Name: Digital Supply Chain Transformation at Scale
Uploaded: 2026-03-30T16:16:07.881Z
Duration: 1 h 8 s
Description: Digital Supply Chain Transformation at Scale

Transcript for "Digital Supply Chain Transformation at Scale": Before Amazon Business, buying for work was chaotic. Now it's easy to find products from thousands of suppliers in one place. Save on every type of purchase from individual items to bulk orders. At any time, you can view your spending on prebuilt, easy to use dashboards. Plus, you can free up cash flow if you choose to extend payment deadlines and view and approve your team's purchases easily. With Amazon Business, things just got a lot more manageable. Hello everyone and welcome to our webinar in partnership with Amazon Business. Today's webinar is part of an ongoing series of Amazon Business, where we explore the big issues impacting procurement, supply chain and manufacturing. You can catch all the webinars in this series on the Supply Chain Digital website or on our sister brands Procurement and Manufacturing. Before we get started, I would also like to mention our Procurement and Supply Chain live event series, which in a few weeks time will be arriving in Chicago. My name is Ella Wilkinson from Supply Chain Digital, and today we will be looking at digital supply chain transformation at scale. Supply chains across industries are undergoing major digital transformation, becoming more connected, intelligent, resilient, and increasingly sustainable. In today's session, we'll explore how organizations are adapting to this shift, scaling technologies, and how they are using data, AI, and automation to drive performance. Joining us today are two incredibly placed experts to help us navigate this important topic. We will welcome Simon Jannig from Integrity Next, who has dedicated his career to helping firms gain deep visibility into their supply networks to become more digital, transparent, and sustainable. And we are also joined by Ryan Gliani from GEP, an expert advising enterprises on how to move beyond legacy systems and embrace unified platforms that leverage AI to drive smarter, faster, and more resilient decision making. So before we jump in, I would like to welcome you both and get you both to tell us a little bit about who you are, your respective companies and why you are an expert on this topic. So Simon got actually on my screen, so I'm gonna throw to you. Perfect! Yeah, thank you Ella. Thanks for the introduction. So as said, my name is Simon Genich. I'm one of the co founders of Integrity NEXT and we founded the company ten years ago. I can't believe that it's that long ago with the mission to set up one of the largest supply chain networks in a sense in the world, where companies globally can share all the kind of practices around human rights and labor, environmental practices and so on, to get more visibility into the kind of sustainability performance of supply chains in general. When we started ten years ago, I think a lot of hesitation has happened towards that kind of topic in general. And I think over the time companies have pretty much understood: the more I know on my relationships and my supply chains, the better it will work out for me also from a resilience perspective, from a performance perspective and even now quite a big and important topic is companies also have understood there's a significant kind of margin contribution if I can work with better kind of, you know, less resource intensive kind of materials in my supply chains in a sense. Yeah, I'm happy to speak about that today and talk about how we help organizations to transform their supply chain to be more sustainable. Well, it's a pleasure to have you here, and we'll also talk about that ten years, so the ten years you've been in business and how rapidly it's changed over that time. If I throw it to you, Ryan, quickly. Yeah, thanks for having me. So I'm Ryan Gianni. I'm part of GEP, part of our supply chain consulting practice based out of our Chicago office. I focus primarily in more of the consulting process improvement space, and I look for opportunities and leverage technology and now AI as an enabler to improve business outcomes. Now, I've been with GEP now for about five years, and I've done this at the big four firms for the last fifteen years. But at GEP, just a little bit about them, or us, we focus not only in supply chain process improvement and business outcomes, which is the main thing for our clients, but then also, we provide our own proprietary software in the supply chain space. And so, really, we've found over the last several years now is that the further integration of the two spheres within our business, both services and software, because the supply chains are becoming increasingly, as they've always been, but really lately, increasingly digital. And so, all of our practitioners, not just our technologists, but our business practitioners are leveraging technology and AI and those outputs in order to increase value for our clients. And so, I'm really excited to talk about this space today. You're absolutely right. AI is rapidly changing everything we do and it's changing not only five months ago, but five minutes ago. So it's new every single minute, really. So it's a pleasure to have you with us as well, Ryan. So let's actually start today by looking at that evolving supply chain. So the supply chain we see today versus the supply chain we saw ten years