatically group bad data points together and inform the user why they've been categorised as contaminated. What should be done about it and how can be avoided. So if you manage to fix the data accuracy layer, then you start looking at the supplier side. So we mentioned what's in it for the supplier. One of the biggest things that we've been doing with suppliers and that AI can do is actually personalised training aids, which is something that we did not have the forefront of our mind to start with, but really helping non-sustainability literate personnel in the supplier resources to understand what they can do to ensure that they can fulfill their sustainability data requests and fulfill the requests that's actually coming from the customer themselves. And the last area around AI Tom that we're, we're probably seeing the, the biggest impact is if they can get over the data accuracy hurdle. And if they can get over the hurdle of getting good data from suppliers, it's what we call is identifying decarbonisation levers. A lot of these sustainability and procurement teams that are working on decarbonisation, very intelligent people, but they're time poor. They're spending a lot of time analyzing data, spending a lot of time gathering data. And when they have all this data, they don't know what to do with it. So AI is playing a big role at the moment in uncovering projects for them, categorising and prioritising these projects, informing them what's the financial ROI, what's the environmental ROI, and how many resources you're gonna have to put against this project. So we're seeing now the teams within our customer base are havming a lot of value add conversations about how they can move the needle and drive change as opposed to gathering data. But they're the three main areas that we're seeing AI actually driving value. As opposed to people talking about it. And identifying shaky data is that both for supplier data as well as internal data? Correct. So one of the, biggest hurdles I'd say that supply chain teams and sustainability teams are facing currently, especially with the economic environment, is trying to get buy-in from other departments and see sustainability data sets as a value add to the business. It's really, really valuable that if you can identify poor quality data in datasets such as your BOM, well, you have the attention of procurement and supply chain, and it's a core data set to the business. So AI has the ability to identify incorrect part ID mappings, incorrect supplier details, incorrect units that are used for weights and also from a supplier perspective, is the supplier. Providing abnormal quantities of a material you may have ordered from that supplier over a long period of time is a really tangible example. And how can they be informed automatically that we're not sure this data point is accurate and what can they do about it? So it goes both ways. Okay, good. And you've worked with both tech giants and manufacturers. What lessons do you think Big Tech could learn from industrial supply chains or vice versa when it comes to sustainability? It's a really good question, Tom. So from a big tech perspective, look, obviously securities is front of mind. And data privacy is, is paramount. Now, I think that there is almost a contradiction occurring currently within big tech in the sense that from an organisation perspective, their security protocols are very strict and they're very clear. They're not allowed to use for example, if you'd use ChatGPT or perplexity internally in the finance department, that's a massive red flag, right? But they're neglecting the fact that their employees are, are using these tools offline. They are using these tools potentially to do their end of week reports in Excel. So at a organisation level, the script and the narrative is correct. Look, we want to find out where data's being stored. We need to find out where data's being used, but they're actually most vulnerable from our point of view, from the people that are using these tools, which isn't a negative thing. But they almost have to realise that, and I think the smaller companies have realised it and they have embraced it, and they do have policies in place around it. But the larger tech joints, although they're pushing the boundaries of innovation, they're a bit slower to move. So I think in the next 12 months, there'll be a big shift here because their data is going into these models, whether they realise it or not. And how do you make the case internally, especially to finance or operations, that Scope 3 data isn't just compliance admin, that it's strategic, cos you mentioned, how important that is. Yeah, look, cost savings is front of mind, and that's something we take very serious at Mavarick. We need to bridge profitability with sustainability to get the attention of the financial colleagues, for example. The financial ROI of the project has to be very clear and we're working very hard with sustainability teams to help them build a business case for their Scope 3 technologies. One of the way that we're seeing that's very successful is that if they can bring value to a number of different business functions, so if they're enriching a given data set that gives