Duncan Angove & Nunzio Esposito | AI & the Autonomous Supply Chain

Supply chain used to be all about operations. Increasingly, it's a capital allocation and competitive strategy story. Disruption, it's all around us. And after years of trying to buy resilience with more inventory, companies have stranded hundreds of billions of dollars on their balance sheets. Firms can no longer withstand that unproductive use of capital. According to our guests today, more stockpiling is not the answer. Rather, a new operating architecture is required. One where AI agents replace fragmented workflows and humans supervise outcomes and software responds at machine speeds. Joining me to explore that shift are Duncan Anggo, who's the CEO of Blue Yonder and Nunio Espazito, Blue Yonder's chief design officer. Jent, welcome. Good to see you again. >> Hey Dave, great to see you as well. >> Thanks for having us. Yeah, I'm I'm excited to dive in with you guys and go deep into into supply chain. I mean, the state of supply chains is kind of worth noting right now. Obviously, we got tariff tariff volatility. It's like daily. You've got this sort of reshoring pressure. Obviously, we we hear every day in the papers of geopolitical disruption and this massive scale, but it's also making supply chains sort of hit the boardroom as a topic. It's like right there with security these days, especially as it relates to profitability and other ways to sort of allocate capital. So Duncan, set this up. What does this mean for supply chain leaders? What's your message to them? >> Yeah, I mean I I unfortunately what we're living through is now business as usual if you're a supply chain professional, right? So you're in a permanent state of disruption, you know, like you said, whether it's geopolitics, tariff, climate change, labor shortages, you know, volatile, unpredictable consumer demand, you know, all the stuff we're seeing in the food supply chain here with beef and all of that, right? It's just this is business as usual. And you know, the the old supply chains were built around forecasting demand and then orchestrating an entire value chain around that. And that just doesn't work in today's world, right? you need a a new operating architecture that is continuously adapting um and orchestrating all the resources, you know, in a value chain in in real time and and the emergence of AI and what we've got at Blue Yonder with the Blue Yonder network enable a fundamentally different kind of fabric that allows companies to do that. >> Yeah. Thank you. And Nunzio, you sit on the design side. I'm interested in sort of what that that means to you and how this sort of the changes that we're seeing in the market. you're seeing completely new client services whether you're a developer or you're a business user. It really, you know, affects who's actually looking at the screen. You know, it used to be maybe somebody was doing planning. You know, they're getting into the systems. You know, the CE CFO now, you know, wants sort of visibility on this. What's my exposure? Um, so take us through sort of your role and and how you're approaching this design problem. >> Yeah, sure. So um you know we've done it we've done design a little unorthodox here at Blue Yonder. Um and I think it's a testament to uh supporting uh the new operating model and the infrastructure that needs to be in place. Um design not over uh just doesn't uh serve essentially what we call uh consistency or continuity of the total solution. it actually is thinking uh broader um you know our thesis around uh being the only end-to-end supply chain provider uh for the world means that the experience needs to account for edgeto edge type interactions. Um it needs to be able to embrace uh the speeds and the feeds of how uh roles evolve and change uh day-to-day. um and it needs to keep uh the business a breast of what those changes are and you know and in synthesizing all those different uh data inputs uh to reasonable uh based decisions. So um what our team oversees uh is actually a a big piece of our uh platform uh shared services. Um our portfolio obviously leverages those shared services um but it's way bigger than just componentry. Um we really are here to uh serve our customer and its constituents which we call operators. Um and those operators need to be able to feel not just um supported in their day-to-day functions but more importantly um empowered for what's ahead. Uh so we're a we're a cross functional group of uh product managers uh designers and engineers. And I think through the profound effects of uh AI and and how it's changing our SDLC, we actually just call everybody on our team an engineer, >> right? I mean, it's just we're completely changing the way in which we interact with technology at at um your user conference this year. It's called ICON. Duncan, you had a great line. You said the agent is the app. Um that's sort of interesting and intriguing. you know, the user now is is has a whole new way to engage with a whole new system. And that system is intelligent. It's infinitely patient. Um, you got this room full of supply chain peeps who they spent their careers living inside these sort of really detailed dashboards and they had this sort of unique expertise. So, how does that change the dynamic? What's