3DRC: 2025 Grand Challenge Winners Shareout | PI Apparel

Good morning, good afternoon, good evening. Welcome to everybody for wherever you're calling in from. Welcome to the 3DRC Grand Challenge, and I'm going to be your host for today. And Harry, may I just ask if I can get the slides back? Yes, thank you. So welcome everybody to our Grand Challenge winner shareout. I'm thrilled that we are able to be here today on the call and I will be your host for the webinar. I'm Sandy Gagnon. I'm one of the co-founders of the 3D Retail Coalition. And if you're not familiar with who the coalition is, I'd love to just take you through our mission statement quickly. We're really just a collaborative and collective group of retailers and brands that work together to advance 3D technology. And our mission is really very simple. We're here to give direction and resources and mostly network opportunities so that brands and retailers can unlock their potential with the true value of 3D digital transformation. And some of the things that we engage in throughout the year are webinars and committee work that is working to develop answers to some of the biggest questions our industry has. But one of the things that we've been engaged in for the last 7 years, as you can see, is our annual Grand Challenge. And I think what makes this so fantastic is we each year try to focus on a theme that really has relevance for a problem we're trying to solve in the industry. And collectively this year, the theme was decided on revolutionizing product development, both from data-driven AI and 3D innovation being brought together to really create and revolutionize something new in the way we think about product. And I am thrilled that today we are presenting the winners to you. And before I introduce them, I just want to give you a little bit of context into how the contest works. So we put out the call in one of our first PI Apparel shows of 2025. We solicit entries, and when they come in, they're reviewed by a peer group from the industry in several rounds. As well as our board of directors, our steering committee, and any external judges that we feel are relevant related to the product category. So it goes through a pretty rigorous process, and we select two winners, one from an academic category and one from a startup category. And we're here today to not only meet but to learn from and to celebrate those two winners. So without further delay, I'd like to introduce to you in the academic category, Lee Lavange from Fashion Institute of Technology, who will take us through her presentation, and then she'll hand over to Christian Sunz from Pixascale, who was our winner in the startup category. At the end of their presentations, we'll come back together, the three of us, and we'll have a chance to have some questions and answers. So if you have any throughout the presentation, please add them to the chat. I think this is what really makes these webinars super engaging is when we get great conversations happening and you can hear what's going on in the fantastic work of these two winners. So without delay, I'm going to hand over to Lee. Lee, congratulations, and take us through your presentation.