them accurate Scope 3 emissions, but also shows their supply chain colleagues that suppliers are sourcing responsibly, that they're able to save money on shipping, that they're able to source alternative materials that have better margins, but better material profiles. If they can really engage other departments who are finding the financial teams are coming with them. Outside the Scope 3 Tom, like we're seeing a lot of companies and you'd read about it, they're really worried about energy usage. Right? From a Scope 2 perspective. There's an obvious reason why companies start their decarbonisation is because there's a lot of money to be saved and the financial colleagues, they row in behind it very quickly. But we find from a Scope 3 perspective that if teams can get an early win on the board from a project perspective, and typically it is with engaging their colleagues and bringing value and making their lives easier, then financial departments, supply chain departments, procurement departments, sustainability becomes a, a central fabric of their day to day because they see it as a cost saving exercise and this sustainability team can save emissions as a result. Do you have any real world examples of companies that used AI and data, right, or actually made a dent in their supply chain emissions? Yeah, so one of our medical device customers, they provide orthopedic implants in the north of France into the hip industry. Now, by utilising AI in their supply chain, they were able to increase supplier engagement by 26%. How did it that Tom very simply was, just making the suppliers what's in it for them? And it's something that I constantly repeat with them, but they realised that their suppliers were selling into, on average, eight to 10 different manufacturers with a similar profile to our customer. So our customer turned around and said to their suppliers, look, if you work with us, you're gonna have a data set that you're gonna be able to use in tenders and contracts going forward. So there's a big win in it for you. Now dealing in such a highly regulated sector, obviously the weighting that's attributed to such data points in tenders is very large, so that was a big enough business case for their suppliers to get in check and to work with them. Now, on the other side, we've seen it in the automotive industry where the same approach didn't work, and it's really shown us that not the same approach is gonna work for all different industries. So in the automotive industry, the biggest win there for them was actually showing them that they could reduce their Scope 3 emissions by sourcing materials locally. And these sourcing of materials locally, especially with upstream transportation distribution, this actually helped them reduce their emissions as a byproduct straight away. Now, there is regulations in Europe such as CBAM that's actually helping the business case of local supplying, and as a result, that's getting the attention also of sustainability teams within the OEMs. But the biggest learning, or the biggest takeaway that that we say to companies is the sector regulations dictates what you are and aren't able to do. An automotive company may be able to look at alternative materials in the next two years, but their counterparts in the pharmaceutical sector, they're not gonna have the same options. So their decarbonisation ability or roadmap is gonna look very, very different because regulation is impacting it. For people listening, Paul, who might not be familiar with CBAM, it's the Carbon Border Adjustment Mechanism, can you just give us a couple of quick words about that to explain to listeners what it is and why it's impactful? Yeah, so CBAM essentially can be viewed as a, a type of tariff. And it's a tariff that is taxing goods such as steel, and raw materials that are being brought into the European Union from other regions. Essentially, it's trying to create a level playing field somewhat for suppliers into different industries in Europe that if their counterparts in other industries are producing materials that cost less but could be more harmful to the environment, that there's an adjustment mechanism placed on top of it to ensure that locally based suppliers have a, I wouldn't call it a fighting chance, but they have a better business case when it comes to providing goods into their customer base. And we're seeing actually as a, as a regulation, Tom, that's probably the most common one that's been discussed, especially in hard to abate industries. If then you had to build, I don't know, a supply chain sustainability starter pack for a mid-size manufacturer, what would be in it? A good question. So the, the first thing we would look is saying, look, okay, can you build out your business cases and what's your best use cases here? So when it comes to supply chain, let's say sustainability, obviously within Scope 3, you have 15 categories of that that you can serve. And we typically say pick the top two, to three categories that make up the vast majority of your emissions. The majority of our customer base, it's purchased goods and services and it's upstream transportation distribution. Once they can focus on their first categories, then from there they can