what's replacing it? You know, you're widening the aperture of folks who can interact with the system. What does that all mean? Gosh, you can spend a lot of time answering that one, Dave. I mean, it the most interesting thing that's happening now. And it's not just in supply chain, it's in what the role of the firm, how work gets done, how expertise travels through an organization. I mean, all these things are going to reshape really the world. I mean, you know, when you graduate university, you go to work. I mean, that's kind of what you do for most of your life until you retire. It's a central tenant of society around which you know with humanity is built and lives right and AI is changing all of that and you know so when we say the agent is the app I mean apps have traditionally been the tools that these people use to run supply chains whether it's a warehouse management system it's a transportation system or it's a forecasting and replenishment system right these are the tools that these users have used first of all the thing I will say is this is something nuns in me believe passionate in we hate the word user uh I was in the bookshop in London this weekend and there's a new book out that's called users and it's basically it's about the internet economy and how as we were both users and an input to it we gave them our preferences our clicks our intent and then they monetized it and all the value acrewed to those platforms right and that's where users come from it's almost like being a drug user so we use the term operators um there's this amazing machine called the supply chain that runs the world and it's underappreciated and we stand with the operators that perform magic every single day. And that's what nuns wake he wakes up every day trying to figure out how we can make things more magical for the operators that run the world for us. >> Okay. And these operators, okay, you've got operators, but nuns, you also have agents you're designing for. It's just this whole new class of things. Um, so Duncan's, you know, narrative there was was was pretty bold bold and I was sort of rethinking how how we think about just interacting with systems, >> but you have to sort of create that that that interface, that new experience. So how how are you thinking about that differently? How do you build an interface when it's not just, you know, people, it's also agents and people interacting with agents and and supervising agents? What does that mean? Yeah, I mean well uh obviously uh you know based on what Duncan is saying and how we're supporting it uh we don't believe an agentic experience is just chat. It's not messages to messages. Uh it really needs to help to provide the assistant and the augmentation that the operator really needs. So uh the way that we're looking at it is ensuring that it is infused into the fabric of the experience. Um it's not a sidecar or bolt-on. um we are as a forcing function and I know that might sound negative in its context but it's actually not. We're changing the way in which a user navigates the way it um a user i.e. an operator makes a decision and um what we're doing is holding ourselves accountable to our manifesto of commands over clicks and based on the type of uh feedback loop that we get from our operators uh through our telematics and our insights program um we are really starting to do the thing that's really hard which is ruthless aggregation and simplification which basically means like if it doesn't really need to be there. Why does it exist? If the system is so intelligent and it doesn't necessarily need a confirmation um or the operator has to physically interact with the solution, is there a quicker way to be able to um you know accept and move on? So, um it's it's less about menus, toolbars, and information architecture. I'm not saying that that is not important. um it still exists but it's more around how do you take the humanto human type of dynamics in the way in which uh decisions information um learn knowledge is obtained and you know and how do you build a infrastructure that supports it um and that's essentially what we've been setting out to do um and I would tell you probably the latest uh solution that we have that really starts to encapsulate a big piece of That paradigm shift uh was led through our new uh space planning and category management experience and through there we learned a lot around the readiness of an operator and the willingness to be able to do the thing that is the hardest barrier of entry um for any type of uh enterprise technology which is the uh change management itself. like how do I go from this that I've learned and done for x amount of years and now I have to have a new behavior uh and changing that behavior is actually the hurdle that needs to be um addressed and I want to ask you guys about you know there's a lot of talk and fear around agents replacing people and robots and and the like but before we get there George Gilbert has has done quite a bit of work and he and I have sort of studied this this this interaction between the the the front end, the client, the new surface and the back end because a lot of times humans are going to be involved. There's going to be exceptions and the system has to learn from the reasoning traces of the human that sort of