Thank you, Sandy, and thank you all for joining. My name is Lee Lavange. I'm an assistant professor at FIT, or Fashion Institute of Technology, I teach in the Tech Design Department, which is part of the School of Business and Technology. Previously, I spent about the last 15 years of my career as a patternmaker and a technical designer. And now at FIT, I teach computerized patternmaking grading using 2D and 3D CAD programs. I'm so grateful to be here today. Thank you to the 3DRC for creating this environment in which we can share our ideas, and more importantly, engage in some discussion, which I'm looking forward to after this presentation. Today in my presentation, I will walk you through the proposal that I submitted for the contest this year. In that proposal, I prove that we can integrate AI with 3D technology to automate custom sizing in a way that is fast, scalable, and profitable. And by doing so, we benefit not only the end consumer, but the brands as well. Oops, behind on my slides. The problem I sought to solve in this proposal was fit and sizing, which go hand in hand. Fit is one of the top reasons for customer returns and the top reasons why consumers don't return to particular brands. When we have poor fit, we end up with customer returns, excess inventory, and excess waste. If a garment does not fit well, there is no secondary market for it. It will inevitably end up in a landfill. The other issue I sought to solve for was human anatomy. Our human bodies are changing as a result the result of many different environmental, social, and cultural shifts. The methods we use today to size garments and mass produce them and sell them are no longer sufficient to meet the demands. We've all watched brands try to launch custom— sorry, not custom— extended sizing that ends up failing, or not selling, I would argue that it's not the lack— it's not, um, the lack of sales is not due to the fact that the customer isn't there, but that we cannot feasibly take a base size Missy 6, size it up to a plus size 24, and expect it to fit well. With the broad ranges of sizes we have in our markets today, the body shapes do not necessarily grow and shrink proportionately. And our bodies are not growing proportionally. How can we expect a standard 1-inch grade up and down the size scale to reflect our consumer? So that brings me to the idea, um, that I proposed. Um, when the industry moved to standard grading, it was to enable mass manufacturing, which enabled higher profits. With today's technologies, I believe that we can go back to the world of custom fits while still delivering product quickly to consumers and doing so in a profitable manner. The model I propose starts with the body and the avatar. Brands who utilize 3D typically fit on a standard avatar. Oftentimes brands may, in addition, have a custom avatar that mimics their fit model. Using a smartphone today, customers can scan their bodies to collect base size measurements. These measurements are then calculated against the brand's base size avatar to collect a set of deltas. Traditional grading uses the method of slashing and opening or closing a pattern in locations to maintain the shape and proportion of the original pattern while also scaling the entire piece. In my model, we use these same techniques to apply the deltas collected from the body scan to the pattern pieces at critical fit points. And hopefully I have a video— yes, I Okay. So this video I have playing in the background just demonstrates the proposed flow. We start with the approved fit on the brand's avatar as we flow through the body scanning apps, the comparison of the avatars, and the changing of the avatars to match the customer's measurements. We then move to pattern correction. So This is a combination of pattern correcting while also scaling to meet the measurements of the customer's avatar. I will say this example that is being shown on screen right now is a very straightforward change, in which the changes are relatively proportional from shoulder to bust to waist to hip. However, this is not always going to be the case. And so because of this, we had to expand this method much further to incorporate some of the complexities that we see with body types and fit. And with that, I will move on to the process. Thank you. So when we add fabrication, styling details, construction, and wearer's preference, the methods for customizing fit become very, very complex. So the first thing I needed to do was ensure that the 3D softwares that we use today could execute and more importantly repeat the steps to scale the pattern pieces based on the values received from the body scan. Secondly, a set of rules and hierarchy needed to be established. These rules are based on the fabric types, the product types, product components— does it have a collar, a placket, a cuff, you know, any number of things— and then also, most importantly, the pressure and the ease over the body. So, for example, in the image on your screen, We have a very tight-fitted stretch knit dress. The top image shows the base size avatar with the original fit, and you see the pressure map next to that. The customer avatar qualified technically as the same base size, so that customer in the store would buy that same base size. However, this customer had a bust that was 2 inches larger than the avatar that this garment was originally fit on. However, because of the fabric stretch, the amount of spandex in that fabric, not much change was needed. So in this case, the fabrication is one of the biggest factors in the decision-making. In some cases, the rules provide a guideline for which points of critical fit and which proportions are most important to maintain. So my best example is a pair of non-fitted, uh, fitted non-stretch women's pants. Um, let's say we have a base size 6. The customer's body is scanned, and the specs indicate that she needs to add an inch in the waist and 2 inches in the hips from the base size. In today's world, that customer would go into a store and she would sacrifice. She would either end up buying a pair of jeans that was too loose in the hips— too loose in the waist or too tight in the hips. But in this system, we can accurately adjust the pants to fit the waist and the hips and still maintain the overall look, the designer's intent, and the drape of the original garment. So this brings me finally to The impact, um, the impact of this project is far-reaching. It has the potential to upend how we design, how we fit, and ultimately how we consume garments. With this model implemented within a brand, it would completely eliminate the need for inventory. It would allow an on-demand production model. Consumer returns would hopefully be eliminated, or at the very minimum reduced. If customers are seeing that garment on their own fit avatar prior to purchase, I believe we would eventually get to a world where we're eliminating customer returns. This model also opens up all kinds of opportunities for personalization. If we can make custom changes on demand, A brand can allow for more unique personalization options for their customer. Perhaps it's a dress that a brand wants to offer in 3 different lengths, or a shirt that wants to be offered in a short sleeve and a long sleeve option, a different collar shape for a button-up shirt. But it also allows for brands to offer other types of personalization, custom graphics, custom logos, etc. But the important thing is the model allows for any interpretation of customization or personalization. Lastly is the calendar. We in the fashion industry all just kind of despise the calendar. In most companies in this industry, we work on a 12 to 18 month calendar. This is a long lead time and can also be a huge risk for new startup brands. But by offering custom sizing, we also reduce the development time significantly. By reducing the calendar and producing on demand, we also create an environment in which brands can be agile. If there is an issue or something is not selling, a brand can pivot quickly if they're not sitting on thousands of units of inventory in a warehouse. I imagine a world where a customer can shop online, choose a style of their liking, see the garment on their own avatar, make personal selections based on their preference. Maybe they prefer things a little oversized, maybe they prefer a tighter fit, make their purchase, the garment is then produced for them and shipped directly to them. Okay, and finally, just, I wanted to say thank you one more time for joining, for listening to this proposal of mine, and again to the 3DRC team for putting this webinar together. I just would also like to say I do believe together we can build a more efficient and a more sustainable fashion industry. When we work in silos, we make a very small drop in a pond. But when we work collaboratively, I do believe we can make a big splash. So I'm really looking forward to the question and answer segment. And I will see you then. Oh, I will pass off now to Christian for his presentation. Thank you.