start building out potential pilot projects internally and assigning metrics towards them. So for example, if they're looking at purchased goods and services, one of the key metrics we see companies from a mid-size perspective, looking at initially is, okay, can we firstly get a baseline of supplier engagement, and then can we increase that supplier engagement by, let's say 15% and we want to do it within 60 days. Now, once they undertake that project perspective, at the end of the 60 days, they will look at the results. They will validate, okay, is this worth progressing further? And if it is, they look to do a further rollout. But the interesting thing for us is that, our biggest advice and what we're seeing is working is that project approach. It's going after the projects that will have the biggest impact, and in many ways will get the most attention and drive the most value internally. And from there, then they can build a proper scope, 3 sustainability, decarbonisation roadmap. But a lot of companies come to us and they want to do everything Tom, they want to do all 15 categories, but when we really peel it back and educate them, they see that the way to get their colleagues bought in and the way to get their supply base bought in is if they take a strategic approach that's based on actually moving the needle, and that's by starting small and scaling appropriately. Makes sense. It's the whole thing of eating the elephant one bite at a time. Right? It's, you can't try and do it all at once. Yeah. No, makes a lot of sense. Now, there's often a perceived tension between decarbonisation and sustainability and profitability. Do you see that? Are you seeing that trade off, breaking down, you know, where sustainability and profitability clash? Yeah, I think if you asked me 12 months ago, my answer would've been very different. 12 months ago. I think sustainability in isolation had enough merit in the majority of scenarios to self justify and not to require external validation. I have found in the last 12 months, and our colleagues are seeing it directly on the ground every day. That cost is coming into the conversation a lot earlier. And profitability is. A lot of the industries that a lot of our customers function in, some of 'em are very low margin, some are very high margin. The lower margin industries, for example, the automotive sector, that's becoming more and more of an important conversation. Some of our customers are setting metrics where your project has to show an ROI within 12 months. When you're dealing with a, a large enterprise and a complex supply chains, you know, that's an aggressive, aggressive target. But it forces you to bring value. So every decarbonisation project that we would do with a technology with those type of customers, you're on the back foot initially, but if you can bring value, then the upside is massive for the customer. But we see the tension between profitability and decarbonisation from a conversational standpoint. It's becoming more and more tense. But for sustainability teams we're just saying if you can turn this into a cost saving exercise conversation. An efficiency gain conversation. Then as a result, it's a different conversation. So it's the way they approach it and the way they actually position it to their colleagues. And, how do you do that? I mean, do you have examples that you know you can share with your customers or with us of where you've had efficiency wins or you've had sustainability wins that led to efficiency wins or profitability gains? Yeah, so, so I, I can, one of our key accounts, and this is publicly available, is the Volkswagen Group in Germany. The Volkswagen Group in Germany, initially when they engaged with Mavarick, they were actually looking at their automated guided vehicles. Essentially their fleets of robots that deliver very heavy parts around their factories. And these automated guided vehicles are battery powered. Now from that perspective there was a number of challenges from a sustainability perspective that that brought. One was obviously on the Scope 2 side. It was energy usage. It was energy cost, but then it was obviously the battery themselves, the sourcing of battery, the materials used in the battery. What was that from a sustainability perspective. But then on the efficiency side, Tom, it raised a question for us and we're through the project of, okay, well where do these charging points need to be placed? How many of these vehicles do you require? When should they be charging? Should they always be turned on? Very operational questions, right? But it brought the sustainability narrative to life. So what we ended up finding after a six month project with them initially on phase one of this, was that we uncovered the ability to reduce their emissions within that fleet by 38%. And that was simply by changing operational behavior. Things such as smarter charging strategies. Things such as ensuring that the correct product was delivered to the correct place by the correct vehicle. Very simple guidelines, which is great because having come from operations, I like simple, but almost obvious steps that you wouldn't be in aware of had you not actually looked at the data itself. So that's