tacit knowledge. We talk about that all the time. Um and and so it's not just a user interface. It's not just a pretty front end. It's actually got intelligence that has to interact with what I call the back end or the what we call sometimes the system of intelligence. Alex calls it the ontology. Um all the buzzwords are there but but basically that closed loop system where that entire system is is learning constantly. Can you address that? Is that fundamental to your architecture? >> Yeah, it is. So I mean I we boil it down to sort of think of it as four design tenants, right? So first of all um you have to design for two participants. One is the agent and the other is what we could call the augmented operator. So the agent using these tools whether it's a WMS, it's a picking thing, it's a planning engine, right? You need to make sure that it's built for agents. So you don't feel like you're having a DOS attack every time an agent hits your apps, right? So it has to be performant, it has to be headless, it has to be secure, you need it, it has to have machine speed access, right? So that's one. And then the other is the augmented operator like nuns was talking about they're no longer clicking right they're actually expressing intent guard rails objectives they're managing agents right so that's when we talk about the agent becomes the app and the applications become its tools right that's that's sort of one second though is that unit of transformation shouldn't just be the end user right it should actually be the entire operational system you know so also design for that um and then to your point about learning is you want to scale expertise and perspective and every exception you encounter should be hill climbing the underlying engine and model right so Jean P who was the Swiss psychologist basically said that intelligence isn't what you know it's what you do when you don't know and it's very very profound I mean you think about software it's deterministic it's if then statements it's what you've encoded what you know in rules and what happens when something's wrong like you you've built a wave plan in a warehouse and there's a short order alloc allocation problem, the software can't solve it. That's where a human's in the loop and they have their institutional knowledge. They're a super user. This is what you do. Now you have agents that can actually reason through that problem, right? It's a new type of of of thing and that's what you want to hill climb so that next time that knowledge is in the agentic architecture and the system itself is learning continuously. >> Okay. and and there's a big gap between sort of the the data oriented uh tasks that we're doing today and what AI can do and and just being able to absorb that that tacet knowledge of the of the enterprise and act on it. Um there's there's a big gap there. So it feels like we need humans for a while. In fact, you've been pretty vocal that the people who talk about, you know, agents replacing people is the wrong question. Although look there there is an economic incentive out there to replace headcounts and public policy basically rewards us. You can write off a capital expense in in one year now. So there is that economic incentive but we'll put the public policy debate aside. But what is sort of the the right way to think about this? What does you know success look like? What how do people's jobs change? You know what do they actually become? >> So it was interesting. I was actually on with, you know, I'm a graduate of UCL in London in economics and I was on with them earlier today and, you know, if you look at it, the number of graduate job openings this year was down 85% from a peak in 2017. So companies are no longer hiring early talent, right? And the challenge is is that it's you don't just hire those people to do lower skilled kind of things that require years of institutional knowledge. You hire them because that's how you build expertise in the company, right? And you know there's almost 200,000 new students going to university for the first time this year and it's sort of like what do they do? What do you tell them? Right? So it's a it's a profound issue for society. You know again back to economics. There was a guy called Ronald Coast and he wrote the the theory of the firm. Why does the firm exist? And it exists because it reduces transaction costs. If a firm didn't h didn't didn't exist, you'd have to go out for every task. You'd have to go out and hire someone, get a contract, monitor performance, all of that. So it was it was cheaper to insource all of that in your own hierarchy and that's why the firm exists. And then um agents come along and they actually compress those transaction costs, right? They can do all of this. They can find talent. They can negotiate something. They can do all of it. So it starts to challenge why a firm exists to begin with. And then you could even go further and you could say you have this arbitrary thing called an income statement. Um but all that does is it categorizes spending according to where people sit. I've got people in development, people in sales, people in services. It literally reflects the firm and people. Well, if an agent one day can be coding and the next it can be doing customer service and then it can be