Thanks so much, Leigh, and also from my side, welcome to the call. Thanks for joining. And of course, I also want to, yeah, reach out to the RC team, the 3DRC team, for organizing not just this event today but also the whole grand challenge. It was really a great experience for us. We got really good, not just for winning and the acknowledgement and visibility, but also for the great feedback that we get during the process. So I can only encourage everyone to also join one of the next calls from the, for the ERC team. So thanks for that one. Yeah, today I will talk not too much about Pixascail specifically, but more about the topic of AI agents, which was also one of the topics that I had within, within the submission to, to the challenge. And I think I'm successful with this talk if I can spark the same enthusiasm in you like I have with AI agents, because I genuinely think that this technology will transform how we work within the next decade, similar like internet did a few years ago. And I want to provide to you maybe a little bit an idea how that could look like in, in the fashion industry. But before getting into this, uh, let me quickly also introduce myself. Who am I? I'm Christian, I'm married to kids. Kids is not really— well, it's two small ladies now, uh, the photo is a little bit old. I'm totally passionate about technology and product creation, and if you look at my career, which is depicted on the bottom, um, this is like a red thread through it. I started in aerospace and automotive, always in product creation, always how can we make, how can we leverage the 3D CAD data that we have for more than just for the construction. Today you would call it digital twin. That was not an expression then, but we looked into exactly more or less the same topics. Then I did a little detour. I went to the German Research Center for Artificial Intelligence, did my PhD in, in the area of computer graphics and AI before I then joined Adidas 9 years ago, and there I had kind of the once-in-a-lifetime opportunity to contribute in the digital transformation of the core process of such a great brand, right, of the creation process. First, I led the apparel work stream from a tech side, and then also the overall program where we digitized from concept down to the digital showroom the whole company, and I was responsible for the technology part leading in the peak time more than 100 people. Really a crazy ride with amazing team, amazing colleagues. Um, great. Um, so, uh, but then after 9 years, it was also— I needed something new. And then I was wondering, all this experience, and it was not always successful. We also failed a lot. That's natural in such a process. Um, maybe someone can benefit also from the same experience. And this is why I founded— then beginning of the year, I quit Adidas, and then later founded Pixascale. The idea of Pixascale is to partner up with our customers in the process from— sorry, that one, that's one of these ladies. So, um, so that's, uh, that's kind of the mission. We want to partner up with brands to start from digital product creation and end up at content creation because we think in the future this will be similar like in other industries that you will have a digital content supply chain where you create your content in parallel with the product creation, right? That's not just true for the product images, but also for story images and other kind of content. And then on the left-hand side, this when it comes to digital product creation and automation, it's more about around efficiency. How can I make this whole process more efficient where we offer a 3D asset management system and scope management system to enable more automation. And then on the right side, it's more of a top-line business case. How can I create novel content? How can I create more personalized content? How can I help brands with this ever-increasing number of channels to have the right content for the right customer? And for these two areas, more or less, you can also divide them into two kinds of disciplines of AI. In the last three years, we talked a lot about GenAI, how with GenAI we can create these great pictures. And of course, this is also a big topic for us where we focus on how can we create in particular product-correct images, imagery at a large scale. And we think that 3D creation is the perfect brief for an AI to create this kind of images. But for today, I will focus more on the other part where it's more about how can I make the whole process more efficient. And this is where AI agents is a super cool technology because we— the AI can then execute simple repetitive tasks autonomously. And this sounds a little bit abstract, but I want to, to make this really as tangible as possible. So let me start with this one. I think we all agree that Digital creation is great for a lot of products, in particular if you have something totally new where you can convert so much faster to your final product, or if you have stakeholders where you need to communicate already good images in an early phase. But then there are also areas, and this is maybe an unpopular, um, opinion in a 3DRC audience here, there are areas where 3D creation still not— did not catch up with the efficiency of the classical 2D process. And, um, I'm— I was just collecting some of the areas where we always found that it's really hard to catch up when, when it's just to create a new colorway or a new material way, very simple things that have more administrative work than maybe in a 2D process. Takedowns is a example that can be handled very efficiently in the 2D process, but it's always a lot of pain to replicate a design in 3D. Graphic shirt, another one where in the end you're more or less only interested in applying graphics on the standard silhouette. Maybe you want to use the chat as well to see, to disagree or agree with that opinion, and also maybe share a few areas where you still struggle with the efficiency of 3D creation, right? And I think also in the last PI Parallels, it's something that I sensed a little bit. There was this promise that the maybe additional effort of 3D is offset by automation that we will get then later. But this automation promise somehow did not realize. And I think mainly because there are so many different processes in a company, in a brand, and it's hard to find then the one-size-fits-all automation, maybe other than rendering an image. Yeah. And this, um, from that perspective, I always saw— I, I sense a certain digital creation fatigue with a lot of brands. And from that perspective, it's super interesting to see how can I now address this blue, blue share of products, which may be— the size may be different from brand to brand, of course. But how can I make, um, these parts where you have less creative work and more administrative work. How can I make that more efficient also in the 3D world? And to make this again as tangible as possible, unfortunately— maybe step in here— is unfortunately I cannot show a live demo today. I will later explain, also explain why, but I still try to make it as tangible as possible where