probably a really good example for us internally where it showed us, okay, this sustainability project, broadly as a whole actually really turned into an operational supply chain project because it was so linked to cost savings and operational efficiencies. And when we actually break that project down to our customer base, it gives them ideas, albeit the application will be different for them, but it makes them think, okay, operationally, what elements do we have in house that we can improve, but would, as a result, would reduce our emissions, or our organisational footprint? Yeah, and, and it's, it's a mantra that I've said many times on this podcast. It's that an efficiency win is very often and almost always, in fact, a sustainability win because what you end up getting with efficiency wins is more output for either the same or reduced input. It, it, it's a very good point. And on on that, Tom as well, like, and you've probably seen in the past, based on your experience, we found that personnel in departments that prior didn't have a lot of love towards sustainability, were excited to get involved in this because they saw the value it could add, whether that be in a specific logistical department. The value it could add to their day to day and how it could make their life easier and let them make more and move more product to your point. So from a people perspective, yeah, it was a big learning. And we talked a little bit earlier about AI. And how, you know, it can help. But there's also the flip side of AI, the environmental cost of AI itself. You know, compute power, energy use. Is Green AI, if we want to call it that, a real goal or is it just another buzzword? It's a really interesting one and it, it's something that we're finding we're getting more and more questions about in the last six months. The questions are tend to come from two groupings within companies. The first is sustainability teams. They're just a bit confused about how they report on it, and some studies are showing potentially that a new Scope 3 category will be created just for AI to track it. And the others coming from IT teams. So IT teams that are left managing these monoliths in, in many ways. How is that gonna scale with our infrastructural roadmap? Now, one of the, the biggest environmental issues with AI is water usage. So these models are being run on very powerful hardware within data centers, and it's taking a lot of water to cool down this hardware. It gets very hot. To ensure that they stay within their operational temperatures. The next potential environmental issue around AI is, as you be well versed in, is the materials used in this hardware. So a lot of this hardware actually contains a lot of very valuable rare materials that are difficult to mine and bring their own challenges when actually constructing GPUs. And the most common, or the most spoken about in the media is energy usage. These models require a lot of energy to actually train. Now, when you talk about Green AI where this all stems from is these models have a lot of things called parameters. And these parameters can be viewed as settings, and the settings have to be placed in a certain way that give you the best output of the model. So the more parameters or settings you have in a model, the more time it takes to train. The more energy that's used, the more footprint it has from a Scope 2 perspective. On the flip side, when people, for example, are using publicly available LLMs every time they ask this LLM a query, that model is having to use more energy to respond to their query, and the footprint is growing and growing and growing and growing. We're seeing a number of things that are being done in industry to help combat this. The first is that we're working with a number of data centers around looking at things such as smarter workloads. Think of it as running your washing machine at 2:00 AM instead of 2:00 PM when the demand for electricity is a lot less. There's a lot of smarter things going on around it. One area, Tom, that we're seeing a lot around is surrogate modeling. This is the idea of using an AI model that has less parameters. So it's cheaper to run, it uses less energy, but it gives you the same output. It'll take you a bit longer to get there, but it gives you the same output. But the biggest thing that we're we're saying to teams, or educating them on the moment around AI, and it's something we take very serious at Mavarick, is responsible computing. So do you really need AI for this project? And if you don't, that's brilliant because you're going to be able to interpret how your method made a decision easier. Are you using a small model? So when I say small, are you using a model that's fit for purpose? Do you need to use a model with billions of parameters? Probably not. And the third thing to look at is where is this model being run? And by run, I mean if it's a web-based model, are they using green energy to actually drive this model. So for example, in France there's a really publicised story where nuclear energy is used to run certain models and as a result there are emissions of these models are decreased massively. So it's a conversation in itself that is happening. There are things