generating demand, where does it sit in the income statement? So, there's all these like tenants we have as a society around why a firm exists, what is an income statement, how does expertise climb in an organization that all of this technology is going to fundamentally challenge. >> Yeah. And you're you're alluding to what I've said is you're going to see all these these fragmented departments dissolve. I mean, we talk about determinism. It's sort of a myth. We have determinism in the finance department. We have determinism maybe in the logistics department and and and maybe there's some determinism in in the analytics group. But you put all everybody in the room and they're arguing about what the actual truth is, which gets me to trust. Nunio, how do we trust these agentic experiences? Because the more autonomous these agents get, you know, this is have this black box problem. The humans really don't know how the decision was made. So how do you ensure that that transparency that a supervisor or planner needs? >> Yeah. So I mean uh like a lot of the uh frontier labs how they are capturing um input and feedback um you know we're doing very similar things from our tieatics. So um you know we definitely do uh nudges ask for uh feedback. It's very fluid. It's not non-intrusive. Um we are also monitoring um what is said and and the results of what's being said to help feed you know create that feedback loop that's back in there. But outside of the actual uh interaction or engagement at hand, trust really comes down uh back to what Duncan was um uh saying around the operational change management and the willingness to do something different. And that's what we've been actually learning um as we are uh deploying our agent capabilities um and our obviously our our newer solutions uh on the Blue Yonder platform. Um what we're seeing is there's a there is a barrier of entry around like role mapping. It's the things that uh at least for those that are in design that might be listening to this, it's academia kind of classifies it as service design. And it's essentially like the jobs I do today to the jobs I need to do tomorrow. And what's hard about it is the tomorrow is unknown. And in order to get from point A to point B, you have to throw something out there. And that's the reason why design is so positioned well to be able to facilitate and and and find those answers. It's it's no more the limitation of like what is it going to take to make it. It's more around like what is it what if and what should it be. So we ask those questions with our customers to help them through this transformation and this change management approach. Um, and it's things like, you know, I'll give you a use case example. Um, to create trust and confidence. Um, it's it's a little bit about like, okay, you have X warehouse managers and a warehouse manager covers a shift and they cover one warehouse. Okay. So, what if there was one warehouse manager that covered 10 warehouses and they were they needed to do it at X different types of shifts. And how do we ensure that there's well-being there? How do we ensure that there is um we can synthesize uh and create the kind of coverage that is needed the blocking and tackling and etc. And and if that's where the customer wants to be then the technology needs to be able to obviously support that and we need to be there to solve on how an operator uh can welcome that. And you know the success criteria that we put out there um and Doug and I actually talk about this a lot is like okay you know how do we ensure that they don't go over in their shift so that they can get home in time and they can they can see their daughter or their son you know at at at you know intramural sports. I mean like I know that sounds crazy but if the supply chain rules the world and the operators are part of it those are things we actually need to think about. >> Yeah. Yeah, Dave, it sounds really trivial, but one of the biggest opportunities we've seen is actually in good old change management. And what I talked about how expertise travels through an organization. There's a reason all these frontier firms have created deployment companies, right? You know, because it's really really hard at the frontier to actually figure out how to implement AI. And what we're seeing is I mean, just take a warehouse again. You know, wave planning or order planning is really hard to do. It's the heart and soul of the operation. And you can train someone but as soon in a classroom but as soon as something goes wrong they don't actually know what to do. So they would have a super use that has all this institutional knowledge and code in their head and then they teach people right that can all be done agentically. Now that agent can start to understand all of this and actually help the operator execute and operate like they're a like they're a super user and it hill climbs and learns the whole time. It also does things like it brings eclectic decision-m into it. Ironically, it enables crossf functional collaboration. So, let's just say you're running a warehouse and you're 14% behind plan and you say, "I'm going to add a second shift." Well, now the agent can bring HR's perspective to it. You can do that, but that means you might have absence and attrition next week. You can