we are currently heading to and what we also in our product. So let's use a very simple use case here where you want to maybe have an existing style and you just want to have another colorway. In this case, just saying this one is in Dusty Blue. And instead of now, I think you most probably think already about all these processes that you now need to start in a brand to get this another colorway started. But let's maybe look from the other side, from the supplier side. If you give this kind of prompt to a supplier, what would the supplier do? He would look up for material he can find about that style, maybe in your PLM system, maybe somewhere scattered around through SharePoint and wherever to find the tech pack, to find the BOM, to find the pattern files. He would try to find some information about Dusty Blue. And then start creating a physical sample. Very simplified in, in that case, right? Um, and I think it would be very similar with an AI agent, right? If you now say, and I give this prompt to an AI agent, um, the AI agent would start creating a similar plan like the supplier. He would— the AI agent would try to find that style, uh, would find try to find more information about Dusty Blue and start creating a digital sample. And here I need to dig a little bit deeper into the technical part. What is an agent actually? So in, in the past years, we all naturally used, uh, ChatGPT, Claude, whatever, an LLM, and we mainly used it as a huge knowledge base, right, that knows so much about our world and can formulate things nicely. And the idea of AI agent is to use the same technology actually not as a knowledge base, but more as a cognitive engine that can plan and reason. So that's kind of new that the— and, and, and but you're already using it on your daily basis. And ChatGPT has that functionality. If you ask them something, it thinks about, can I answer that myself? Do I need to do an action and schedule a web search? And this is exactly what this agent is doing, right? It's building a plan and then it has tools, which is the other important component of an AI agent to then do actions. And so based on that, right, what would an AI agent do? It would know that it can maybe get that style, maybe a clo file from our Pixascall asset management system based on the style number and would look it up and find it. It can, in the same way it would approach the PLM system to maybe get the color code and RGB values for the dusty blue, then it could create inside your digital creation tool a new colorway and then start recoloring the parts with them that it finds that has the main color. And You could now say this is most probably something where you could also have a button, create a new colorway, and you enter a color and you create then in a very classical automation a new colorway. What is kind of interesting, and that makes it really, really amazing, is that you can now also give the LLM business rules, right? And I give here two examples. You can express these business rules. You don't need a, programmer, you can express these business rules as natural language. And I gave you two examples. So one could be that the color you, you are choosing should be in your seasonal palette. If, and it instructs here the agent to propose close alternatives in case that the color that you want to use is not in the seasonal Similar, I think something that we stumbled in the early phase upon is that our designers were suddenly like, oh shit, now that I work in 3D, I suddenly need to define 6 different seams, the color for, right? And that's something that I never did in 2D because we have a standard operating model with the suppliers how we would act with seams. And you can similar here decide what should be done with seams. And with these kinds of policies, what, how such a dialogue with the AI agent could look like is saying, I want to create that style in dusty blue. The agent is looking up, doesn't find dusty blue, gives a few alternatives which are close to, to a most probably dusty blue color. Then you can decide for one of these colors or override the decision. You can, then the agent asks you how do you want to proceed with the seam colors. It can show you visualizations of the two alternatives and then you can choose if you want to have a tone-in-tone or contrasting seam color. And once kind of the recipe is complete, the agent will pick up your digital twin file, CLO, Browsware, whatsoever, and would apply these changes to that file and make it available to you. And I think always important is then the last iteration, we cannot fully trust the AI, but you start from a completely different base and you can start, start them from there. And similar to that very simple use case, you can add more and more and more complex use cases on top. What I absolutely want to solve is the carryover thing. This is such a common use case, and I think we want to solve that. So it should not be more than 1 or 2 sentences with an LLM to carry over style into a next season. No downloading, uploading, renaming whatsoever. Also, applying a graphic to a standard silhouette should be easy. Maybe the positioning is nothing that the AI can do perfectly, but it can do it based on standard operating manuals. And if you want to then adapt, then you just adapt the— do the fine-tuning. We see here some other takedowns, and then the last example is a little bit where we also connect to this ask for data-driven creation where then you could— where one agent could come up with a demand that you detect maybe from, from your e-com experience, from your, from other teams or from, from accounts. And then you just can ask the AI to figure out some proposals very quickly. And then instead of having huge and long iterations to, to come to a answer to that demand, the AI could propose you 6 designs and you then in a team decide from which of these proposals you continue. Yeah, I still owe you why I couldn't show that live here. So the— luckily a larger brand is interested to go into to a pilot with us. And I think to enable this AI agent, one totally important foundational thing is then to have your styles, your assets, your components, your libraries very well organized. Otherwise, also an AI will not be able to solve that, right? So you've seen this call. I need to find a certain style. I need to find the components to apply that. That means you need to have your assets very well organized in a way that an agent can actually also find them and make sense of them. And then our system would also provide then the tools and actions to change your digital twin and do, and to do simple modifications on basis of your digital twin. And this is why we pause that now a bit, and I'm super glad about this one because I know then in 6 months from now, we can really do the first baby steps in this area. And as I said, I expect that this kind of technology will totally change how we will work in, in a few years from now. Yeah, with that, um, thanks a lot, and I hope there are some really good questions. Don't hesitate to ask and also not to reach out. And with that one, I hand over to Harry.