that are taking place, but I think in the next 12 months people are gonna become more and more and more curious about having the answers of how environmentally friendly are my AI methods. At the moment, it's very much of, it's doing the job, it's helping me make efficiencies, it's helping me save time, but the needle is going to move. So there is a lot of work being on. It's something we take very serious at Mavarick, and it's an area that we see as a kind of standing out in the market as well, is that we're quite early movers in terms of adopting it into our technology. Right. So Paul, are you coming across anyone who's using the likes of chasing the moon? That kind of strategy and it for people who might be listening who are unfamiliar with the concept, the idea of chasing the moon is that you are using nighttime electricity no matter where you are in the world. So. obviously, the likes of OpenAI and Microsoft's Copilot, et cetera, they're global companies and they have server farms all over the world. So we could be here in Europe. It's the middle of the morning, we don't care when we put a query into any of these, whether it's happening locally or whether it's happening on the far side of the world. And if it's nighttime there, the electricity there, nighttime electricity is obviously going to be greener. So is that a concept that you're seeing any of the manufacturers or any of the people you're dealing with using? No, Tom, I haven't seen anyone speak about that, but it's, it definitely sounds very future thinking. Alright, well. Maybe it's something that we can, we can start investigating for people. It could be, it could be, I don't know if anyone's offering it yet. I haven't heard of any of the AI AI providers offering that as an option yet, but I think as more and more people are talking about the footprint of AI, it's something that could be a differentiator for some of these large AI providers. Anyway, we'll keep going. How can companies do you think future proof their emissions strategy, knowing that regulations, technology, and supplier landscapes are all shifting really, really quickly? The easiest way to do it is actually at the design stage. And we hear this constantly, right? It's like, why are we retrofitting products that were never built to be sustainable, to be sustainable, and why aren't we focusing on our R&D and, and building for the future? Now, in many cases, the problem is a lot of design teams don't talk to operations. There's a disconnect there around that. That's what I, I would say, Tom, is that like in, in reality people are going to have to look at their existing processes. Now, when you talk about existing regulations, if I bring it back to the, like of the pharmaceutical or, or medical devices, if want to, for example, look at a, a spinal screw and you want to reduce the product carbon footprint of that screw, and you look at your internal processes, it involves, okay, changing machining approaches, looking at different materials, looking at the way it's packaged and cleaned. Well, straightaway there's big issues around revalidating processes that take a very, very long time and almost upsetting the industry norm. So if you're looking at implementation changes like that, that drive decarbonisation, you've a massive uphill battle and it's, it's very, very unlikely that you'll actually be able to drive change that way. So that's why we always are telling companies, okay, look, you need a long term decarbonisation roadmap. What does that 10 year goal look like where you can work with, let's say, regulators to ensure that you can make more sustainable products. But in parallel to that, you need your short term, and that's why it comes a lot of the time to optimising the way that goods arrive on site. To optimising the way you're using energy to look at renewables, you're using on site to work with suppliers to look at how they're sourcing goods and what you can actually change. But the design stage, as I said, is key at the start. We are seeing companies in the plastic sector, primarily in the UK and North America, that are really investing heavily to design more sustainable products because they're getting bombarded with request from customer base because in many ways they've been deemed the dirty industry, whether they're fairly or unfairly, and as a result, it's forcing them to change. But I think a lot of good is gonna come from them in the next 24 months. And looking out five years, what does good look like in sustainable supply chain data and emissions tracking? You know, what will we be doing differently? I think primary data will become table stakes, if I'm being honest, primary data will become table stakes and good will very much look like those that are taking action. Those that are actually being very transparent with their customer base and providing proof that they're not only talking about change, but that they're, they're making change. At the moment, in many ways, a lot of larger entities especially, their sustainability reporting in many ways, can be viewed as somewhat of a marketing exercise. But those that are showing tangible examples of how they're reducing emissions, whether that's through greener materials, whether