bring safety officer into it. You're going to have a higher incident rate. You you can start to get this kind of eclectic decision- making and again, you're making the operator basically execute at a fundamentally different level. >> Thank you for that. I mean that brings me to sort of another interesting topic. I'm glad you mentioned this sort of deploy co. It's a hot trend right now. I mean you're you're definitely seeing these as you know Duncan these you know software heavy portfolios at at PE firms. They're bringing in these forward deployed AI engineers to AIFI their SAS so they can survive the SAS apocalypse. But there's also a a big conversation going on right now about open versus closed models. The whole US versus China competition and all that. You guys recently shared some news with Nvidia which is now they're leading the charge with Neatron. You saw Jensen's first expost hit like a million followers first day I think. So it's open source it's open weights. I'm interested in the work that Blue Yonder and Nvidia are doing and and I'd love your take on the importance of openw weight models. You know why people should care. You know what it means to we could talk go forever what it means to the the closed models because this stuff's not easy. And I actually think the the the anthropics and the open AIs of the worlds have a great opportunity to simplify this. But what's your take on all that? And take take us through the NVIDIA announcement. >> Yes. Yeah. Uh I mean first of all let me just I can't let the FDE comment go without comment on commenting on it. So let's just understand what FTEE is. And it by the way unfortunately it's a way to rebrand consulting across the entire industry right just because you rename a consultant FTE that's not actually what it is. And let's understand why it existed. you had to put an engineer in because the platform wasn't self-service. It was so complex that it required someone in engineering to actually come in and configure it and make it work. The second thing is we had to put an engineer in because we don't understand your industry and we need to learn there so that we can actually build something that's relevant for you. Right? So I think those two things are actually the opposite of what you want to do when you build a platform. You want a platform that doesn't require an engineer. It's agentic and you can actually you can inference is bigger than configuration. Inference actually does your configuration. The agent does it for you. And the second thing is it's built purpose-built for supply chain and the people that built it understand your industry inside out and don't need to learn on your dime. Okay, there's my FDE rant over um just your point on open versus closed. I mean, we announced this back in May and actually I think we did it before anyone else in the industry was really grabbing on to it before that big open letter came out and we talked a lot about the importance of open um so we're big believers in it. You know, we we announced a partnership with Nvidia back in May around training openweight specialized models for supply chain um that would be dramatically cheaper than global frontier models. There's always going to be a use for those, but for some use cases they're not. again in a warehouse you want something that has zero latency is super fast um as a result and um and is dramatically cheaper uh and is trained for this particular purpose. So that's what we set out to do with Nvidia back in May. We've made tons of progress. We continue to benchmark it against other models. It's dramatically more accurate, faster, and cheaper. Um so we're we're a big big believer in that. We also believe in the recent discussions around sovereign AI. This is something else we talked about and this ladders back to sort of the internet economy. let's monetize user model. This idea that you should be giving someone frontier models your intelligence and they're rented back to you. We also don't believe in that. Right? So the way we built our um open weight models is that they're trained on synthetic customer data, not our customers data and then they are encapsulated at that customer and only they hill climb it. Right? So we um we were pretty early in both of those making both of those statements. >> Yeah. The um the sovereigns a hot topic. I actually want to come back to that but but so the but the Nvidia relationship it it relates to Neotron correct and so you guys are doing stuff with with Nvidia and Neotron >> do you not I mean do you do you use closed models selectively or have you sort of eliminated those where do you stand on that? Yeah. So in our harness we have a model forking um algorithm. So there are cases where a a bigger broader frontier model um will actually make sense. You have to make the token economics work as well. But a lot of cases I would say at least 80% of them you don't need that right. You the open weight model gives you a better answer particularly if it's trained on a specialized supply chain domain like ours are. >> Yeah. And so the sovereign is interesting. We're digging into that with our our our research team here. And you know, it's not just about token costs, right? It's it's about the outcome and the value that you get. And I I know you you understand this. And