Thank you. Those were fantastic presentations. So thank you to both Lee and Christian. I think, you know, as listening to it, I hope those on the call really see these are the bright ideas that we're looking for when we, you know, decided to have the Grand Challenge this year focus on data-driven AI and 3D technology together. And I want to kick off by saying kind of that I think elephant in the room statement. 3D has been fantastic, but has it solved all the things that we thought it would? Not necessarily. And I think what both of your entries are showing is that next level that we can get to, right? Because consistently over all the years when we've talked to so many brands and retailers using 3D, one of the conversations has been, but we still have to do things one on, you know, one by one in a manual way. And I think both of your proposed solutions really start to address that. So it's just super exciting to see, and thank you for all of the hard work and dedication in bringing this to life, because that's really what is the exciting part. We're going to pivot now and we're going to move into a little bit of question and answer. I love this part of the presentation because I think this is where it gets to be really fun, right? We get to hear from both of you in terms of What inspired you to even get working in this? And I know you've talked a little bit about that, but what I'd really be curious of hearing is where was that point in your process where you had maybe been doing something classical, right? I.e., 3D, and you started to really say there, there has to be a better way. And I think it's AI. And what I'm curious to hear, and I think this is what I think that the audience will find really fascinating is what were the things that you put those AI tools through to help you select the best one to solve your challenge, right? Because there are so many available now. And I think one of the things that can overwhelm new players in it is, where do I even begin? Much like we had conversations with 3D 10 years ago, right? There's a lot of tools. Where do I start? How would you both recommend that somebody would go through a process of identifying a tool for AI, much like you've done in your entries, to help them start thinking about some of these solutions in a different way?