that's through greener methods of transport, whether that's through looking at the use of sold goods, especially in the automotive sector. If we look at the exhaust emissions themselves, how they're gonna address that, there's massive potential there. But primary data will become table stakes and I'd be very surprised if spend based data is still in the conversation. Great. And for supply chain leader listening today, feeling overwhelmed by Scope 3, what's one small, practical first step you'd recommend they take tomorrow? Map out their data. So sit down, get the key stakeholders and do a data mapping. Identify where you're data rich, where you're data poor, and ask yourself where you are data poor, is it worth gathering data in that place? Often it's not, and that's fine. That's totally fine. But once you have an understanding from a data perspective where you lie, then your potential avenues for change become a lot clearer and it also becomes clearer what other departments you need to engage to ensure that if they are the custodians of said dataset, they're aware of what you're doing at a very early stage. So if I had to give one piece of advice, that would be it. Okay. And imagine you're pitching AI driven carbon accounting to the Dragons on Dragons den. What's your 30 second hook that wins them over? I think for them it's, we can do more but do good and we can make more with less. When we actually communicate that, especially to C-suite teams within the manufacturing environment, let alone dragons themselves, they become very curious because we have to hit people in the pockets, but in a positive way. And we, we take that very serious at Mavarick, and it doesn't always sit very well with sustainability personnel, but profitability is key if I'm being totally transparent. But if you can really show teams that look, you can make more by doing good, and by using less, then in itself, a lot of people, at the very least, will say, okay, how can we move the needle here with AI for decarbonisation in our supply chain? And, left field question, if you could have any person or character, alive or dead, real or fictional as a champion for AI based Scope 3, sustainability reporting, who would it be and why? That's a very good question, Tom. That's a very good question. I think we've, we've enough experts or inverted commas experts floating around, so it would have to be someone that would be able to shine a light on this I am a, a massive football fan, so for me B ill Shankley, the former Liverpool manager, I'd have to bring Bill in because Bill would be able to build it from the ground up and get the people on side and show that it's a societal problem and not only a problem they read about in the news, so it'd have to be Bill Shankley. Very good. Very good. We're coming towards the end of the podcast now, Paul. Is there any question I didn't ask that you wish I did or any aspect of this we haven't covered that you think it's important for people to think about? I, I think you, you did touch on it, Tom, but just to reinforce it from the AI perspective, what we're finding is, look, a lot of teams, they don't know where to start with AI. But it's just to tell people that it's normal, and everyone is in the same boat. We're finding that a lot of companies, albeit externally from a marketing perspective, they are AI first. But when we talk to people on the ground, there's a disconnect and they haven't got started with the technology, but that's okay. But what I would say is that people need to start looking at it, at the very least and taking a step back. And it's not always the, product features that benefit from AI, I know you and I have spoke, but in the past, it's internal processes. So, where are you spending a lot of time? What's your high friction points? And I think if supply chain personnel and sustainability personnel can take a step back and look at that, then they'll make a lot of inroads in their day to day and they'll enjoy their day to day a lot more because they'll be bringing value. Paul, that's been fantastic. If people would like to know more about yourself or any of the things we discussed on the podcast today, where would you have me direct them? Yeah, so the company website Tom can be found at www.Mavarick.AI and that's Mavarick with an A, so M-A-V-A-R-I-C-K. Also available on LinkedIn. And they can connect with me personally on LinkedIn also. Tremendous. Paul, that's been fascinating. Thanks a million for coming on the podcast today. Thanks, Tom for having me. Really enjoyed it. Have a great day. Okay. Thank you all for tuning into this episode of the Sustainable Supply Chain Podcast with me, Tom Raftery. Each week, thousands of supply chain professionals listen to this show. If you or your organization want to connect with this dedicated audience, consider becoming a sponsor. You can opt for exclusive episode branding where you choose the guests or a personalized 30 second ad roll. It's a unique opportunity to reach industry experts and influencers. For more details, hit me up on Twitter or LinkedIn, or drop me an email to tomraftery at outlook. com. Together, let's shape the future of sustainable supply chains. Thanks. Catch you all next time.