so I want to ask you about sort of you're building some very sophisticated agents for supply chain and that's clearly the future, but there's a lot of legacy systems out there. you know, you know that you've been in the software business for a long long time and customers need help migrating from those and and implementing agentic systems. So, what capabilities or tooling or other functions are you building to help customers with those issues? >> Yeah, I think it's it's no surprise that the area where AI has sort of proven itself um is in coding, right? I mean, that's obviously the place that it it's been and if you look at the revenues of of a lot of these companies, I mean, that's where a lot of it's coming from. Um so we took that mindset and we applied it to the technical deployment of software whether it's a migration an upgrade or a new deployment and we said here is a traditional we I hate the word statement of work it just sounds like it's ancient from the industrial era but what are all the activities involved right configuration data management integration testing all of these types of things and how much of that can we automate agentically and um we went after it aggressively we called it um frictionless outcomes and how you how you can basically automate agentically the implementation or migration of software and we did it I mean again created destruction it's 20% of our revenue um and we went after it um because we think it's the right thing for customers and the right thing for the market so you know we can migrate customers in 24 hours whether it's a planning system or it's a warehouse management system and they go from either an on-prem legacy system or they're running a hosted single tenant version of the cloud uh which is most of the planning solutions out there today and we automate moving them to a cloudnative AI first um agentic cognitive solution. So, Nunio, I mean, you've got these agents, you know, in new places in and out of the organization, sort of learning all learning the sort of tribal knowledge. Are you able to sort of how are you able to track that, get telemetry, make sure they're behaving properly, secure them? Yeah. So, yeah. So to to what uh Duncan is saying as far as creative uh deconstruction, I mean we're we're also looking at uh ways to ensure that there is on day one there's a onetoone mapping of where you're coming from to what it means now um in this you know this new solution this new arena and and ways to ease the friction points which is hey it's onetoone you know I work in this workbench which was called a workbench and we ensure that the new view and it's called a workbench. But at the same time, we're also educating the operator that there is also end many ways to be able to get that job done and and how do you nudge and support and educate so that it removes friction and really isn't a force function uh for the operator. It's more around um their own um appetite, willingness to try something new. And through those telmatics we are capturing that at a role by role level. Uh you know you could say a knowledge graph context graph I mean you know that they work together. I mean, you know, not to get into any kind of like marketing jargon, but we are cataloging it and through that data and through those interactions, it helps to inform the next best step or the next piece of content or the next popup or nudge or so. There are uh different mechanics uh numerous ways to handle the way in which the system helps to um you know coach the operator from point A to point B and it activates uh different modalities uh different touch points and it meets the operator where they are um and that's really important because trying to create you know SOPs and do all this onboarding And then, you know, our customer in the IT department has to set up another SharePoint site. And I think we all know that there's like 7,000 millions of those at at a at a at a site. It it it just gets lost in translation. And what we don't want to see is the value that our uh solution is providing not actually getting the adoption that it that it it needs to have because it's been designed for them to operate better. uh for them to make the decisions that they need and the better decisions that they make then the smarter the system gets. So, you know, those kind of hooks um are extremely important and the you know, the bigger barrier there that I just want to communicate as Duncan is talking about FTEE. Um I would basically tell you that um from what I have seen and where we are uh a friction point is the willingness for an IT department especially at our customers uh sites to allow for a user to decide how they want to work and that has been very hard and a challenge that our group is undertaking now. And what I mean by that is turning something on and allowing the user to go they can get that done this way or they can go get it done that way. And we need our customer to be okay with that. >> That's how this system gets better and and we get better cataloging. Hey, just on that n so just one thing that's again change management is the huge obstacle here >> right the all these people are used to doing things this way and AI changes fundamentally um how you do it in the p in you know going forward and I say to nuns we want our