ago is vastly different. How have you seen digital transformation reshaping these supply chains in the industries that you've worked in? I'll stick with you, Ryan, because you're still on my screen. Sure, sure. I mean, so this is, you know, kind of an interesting question. I mean, from our perspective, there's been a shift in supply chain towards digitization for quite some time. I mean, you can go back to like mainframe computers and now to where we are today with modern ERPs. I mean, the last fifteen years of my supply chain career has been focused on driving digital transformations within the supply chain and overall business. These have typically been of commercial off the shelf, SaaS driven providers to transform a planning process or logistic process. And it typically takes years to do this kind of standardization. And it does achieve a lot of business value, and we've seen that. So that's already been happening. However, lately, what we've seen with the recent shift towards AI is that companies are focused more and more on a micro scale and more speed to value, where they can drive those efficiencies within their existing architecture. Not looking for an ERP implementation. They're looking for what kind of value can I drive with what I have and how do I achieve various efficiencies and cost takeouts? And so where good companies are leading is to really automate, you know, less in the micro space. So, hey, can I automate this small little process over here? But rather looking more holistically of how do I build platform on top of AI? And how do I link AI agents to AI agents to AI agents versus human to AI agent and then human again? And so that's really where we're really focusing on with a lot of our clients today is how do we kind of scale the AI? Because we're already incorporating AI today. We've got myself where I'm going through trainings of how do I create my own agents? We have technologists doing that. So, how do we make this more robust and more scalable? And that's really what we're focused on today. Very true. If you hadn't got on board with Facebook in 2007, then you're behind in an advertising company, and if you didn't get on the digital train, then where were you? But obviously now AI is just ramping up that speed. Is this something that you're also seeing, Simon? Yeah, I mean, I look at the supply chains in like different ways, right? I mean, we cluster typically the supply chains into various different kind of buckets. We say on the one side we have discrete manufacturing supply chains, have been extremely become more complex in the sense that regulatory requirements have also kind of challenged the organisations to comply on not just getting the things to the right place in the right time, but also kind of being able to bring the products over the right borders to a certain time in a sense to comply with the relevant regulations and having here also quite a vast majority of regulations that companies have to focus on. That's like one of the things we see in particularly in some of the process industries like chemical and pharma industry, also post COVID, we have certain aspects of kind of decoupling and kind of processes where entire kind of supply chains are basically been rebuilt as one element. And if you look into like consumer goods supply chains, you suddenly see regulations from the European Union and so on. The companies suddenly have to trace down every individual product back to like the plot of land where a certain tree was basically been planted in a sense to understand like, was there a forest before 2020 or not, right? And so I think the challenge with all those different requirements on the one side is that the core of what you do as a supply chain professional is not any more just like looking into where does my data flow, where does it crosses borders and like make sure it arrives at the right time. I suddenly had like this kind of vast majority of additional regulatory constraints and reporting due diligence aspects that I have to fulfill in line with that. And I think also, as Ryan said, I mean, one of the key benefits of AI at the moment is that it suddenly helps the organizations to first of all predict to a certain extent and make probabilities to where do potential risks actually sit in my supply chain. And it also helps to get access to a much larger amount of information and to the kind of consumption of those informations, because a human brain itself might not be capable of like filtering across, like, I don't know, 1,000 regulations, at the same time understanding which goods do I have in flow and which goods will basically cross the border at what time and where do I have to be compliant with what regulations, right? So I think the complexity has increased and also the geopolitical framework in that sense with kind of more, I would say, gatekeeping and so on has kind of caused this tremendous need for information and the right analysis of the information. So, yeah, I mean, that sense AI is, I would say, to a certain extent, a huge advantage for the organisations to get better hold of their entire operations in that regard. So, do you now think that we're asking supply chain professionals to wear almost too many hats? They've got to be financial minded, they've got to be sustainability minded, they have to foresee problems before they arise. Do you think we're asking too much of our supply chain? I think the point is it's not