Yeah, maybe, maybe I can start. I understand that this is currently super difficult difficult task because AI is obviously a hype topic, right? And there is so much BS around AI. And you see that there's also this fear of missing out in many companies trying to jump desperately on an AI topic and then failing totally because what you actually want to achieve with the AI is not clear, right? And this is kind of for me the two things that, that will not change independent which AI technology you, you choose is number one, you need to answer why first, right? And number two is then that you make your homework to have your data available and ready, right? This is also something that AI will not solve. It's a different kind of interface, but I said it's a cognitive engine. It's a cognitive engine of a junior person, right? And if that person cannot find the data, also the AI will not find the data. So, so these are for me two preconditions that you actually exactly know what you want to solve and that you have done your homework. And then the, the technology choice is then coming pretty automatically. Right? It's, uh, and sometimes even doesn't make a big difference which engine you use and, and so on. Um, so, um, yeah, that's a little bit my point of view, if that makes sense.

Yeah, totally makes sense. Thank you for that, right? AI is not going to solve all our problems. We still have to do some of the work. So I, I love that, keeping it real. Lee, anything from your perspective that people should be thinking about?

Yeah, um, I will just say I, I Kind of my answer is very similar in that, um, I think it matters less what software we're using, what AI system we're using. Um, from, from my side of things as a pattern maker, it's the data, right? We, AI only works given how much data we provide it. And there's so much in pattern making and grading that is in people's heads, right? Um, you could. You could go look for books on grading, and there's a few. There's just not a lot of information out there in how we do what we do as patternmakers. Most people who want to become patternmakers here go and have to work under somebody who's been doing it for 30, 40 years. You know, we teach as much as we can in school, but there's some of that, those small decision points, right, where we we're deciding if we should move this a quarter of an inch or a 16th of an inch, and why we're making those decisions, those are not recorded anywhere. And so that was sort of part of my— the big crux of my proposal is that they are mathematical decisions that need to be recorded and turned into data that we can then feed into a system. And so I think I think same with 3D now, right? Brands are choosing whatever software they want to, and they're all great. I think it matters more about what you're doing and making sure that we're getting the data fed into the system so that it can do some of the work for us.

Yeah, that's fantastic, right? It's all about data and organization of that data and knowing what you're really trying to solve for. Aren't those good lessons in life in general, right? Being organized. I'm actually going to keep answering some of the— or passing some of the questions to you, Leigh, because there's quite a few in the chat. Really lots of love for what you're proposing and saying it's super interesting. I think a couple of things that pop out. One is, what would it look like essentially for a brand to run a pilot like this? Realistically, what would it look like thinking about your tool? And then also, would you think about certain categories or places where it would make sense to start? I think we'll start with that one.

Sure. You know, I've thought about this a lot and wrote about it in my proposal as well, because at the end of the day, this is a brand, a customer-facing and a brand tool. I think that there is I made my comment about collaborative work at the end of my presentation because this is a, this is a solve for a very small part of the process. However, everything along the supply chain has to change to make these big changes within the process. So yes, to offer— I see a lot of questions about the process and the manufacturing. So for a brand to be able to offer this realistically, we do need an on-demand manufacturing system, ultimately. I do believe that there is a faster-track way. I called it my short-term proposal, which, in a sense, instead of custom-making the garment, it proposes the size that would fit best based on the same data for companies that are still manufacturing large units at a time, large quantities of units at a time. So manufacturing is a big one, right? So that, that's going to be the number one thing that has to change. There are, there are places that are exploring this, that are doing this already. So it's, it seems like a big thing to us. But I believe it is right around the corner. The second thing is e-commerce, right? So this model requires a customer-facing interaction, right? So we've heard for years since 3D came out, the virtual fit room, right? Everyone wants to create a virtual fit room. But just like the 3D technology, a virtual fit room can only go so far. If we don't then produce the garment that the person sees on their bodies, right? And so essentially, this requires a form of virtual fit room where the customer's avatar gets changed, like you all saw in my video. And then they get to actually see that garment dressed on their body before it's then made for them. So to me, those are the big— the two big hurdles on either side of the process. That need to have a solution before this model can run as it should for a brand. As for product type, I don't have a recommendation. What I would say is the same way we implemented 3D within brands, for those of you that have gone through a DPC journey, is start, start small. Start with your, you know, big winners. Start with the ones where you've— you're really set on your fit, you're happy with your fit. Maybe it's a small collection, maybe it's a capsule collection, um, and, and start there. Pilot it, see how it goes.

Yeah, yeah, thank you. I really do appreciate you calling out the fact that there are changes that have to happen throughout the supply chain. In order to bring something like this to life. And I, and I think Christian, you might have even touched on this just really quickly in your presentation when you were talking about all of the data has to be available. And dare I say, I'm not sure as an industry, we do have all that data available today, right? Like specific brands or retailers may have it because they've worked so hard through the years to build that. But I think our, one of our big challenges as an industry is to really get our materials digitized, our trims truly digitized, like in easily accessible libraries. And I know some of our committees in the 3DRC are working on those challenges specifically, but I think it's just a really great thing for this call to, to bring to light is you all have wonderful solutions, but they can't work in a vacuum, right? They need that entire ecosystem to really maximize the great things that you're bringing to the fore. So I just call that out. There's also a question in here for Christian, a couple of them actually. Where do you see the limitations or tasks where AI agents currently are struggling and how possible is it to address those?