software to be trainingless you don't need training to use it because it comes with an operator coach that by the way is it represents you they want you to be the human that wins in the organization they share accountability with you just like a a tennis pro has a coach. And so they onboard you, they train you, they teach you how to use it, but then they stay behind. And as you're making decisions, they're working with you to make make sure that you're exceptional. So let's just say you you do an override and it reduces availability. They come back the next day and say, "Hey, that was on us. Here's what we should have done. I notice that you you ponder over these decisions. I notice that Bill in the next cube is better at you than this." And they're constantly working with you to make you better. And to that regard, I mean, nuns of me take huge exception to to the phrase UAT. At the end of a deployment, you have this thing that's called user acceptance testing. Well, first of all, we don't like user. We've said that. But think about the word acceptance. >> Like Dave, if you cooked a meal and then someone your guest said, um, you asked your guest how was it and they said it's acceptable. How would you feel? Like that's our hurdle. These poor operators that we inflict this stuff on and we're done with that. We're going to we're we've completely reimagined UAT as operator endorsement. If an operator's going to endorse it that my people are going to use this and I endorse it, it's a completely different hurdle. >> Well, and that endorsement is is a strong signal to the intelligence engine which then knows how to act and to Nunio what you were talking about earlier. It seems to me you the better your data, the better your analytics, the better the decision- making, the better the ex acceptance rate it goes to now endorsement. Um, and that just creates better outcomes. You know, you guys are really tackling some hard problems. Um, and I want to end on predictions and people when I ask people to look out a few years, they say, "Ah, forget it. I things change so fast." I actually think Duncan it's easier to forecast out four or five years from now or maybe even five to 10 years from now than it is two quarters from now. I I can't predict what you know what open- source model or what China is going to do with the geopolitics. But when you think about the potential, I mean, I go back to the internet days and we we saw the potential and we and the same with the cloud. We I think generally accurately forecasted what the end state, you know, typically was going to look like, but how it got there, it was kind of messy. What do you think if you look out in supply chain five years from now, middle of next decade, you know, late next decade, what do you think is going to fundamentally change that is going to make us forget the old way or look back and say, can you believe we used to do that? What does it look like? >> I think a few things. I mean, first of all, we have a labor shortage in supply chain, right? So, it's not like robots and we haven't even talked about physical robots. I don't know if you've been watching the the robot games in China. Amazing. I mean, that's going to be something else. And there sort of compounding. This is sort of the brain and then you have the physical robot and they're going to compound in terms of what they're capable of doing. So, it's not like we're going to see people I mean, we have a labor shortage. So, actually robotics and AI will help solve from of some of that. Um, and I think that one thing for sure is that people will start to think about supply chain software or even the supply chain you run more as one organic adaptive system versus today it's all served by different departments, different companies and software's grown up to support each of these different things. So I think and we're seeing this already all those silos will go away inside a company departmentally and there'll be much better orchestration across companies as well. So that's obviously we spent the last sort of 5 years building and you know the agentic capability just enables so much more of that sort of real-time orchestration. So I think that's one thing that we'll see and and hopefully this will save you know solve for the labor shortage as well. >> Yeah. And yeah, Usain Bolt now is eclipsed by a by a robot. But you know, humans, we can make things go fast, you know, so big Bolt fan, but but as well, it brings me back to the policy public policy discussion where there is a huge economic incentive to automate and and bring in machines. And I think that, you know, you'll see over time tax policy will change and and support the human condition because as you pointed out in this interview, you know, that continues to be vital. Guys, thanks so much and uh congratulations. It's amazing the progress that you guys have made, you know, from from putting together, you know, a lot of so-called legacy software to really reinventing uh yourselves and what has become Blue Yonder. So, congratulations. super excited to keep in touch with you and monitor the progress. Thanks for your time. >> Thank you, Dave. Thank you for having us. >> Thank you. >> All right. And thank you for watching. This is Dave Volante for the Cube and we'll see you next time.