about if we ask too much. I think the challenge, what we've seen over the last few years, is that if you have not integrated it into one function, you know, it often just recited as kind of silos next to each other and then the benefits of each of those functions has not been completely operationalized. So, I would say with the possibilities of AI bringing the information together and also helping in the reasoning and helping in the kind of decision making processes, there will be the chance that it is capable to have it in one central function. And I think, honestly, to be also quite fair, it is the only way in how to get hold of it, rather than having individual silos looking at individual things, not connecting the dots, not understanding that, for example, if in a certain supply chain I potentially kind of increase recycled content, I can increase my margin by 10%, right? And that function will become more holistically and AI is the one that can enable them. I would say, if you would have asked me like two years ago, I would say yes, for one person it might be too much to ask with the capabilities now. I would rather say there is even more potential scenario building information that might make sense on their tables as well, yeah? So I think one of the biggest issues really is those silos not being able to talk to one another. How are you seeing the most successful firms bridge that gap? I mean, it ultimately comes down with a clear structure, right? I mean, clear data structure. I mean, what we see in most discussions with companies is have now the understanding that they have some sort of a data lake on the bottom and then on the top they have some sort of kind of agentic infrastructure, which basically doesn't talk to their data points. And I think we all know, if you have worked with the different kind of AI models to a certain extent, it also comes really down to the fact that you have to contextualize a certain set of information and not just kind of make the access of the data too broad, right? So you will need a framework in the middle where you say: Well, I still need to understand what is my purchasing data, what is my vendor data, what is basically my product data. And you need to still have a kind of system in between the big data lakes that makes sense out of that, who then basically connects to the agents on top of that and kind of digests the information and distills the information that the agents are capable of working with it, right? And so, I think most companies are now at the point where they understand I need that and basically they need to have some hand holding exercise at the moment where we need to guide them and explain them how this infrastructure looks like, because it will ultimately come down to the success if that has properly been executed. So, Ryan, from a more advisory point of view, do you see when, what do you see the most successful firms doing to bridge this gap? Yeah, I mean, it's a complex question. I mean, you know, we are seeing the kind of silos inherent now as the AI boom is just kind of starting to take shape within companies. You know, they're using AI within their specific functions to do specific tasks, to mine specific data, to do specific analyses. But, you know, the good ones are able to connect across the different functions to provide those more holistic strategic decision support mechanisms versus just individual ones for a finance function or logistics function. So that's really where clients are starting to ask us, how do we actually do that? And how do we kind of get to that level of scale? Because we're seeing, as I mentioned, we're seeing a lot of individual agents being able to run individual analyses and individual reports. And that's adding a lot of value. It's creating efficiencies and effectiveness. You had an individual and a human doing this in the past, doing a lot of analyses, doing a lot of data crunching. Now you can have an agent that does that for them, creates a lot of and generates a lot of efficiencies. So, your companies are happy with that. But then, as I mentioned, the more strategic one of thinking, what's next? How do I connect this and create a more holistic decision support mechanism so that I can make decisions not only on logistics impacts, but also on planning impacts, on sales impacts. And so kind of running the full gamut of my whole business. And so that's really where we're seeing a lot of questions on how to do that. And so, you really kind of have to think more strategically on not just those individual agents, but then how do I build it on an AI first platform? And so, just a shameless plug from my company at GEP, we've kind of now converted our software platform for our supply chain, both core supply chain as well as procurement, we've now converted to what we call GPQI and that's built on the agentic platform so that we can enable those agents to work across the different functions. And so that's what we're trying to invest in. We're kind of demonstrating that for our clients. I think that's the next frontier is to try and get those insights across the different functions and get those efficiencies across the different functions and not just within individual parts of the value So I mean it's very clear that we have incredible technology now. We have, you know, automation coming out of our ears allowing full visibility of our complete supply chain. But in 2026, do you think that's even