Mm-hmm. Yeah, this is a really good question. So I think the technology itself is not limiting too much, right? It's again, it's a question of investment. We see now agents in software development where billions of investment is made to make, to really fine-tune these AI agents to not derail from, from the actual task and do really a good programming job. We see millions going into— you are chatting with your office documents and so on. There's huge investment behind it, of course. Then, if the more you get then domain-specific, the investment gets smaller. And it looks maybe, look now, a little bit too easy with these policies that I was showing. It's really hard to get the AI agent doing what you actually want and not derailing from the task that you actually want from them. And then, maybe also connecting to that one, because, Sonny, you ask also about this organization, and there's the question if that's only for, for big companies. This is, I think, a pretty big trap currently, saying, um, um, for smaller brands this would work exactly the same way, but also they need to organize their data to enable and I see a lot of brands now only introducing a PLM system or not even having a PLM system. Of course, a PLM system would also need to be agent ready because the whole idea is that this agent is reaching out to different systems to collect the data. Right. And this is also the big difference where why we need 12 or 18 months is mainly because creating a product is a highly collaborative project. A product process and you need to reach out to so many people to get fractional amount of information to go ahead. And I think this slows us down a lot, right? If an agent can help you to collect already and assemble the library elements, the data from, from a product brief perspective, from a PLM, from a price point, increasingly also from a environmental impact perspective, then you can have all the information in place. The agent can make a proposal or not, but at least it helps you to get all this information together. And this is where my big passion is also. I see we have an increasingly complex world that leads to increasingly complex and more chaotic processes in the companies. I don't see— when I joined Adidas, it was like we will all standardize our processes into a few and they always got more and more, right? I think it's totally unrealistic that the, the chaos in a company will decrease, but I see the chance that you introduce the right tools to cope with this, with this chaos and with this complexity. And here I think the agents are a wonderful tool to help you to collect your data, to understand your product better to enforce some policies and rules as well, or maybe not— you can decide, enforce or, or guide you. So all this is then something that you can decide as a brand. And from that perspective, this is why I am so fascinated for this one. I think this is the first time that would not work in that industry to build a brand per brand rule system that would cover all these, um, all these things. And, yeah, this is why I'm so passionate about this one.

Well, I'm going to build a little bit on that comment about chaos, because this next question, I think, is kind of, you know, thinking along the same lines. We have digital twins, and some of the chaos that comes in is around everybody working on those digital twins, making changes and that kind of thing. How do you keep version control and make sure that at the end of the day, the ones that they're working on are not the chaotic ones, they are the actual source of truth?

A very, very good question. Um, and, and I think this is also one of the, um, where our tools today are, um, these heavy desktop tools are not really designed for collaborative work. Again, totally different if you look into other industries like software engineering and so on, which is highly collaborative tools. And, and I want to transfer this kind of mechanism to creation as well. Because also I've seen in Adidas, when we provide these collaborative platforms and people can understand who has changed what in that file, when also maybe have a written description what actually changed, right? That you do not search for half an hour in your patterns what actually changed, but have maybe a short description what actually changed that can be computer-generated, then you enable this collaboration. And then, um, so much more in the end results from that, where we have seen that teams suddenly included the supplier so much early into the process, which we never really expected, right? And, um, and this is where then process innovation can actually happen. And this is another topic I'm really passionate about, that we get process innovation into the companies. A lot of companies work like 20 years ago, right? But it keeps adding up the chaos, as you say, and I want to help in the end to get that down again and give the people the space to really create.

I love that, right? Let's have people focus on the things that are important, right? Here's a question I think for both of you, and I think it really comes down to You know, when we think about the work that you're both engaged in, what do you see as the next steps to take it and actually do validation or actually scale it? They're fantastic ideas. How are you thinking about taking it to the next level? Lee, let's start with you on that one.

Sure. Well, as I talked about the complexities, right, on my process side, the— all the product types that there are out there, just applying it over and over again, right? It's, it's like trying to break the system, give me a design, let's see if the system works, right? Fashion is always changing. Fit is changing, fit preference is changing, right? So making sure that we are feeding enough product types, fabric types, component types into the system to ensure that we are able to execute on any type of product. So I think further proof, further practice, and just further developing the idea into something that brands can truly just adopt and implement assuming they have the supply chain set up.

Yeah. Christian, how are you thinking about scaling or taking it to the next level?