enough? Do we now expect that agentic AI to come in and predict disruptions before they even happen? I mean, I think there is the certain kind of there's certain belief sometimes in the market that you have to basically be able in predicting anything. And I think what AI does pretty well is it is probabilistic, right? Those are all probability models, right? And with more information, the probability will be significantly higher. I think if you look at, in particularly supply chain optimizations, and there are a few also elements, if you think at mathematical optimization and decision making support, that AI basically cannot do. I think it's ultimately a combination of probabilistic scenario modeling capabilities with then purely mathematical kind of optimization models that over the time will provide the best insights information. Will we be able to predict in a sense like if there will be another ship in the Suez Canal and everything will be stuck? Most probably not, you know, there will always be those kind of black swan events that no one can predict to a certain extent. But where we have become much, much better is in the combination of different kinds of data elements to understand how well are my partners in my supply chain, where do I have so called critical knots, so we can with a relatively high probability, basically say, well, if I'm buying this or that product, I have a supply chain point X, Y, Z at a lower tier level, that can cause me a potential threat, because ultimately my vendors and myself were buying actually at the same kind of critical knot in a sense, right? So we're getting there, but I mean, to get to the point of like the real prediction to a certain extent, I would say that's in a probabilistic world rather difficult. We will get close to that, but I think we will still need to live with the probability of it. Maybe we'll never be able to predict those black swan events, but how do you think AI is making us more resilient to them, so that when that happens it's less of a disruption? And I think that ultimately comes down to just the pure mass of information that I can contextualize. And I think this is ultimately if you have a certain kind of supply chain and that's I think all about the context. So if you have a manufacturing, discrete manufacturing or a specific kind of vertical specialization And then with the relevant kind of contextual knowledge, you all know where the critical knots are. And with that, you can more targeted focus your observations into those critical knots and try to get as much as possible information around those critical knots. And that ultimately will help you to get significantly more resilient in a sense. And I think what is very interesting also what we've seen in discussions with our clients is that resilience also is not necessarily only focused around kind of risk events. It's also about, in a very, very fast kind of changing economic environment driven by the AI transformation as well, it's also about understanding who in my supply chain are the most innovative, basically, companies in the sense that master the transformation for themselves in the fastest way. So it's also about who will be my partners in the next two years moving down the road, because they cope not just with the environmental factor, but also with the transformative kind of forces from the outside in best possible way, right? So gathering as much as possible data on your vendors, on the supply chain, I think is the key element, because we now have the tools to make sense out of it. So, let's remain talking about our vendors then. Ryan, actually, as we scale our digital tools, are we potentially overwhelming suppliers of new platforms and data requests? And how do we then implement advanced technology across our global supply chain without breaking those very important relationships? Yeah, mean, it's a good question, and I think that that's historically been the case throughout the supply chain. Every time you change vendors or you change softwares, you're asking then your suppliers to then integrate with another platform, another data stack, and data requirements. And this takes time to kind of convert their not only their technology, then their internal processes to be able to meet that requirement. I think what we're seeing is twofold. I think, obviously, if we stick with AI, AI has been helping with this to be able to, from what I've seen, transform that data structure such that the suppliers and the vendors have to do less in order to engage and to transform there. And so, that's what we're seeing, specifically in the AI front. But then second, there are types of software that we like to leverage that are multi enterprise suites whose job it is to integrate with multiple parties and multiple partners. Think of the Ariba networks or things like that. And so they're able to do that and they're able to obviously leverage the latest in technology, AI and other aspects of things in order to integrate there. But yes, I mean, this will continue. Mean, the access to data is required in order to make those predictive models effective. And that's, you know, I think what Simon was mentioning is that ultimately these are predictive models based off of data inputs. The value then has been there for a long time. We've had machine learning mechanisms for demand planning for well over a decade. But now I think the change in that is that not only do we have historical data that we can base those models on, but now we've got other aspects of unstructured data, natural language and emails that they can mine data on much more quickly and effectively than humans could. And so the ability for these companies to access that type of data will kind of differentiate themselves. And so, back to the supplier and vendor requests there, we want to have access to more and more of that data. And we're seeing requests from our customers working with their vendors requesting more and more data. And so, we're able to then leverage the new technologies of AI and things like that to of quicken and lessen the impact with the supply base so that we can comply and get access to the data, but not, you know, kind of, overly burden them to have to transform their business substantially to be able to provide that data. So, I mean, as that technology is becoming more advanced, we are changing the capabilities of our supply chains and that's happening every single day. So this is questions for both of you. What are some of the most useful ways that you've seen AI being utilized by your customers? I can start, I think the use cases are many fold, I would say. Some of the key use cases which we have seen where there's been the most valuable is in the sense where you need to kind of make predictions in the sense of where do I have potential critical knots, which would not have been possible, I would say, years from now. It would not have been possible to kind of look into 3,000,000,000 trade information to understand who does business with whom at a second tier or third tier and identify kind of critical knots. I think that kind of thing is it's not the holy grail so far, but it gets close to like where you pretty much understand and can very highly predict who does business with whom at what lower level and what potential risks does basically appear for me at that kind of lower level. So that is one of the key elements, I would say. And then the second key element is particularly in operationalizing some of the kind of regulatory aspects in a more value driven aspect. I mean, I mentioned it, you know, that companies understand I let my products being calculated from a product carbon footprint perspective and now I understand, well, that has a huge impact on my total, first of all, maybe margin contribution, but it also has an impact on my personal kind of footprint in a sense and combining those data points with each other and kind of eliminating the silos that where we see is the highest, I would say, contribution in a sense, combining the information together. Okay, and Ryan? Yeah, and so there's specific use case here that we were able to do, sticking on the realm of inbound supply with our vendors. We're trying to predict when those vendors are going to deliver to the warehouse, to the hub, and then to the manufacturing facility so that we can plan our production more quickly for our clients. And so, there's obviously, historically, there's promise date from the supplier and when they're able to deliver it and then there's shipment notification. So those are all provisions that the vendor provides. But those can be error prone based on master data, standard lead times and things like that. So, it's when can we actually expect delivery so we can plan our production down to the minute. And so, leveraging AI, we're actually able to not only look at the history of this and just look at when the supplier says they're going to be able to deliver and when they've actually been able to deliver. We've got thousands of shipments to base it off of. That's a great input. But then also there's a lot of external discussions that are happening with the vendors and with the carriers. This goes through email traffic. This goes through Microsoft Teams and messages that in that way we're able to leverage AI agents to go through and trawl through all that data to add and layer into that prediction model, not just the history, but the conversations that have been happening and mine through all that basically natural language to better and more precisely predict when those deliveries are going to occur. So we were able to kind of layer in that on top of just normal kind of machine learning to add to the prediction model. And so that was pretty effective. Now intuitively that makes sense. Know, there's a lot of email traffic going back and forth with the carriers and so you have a lot of data. Now it's in a whole bunch of different unstructured data and a whole bunch of different people's emails and outlooks. And so, you know, we provide an agent the ability to kind of control through that. We're able to kind of gather that information very, very quickly be able to analyze that and incorporate that into the model. So that was a use case that we recently did and it's been quite effective. And it sounds so useful. Simon, you mentioned regulations as well earlier. I'm wondering, do you use a similar model as that to stay on top of your regulations of sustainability, your compliance requirements? Because that's a massive amount of digital verification there. Do you use AI to help that? A 100%. I mean, we are screening thousands of regulations on a daily basis to understand, like, are there changes? Are products like basically our customers are having? Are they basically affected? And if you think about some of the regulations, if you think about PFAS or if you think about REJORS, it's a chemical