Well, scaling in particular is kind of our core idea, right? And we have the experience. We have scaled a product, digital product creation, from a few hundred styles to more than 20,000 styles per per year, millions of rendered images. You can build a totally scalable system in digital product creation. When it comes to the AI part, I also want to be really honest, I don't know. I need here, and that would be my idea, to find then a customer who has the trust to go that path collaboratively, right? And to learn. And similar, like with every hype technology, there will be huge expectations that at some point in time We go through this disappointment phase as well, but still the potential is so big that we need to go through this cycle, right? And, and I'm looking then for, for partners and customers who have the appetite to start with that one, because I think that's the fact— no one will just be able to ignore that technology and not start, right?

Well, that actually leads me into my next question that came across. Have you, both of you, have you been working actively with brands? And if so, what kind of feedback are you receiving? And, and how are you taking that to move forward?

Lee, for me, no. For me, no. This is still in practice theory. After PI, I had a few people reach out. I know there are a a lot of brands out there, or other technology companies that know that this is the next step in the big breakthrough of technology in fashion, right? Everyone knows, getting to that custom pattern or custom design or custom fit or going from straight from AI to a pattern, right? Everyone's talking about that as well. That, that is, this is a big unlock that I think a lot of brands are looking for. And so, yeah, it's coming. But no, I have not, I have not practiced with a particular brand yet. So.

And Christian, you mentioned, I think in your presentation, you were starting to work with a brand. What's the feedback been like so far?

Yeah, I think, right, the experience that we have gives us a little bit of a trust bonus that we now need to also really prove. And people, I think, amazed about the product because it also looks so different than, for instance, a PLM system, and the interaction with the system is really different. So, um, but I'm also honest, right? Um, we, um, I also— we also offer consultancy because I think it's super important also to see other brands from inside because Brand A is not like Brand B and all my experience from now one large brand I cannot just apply one-to-one now to other brands. So I think this is also an important part to be humble enough to say, hey, I don't know how you work, I really want to learn and then also explain you how our system can help you.

Yeah. And I think everybody appreciates that honesty, right? Rather to be a partner in the problem solving than just hop to what could be the solution but presents other challenges along the way. Really quick then, I would say for each of you, what would you be looking for from those on this call? Like, how can we help is the question. Leigh, over to you.

I'm just going to go back to the collaboration. I love interacting with others in this industry who are forward-thinking, right? We all know there's a lot of, this is working, why change it? But there's a lot of great ideas out there. I think when we put them together, we will create something. I would love to partner with another brand, a technology company who can help move this idea to the next stage, to a product that is usable in the industry. And so, you know, that, that's just it. I just, um, you know, this is my plan, this is my idea, I have all the stuff. It's just taking that next big jump. So I love being inspired by other people, having discussions with others that are thinking the same way, and seeing where we can move it together.

So, Christian, how can we help?

Absolutely. I mean, it was a great acknowledgement to win the Grand Challenge, a great acknowledgment to now go with the first pilot, but still we are in the validation phase, right? And, um, and, um, um, so for any kind of feedback, and in particular the critical one, right? In particular those people who say, hey, what you do is absolutely bullshit, that helps, right? So be, uh, and it's really, um, very difficult to get also the, the critical feedback. So, um, maybe here also to the audience, if you think, hey, it's this was, uh, you're heading in the wrong direction, um, this cannot work. Um, but also if you want to, of course, also have positive— if you want to engage, I'm, I'm very open and approachable. I will not hide anything, right? So, um, um, so yeah, reach out. Um, let's have that discussion.

Yeah, thank you. So that's the challenge out to our audience today. Really reach out and engage, which is again back to that core mission of what the 3DRC is all about, right? We've been so inspired by both of your work and what you've brought forward, and I really can't thank you enough for joining us today for the call. It's been great to hear how your tools are starting, and we hope to learn a lot more in the, in the months as we go forward and, and see you again at a PI Show where you'll be back on stage bringing that message and, and hopefully have some fantastic updates. So thank you to you both. It's been great.

Thank you so much.

Thank you so much. Thanks.

And I think we're gonna just go back to our final reminder that if you've enjoyed everything you've seen today, we'll be bringing a lot more to you. So our next webinar is on November 13th, and this is one that we do every year as well. It's a total end-of-year recap on all the things that our amazing committees in SteerCo have been able to work on. Please join us. I think you'll find it extremely helpful to know what's happening in your industry and how you can be a part of it. So thank you again for joining today. Thank you to Leigh and to Christian. It's been fantastic, and have a wonderful day wherever you are.