regulation, right? You're a big kind of don't know, you're a big kind of electronics manufacturer and you have, I don't know, 500,000 products you're typically buying on a yearly basis, I mean, you have no clue which of those products basically have to be compliant with what and how a change in one of those regulations has an effect on each of those products and where are those products at the moment from a shipping perspective, right? And so we use AI in our product compliance and sustainability platforms to basically just tell the customers: Look, there you have a change. Something has changed. You need to basically be careful in understanding that because potentially, I mean, as a next step, there are some regulations, you might potentially not be able to import it anymore, right? And then you a supply chain serious issue, not because like there's the ship in the Suez Canal, but because it's just stuck at customs, because you weren't ready for the relevant reporting. I think to be also quite fair, mean, a few years ago, we had like an entire huge team looking into that. Now we have rather the team focusing on understanding how to transform this into kind of use cases within the products themselves than rather being just needed in a way to manually kind of freeze through the screen through all those regulations and provide those information to the customers. And I think that's massive. I mean, that's not just for us internally a huge gain but also a huge advancement because you'll always stay ahead of what is happening and changing, you know. Just having a huge workforce to never forget anything. So, I mean, it's very clear that we have the tech but do we have the culture? What are the most effective strategies that you've seen for getting global teams to trust and actually use automated tools for making decisions? And I think it all comes down to governance, right? And to kind of traceability. I think still, you need to still be capable with the applications that you're implementing, that you'll always be capable on why which decision has been taken. I think the governance aspect and the trust aspect is, I think, is key element where you kind of be able to convince the people understanding it. And then I think that also the key point is you need to have specifically trained models for the different kind of use cases. I think what people often have as a kind of experience, if they kind of work with JGPT or Claude sometimes or something, just use it as a kind of sense of a better kind of Google search, you know, and they say, well, I'm just typing a question then it gets a kind of see me great answer out of it. And so what really helps is educating and explaining them how contextual knowledge basically will also increase the output and how traceability allows you to basically get back to the understanding of it. And I think it's also culturally the discussion around that people are afraid and losing their chops and I think it's more about: look, there is a different way in how you can spend your time and it's typically more challenging for you as a person individually if you're not kind of jumping on the AI train, you know, but rather you do and you see how it can enhance and enrich your kind of daily business doing in a sense, right? And so I heard an interesting, I think it was Chen Soo Huang who said, you know, if you basically literally have someone like an engineer that said, well, I'm only using my paper and my pen and then not shifting to like the CAD tools that have been introduced, you know, decades ago, you know, it's the same kind of way in how you need to adapt and work over there and, yeah, less fear it to a certain extent as an individual, you know. But that's for sure, it's challenging, but it's all just, I think, working out with honesty and a lot of education and hand holding. It's like if a newspaper stuck to print and never went online. Ryan, is it all about trust for you as well? Yeah, and specifically the way in which we're driving that at GEP is where we have a technology part of our business that drives software and technology and digitization enhancements. That's a core part of our business. And we also have our consulting and managed services side of our house. What we've done and what we've initiated is we've gotten the whole business involved in creating AI and AI agents. And so even myself, I'm going through trainings to try and create my own agents. So we're providing ownership and we're empowering the teams, not just in the technology side, but our services side to have ownership over AgenTx AI and other parts of AI so that they not only are involved in that, but then they also develop that trust towards it. Because we have I mentioned we have a lot of managed services and business process outsourcing. Do. And so naturally, the incorporation of more AI affects their job. And it could put them out of a job, theoretically. Now, we want to continue to push towards automation and efficiencies and acceleration of these things in supply chain. That's where it's been going. That's where we'll continue to go. And so with this mandate, which we started at the beginning of this year, we're empowering those folks to be involved in the agentic AI revolution. So that not only do they develop that trust, but then they also are building their own skills. And they're part of that transition from paper and pencil to CAD, just like in years past. And so I think that we're seeing a lot of enthusiasm with this internally. And so it's something that we've actually started to discuss with our clients that this is kind of starting to work and starting to stick. And so it's process that could potentially work internally in your businesses as you start to embrace these new technologies and this new AI. Fantastic. We are running out of time, so I'm going to move us on to our final topic here, which is data AI and automation driving efficiency and performance. So the first question, you're both going to shout yes at me, but can a digital first supply chain truly optimize for cost, speed and sustainability all at once? It can, but I think you have to have the right setup there. Absolutely. Yeah, that's the short answer. I mean, I would agree. It's possible, you know. Right now, we're just seeing these kind of individual agents working in silos. But if you start to build that across all those levers and all those platforms and all those functions, it's definitely possible and definitely feasible. So let's look specifically at autonomous procurement. So we hear so much about it, but in a scaled environment, how much of that decision making can we truly hand over to AI? And where do you think the human in the loop is still imperative? If I stick with you, Ryan, for now. I mean, I think that it's already happening. As part of our teams that are supporting and acting on behalf of our clients for their procurement organizations, we're already incorporating that into the processes. So we're taking more and more humans out of the loop or human decisions and activities out of the loop as we layer in more agents from PO creation to chasing and all the activities from all the way down to procure pay. So we're already kind of seeing that. I think that slowly we'll get into that. The challenge and that we're still seeing with the need for humans in the loop is more relationship aspect of things. So the relationship aspect with the vendors, there's still a value in that human interaction and discussion in order to inquire on different aspects of where's my stuff, or just to discuss around different aspects of the needs of their business. I think that as we move forward, businesses will start to embrace that interface with the different agents. But right now, we're seeing still that relationship value is still there and still needed with people. And that's where we kind of empower our folks on the procurement side and category management to kind of maintain their focus on the value that they can provide as resources and kind of give more of the analytic components of things to more automation and digitization aspects. So do you foresee a future where agents are actually governed by other agents? That's the singularity of it, you know, where you have machines creating other machines and managing other machines. I definitely perceive that. I mean, that's definitely where I would see this going, where you have a hierarchy of agents, and hopefully you have a human sitting at the top and not another I machine sitting at the mean, that's the holy grail there. You know, we don't want humans to be fully removed from the scope and where are we actually needed. You know, but, you know, it's a scary thought. It's been the subject of many movies. And I don't think it's out of the realm of possibility. Right now, we're not there yet. But that's really where we're trying to drive towards that is having a human and manage a bunch of agents and that's their team. And then those agents manage other agents just like a normal hierarchy. Again, like I said, we're not just yet there yet, but that's the vision and that's where we try to have those, at this point, more philosophical discussions with the clients. Yeah, it's an interesting topic. And we could talk about that topic for ages, but I will move us on because we are running out of time. But Simon, then, what are those practical digital steps that you think a leader should take to ensure that their data actually triggers a smarter procurement or logistics decision? I would separate in a strategic and an operational bucket, right? And I think on the operational side, you do a lot from a kind of automation, from a kind of fully kind of more autonomous kind of way. And I think then you have, as also Ryan said, you have the entire strategic aspect and that's relationships. That's basically, in a probabilistic world, to still decide on is the probability basically been based on the right set of data, you know, in a sense, and do we believe in those scenarios or not? And I think that's ultimately the human touch on that and to a certain extent, right. So, I think separating that into those two areas and then really understanding, like, are my key processes in a sense and how can I easily automate some of the operational processes and then understanding and clustering it where on the strategic layers can it provide me a kind of strategic edge to have more information available and then take the decision on that more information myself? I think that's the key element. And then I think for sure, what Ryan also said, governance at a certain point has to be built into the agents itself. Yeah, because I think at a certain point, even if you I mean, we write a lot of code, you know, even you have to build the security within the agent writing the code to a certain extent to be secure and set up to a certain extent, or at least to have minimal security being built in, right? So, but I think that separation and being clear of what is a human kind of capability and a human kind of quality and what is an agent capable of doing, I think that separation sometimes helps in a sense in driving that successfully. Brilliant! So we have a few minutes left and I would love to get some of the Q and A questions done, so I'm going to ask for very, very quick answers from you both. So our first question here is: what do I do if I have loads of different legacy ERPs? What is the one practical step to cleaning up that data? No, I think, I mean, what many companies basically do is they first start kind of building out some sort of a kind of a data lake, but within this data lake also even having kind of a dedicated kind of structure and then using applications like us for dedicated use cases that, again, even more structure and enrich the certain kind of elements of information on top of that. I think that's a necessity to a certain extent. Yeah, you will need to defragment it in another way, more content specific, and then plug it back into the Nagenetic structure. That's my take. Now, Brian? Yeah, it's a similar answer. Mean, you need to extract the data first you can get those insights across the different ERPs if you've acquired a company or something like that. Ultimately, then companies go down this long line of integrating the ERPs over time, but you need to get access to that data. And you can leverage different types of machine learning or AI to be able to mine through that because it's in different structures and different formats than your other ERPs are used to. And so that's the first step that you want to do as you go down the path of integrating the ERPs. Fabulous. Thank you. So the second question here is about hallucinations and errors in AI. At a big scale, how do you ensure that you build the relevant guardrails in place to make sure that those agents don't end making really expensive mistakes? And again, that has to do with, like, the access to data, right, and the context. You know, the broader the context to data is, the less precise it is. I mean, for everyone who has built a skill architecture, basically, for example, knows the more precise you are and the better the contextual knowledge is and also the stricter the rules are, the more precise it is. And that's why it has to be context specific, agents and context specific as small language models or large language models rather than just kind of providing access with one big model to all data available. I mean, that will ultimately only lead to, like, hallucination, which basically no one wants. Absolutely. Ryan, any further comments? Yeah, just from a business use perspective, that's where we still have humans in the mix of reviewing the outputs of this to make sure that the output makes sense and that it's useful and comports with the requirements that are needed, the outcome that's needed. So that's why we still have this little bit of trust that needs to be worked on. But I think over time, as we tailor these models to more specific data, and we're able to understand the fragmentation of that data and the structure of it, then those outcomes will be better and better over time, and then you'll be starting to remove the human element to it over time. Brilliant, and our final question here I'm going to ask you both to give me one word. As we automate to more tactical supply chain roles, what is one new skill that we should be hiring for today to manage that digital first organization of tomorrow? One skill, one word for for a skill that's needed. That's the question. Go ahead. I'd like to get a thought started from Simon. I would say creativity. Yeah. That's a good one. Yeah. Creativity outside the box in order to kind of see what's next on this roadmap. Because it's going to take us in different ways that we might not foresee, and so those people who can have that vision, I think will be important. See, I would say curiosity, so I think very similar. But thank you both so much. I'm afraid that does bring us to the end. We have covered a lot of ground today from the digital backbone required to connect our global networks to the cultural shift needed to move past pilot programmes and into true autonomous procurement. If there is one key takeaway today, it's the digital transformation at scale required as a unified digital thread connecting every tier of the supply chain. So I want to thank you again, Simon and Ryan, sharing such practical insights. And of course, a big thank you to our partners at Amazon Business for making this session possible. As a reminder, you can find the on demand recording of this webinar along with the rest of our Amazon Business series on the Supply Chain digital website. And don't forget to join us in person for the Procurement and Supply Chain Live in Chicago in just a few weeks. So thank you all for joining us and we look forward to you joining us at the next session. Goodbye for now. Before Amazon Business, buying for work was chaotic. Now it's easy to find products from thousands of suppliers in one place. Save on every type of purchase from individual items to bulk orders. At any time, you can view your spending on prebuilt, easy to use dashboards. Plus, you can free up cash flow if you choose to extend payment deadlines and view and approve your team's purchases easily. With Amazon Business, things just got a lot more manageable.