AI Startups Redefining Merchandise Planning | Profitmind, Sizeo, Intelo AI & Hakio | PI Apparel

I want to start by introducing myself. My name is Mads. I am the CEO and founder of Hakio, and Hakio is a demand planning and forecasting tool specifically built for the fashion industry. I am really excited to be here. We flew in from Denmark, my colleague and I, so really, really excited to come here today. I'm married with Mia, who's a nurse, and we have two daughters together. And one thing you should know about me is that my path into the fashion industry is a little bit unusual. I studied physics and nanotechnology, and that really inspired a deep love for mathematics in me. After that, I got to spend an amazing couple of years as a management consultant where I worked with some of the leading supply chain experts within the fashion industry. And from that vantage point, I got to see a dilemma unfolding in the fashion world. So today, over the next 6 minutes and 46 seconds, I'll try to unfold that dilemma for you and describe how we at Hakio have decided to tackle that. In short, I think I can say that it's never been harder to grow in the fashion industry. Growing in the fashion industry today entails more collections, more drops, a lot more styles in general, and new sales channels that are popping up. In general, that is more complexity, and this more complexity is putting a lot of pressure on your demand planners to handle this complexity. How can they lessen that feeling of gambling in this increasingly uncertain world. I really believe that there will be winners in this economy because people that step into this arena for growth, they will see that the old way of planning simply doesn't work anymore. We at Hakio have found when we spoke to these amazing planners that they have three distinct issues that they're seeing. And they see that the rules have changed. When we talk to them, they say that they want to make better decisions backed by data. They want to spend more time wisely and not spend their time feeding Excel sheets and drawing out data from multiple systems. And I want to be able to do it in a unified and scalable way. And when I talk to these planners, hundreds of brands, I realize that it's not a lack of talent that's out there. It's not a lack of willingness. But what we see is that it's an inadequate set of processes and outdated systems. So I ask myself, what if I could use advanced mathematics deep learning models and create a platform where I could help planners to deal with this increasingly complex world. And that's why I built Hackio. Today I stand with a very mature offering for the fashion industry, and Hackio is a platform for AI-driven demand planning and merchandise planning, very specifically built for the fashion industry. The offering that we've built in our platform can be comprised in 3 different areas. So first and foremost, the forecasting engine, and I really loved everything that the panel was talking about here because that really was what we've been working on the last 4 years. So first and foremost, forecasting accuracy is what our platform is built for. It is taking vast amounts of data and forecasting both your never out of stock collection, your evergreens, but also new products. Machine learning can do a whole lot there. But where we see it really come into life is when you have a modern and user-friendly front ends that can help these planners easily act upon suggestions to replenish or to mark down or to hold back, just to name a few. The HackYou app is going to be a companion also for orchestrating your processes in the most optimal way. And this is all the way from early budgets to last item sold. I'm really humbled and very thankful to be able to work with some amazing brands, some brands that I admire, and some brands that are giving us a whole lot of input. You might know a few of them. A thing that we do at Hackio is that we build it together with these brands. So we don't try to solve something from a remote position. We sit very close to our customers and handle the problems that they see. And despite this being a very amazing set of companies that we work with, I'm really, as a nerd and a geek and a physicist, I'm really like these numbers is something that really is inspiring me. So every day we run billions of historical sales orders that we are analyzing with our machine learning models. We have a library of more than 20 models that will get you the right forecasting algorithm for the right option at the right time. And we are processing more than a million SKUs every single day with more than 1,000 points of sales. And it's with these numbers that we are able to provide real value for our customers. So we're seeing increased availability, decreased working capital, and that is of course coming from our improved forecasting accuracy. And of course, we're also helping companies increase their revenue potential. And to be honest, as a founder and CEO of a company that is fairly young, it is when we get feedback like this that we really know that we're making a change. It now gives me peace of mind to be able to trust in Haku instead of having to wreck my brain for mistakes I might have made. We know that trust is not easy to get. At Haku today, we are more than 25 forecasting experts. Data scientists, and engineers. And we work very, very hard to earn that trust and very, very hard to keep that trust. We have yet to meet a company that is too big for us to handle. We work with companies that are down to $20 million in revenue and up to billions in revenue. We are here today or the full 2 days, Martin and I, my colleague, he's down there in the back. Feel free to reach out and ask more questions if you want. Thank you very much for your time.

Thank you. Can you guys hear me?

All right.

Just go to my next slide.

All right.

Thanks for having us here to introduce Intello.ai. Intello.ai is— and as a company, what we do is pretty much build AI agents purpose-built for merchandise planning. So a lot of the discussions we were having in the you know, discussion before this around starting small and starting at the pain point. That's exactly what we do. We build AI agents focused on specific pain points, right? So, and you'll see some of the examples. Before I, you know, give you more about what the product itself is, just wanted to introduce the core team, which is out here. We are a 25-member team spread across multiple parts of the world, but Myself, I'm Rupesh Nair. Pretty much if you look at the mixture of the team, I am the consultant who worked at, who ran Accenture Retail, worked at Publicis.Sapien running retail, ended up kind of doing an HR tech company which basically used a lot of the concepts of retail in high-volume recruiting, and then eventually came back to my calling. We started Intello about 2.5 years ago with the purpose, specific purpose of kind of really solving those smaller pain points, which then adds up to the bigger pain points. Some of my colleagues who are out here, Jeff Fish, who's more of a, think about him as the tech scaling guy. So he comes from having sold his previous company to Salesforce and then running Salesforce China over the last 5 years. And then we also have a practitioner, Vasilis, who basically, came from being in retail and kind of having seen a lot of these problems firsthand. I joke sometimes, like, as a consultant, I can be there for a year, you know, hand it over and then not see all the problems. Vasilis, on the other hand, has seen that problem across. And so we kind of make a good partnership between these three and of course the extended team out here. So what do we do? We basically, as I was mentioning, purpose-built agent KI. And you will hear agent KI at least 2,000 times this week or in the next 2 days. And I'm sure there's a subtle difference between all the different models of AI. But, you know, in essence, what you need to remember is the fact that we focus on specific pain point and then bring AI agents exactly into kind of those specific points so that you can innovate and you can focus on some of the customer excellence points, which we were talking about in the last in the last discussion, right? So the core pain point theory is that, and again, we touched on it in the last discussion, is there is still lots of missed customer demand and whether it results in lost sales opportunities because of out-of-stocks, because of broken sizes, because of mixed revenues. And then, you know, we talked about how teams are getting inefficient because they're jumping between systems, spreadsheets, I can bet that if any of you are a practitioner, there is a good amount of time which your team spends today between basically skipping between systems which took years to implement and spreadsheets, and they spend a lot of time on spreadsheets still doing some of those plans, which then takes your time away from some of those strategic aspects. Honestly, some of the areas we solve for are exactly those points, those areas where people were spending too much time in that particular area and then fix that specific pain point. Now, that, of course, results in inconsistent customer experience. I think we have covered that in the previous session, so I don't need to go through some of that, right? So what did we do? We kind of built AI agents which can be plugged and played. So for example, one of the problems we talked about, rebalancing. So there was a problem. We talked to one of the customers. We were talking about the rebalancing problem. We kind of took that problem, solved that, and then implemented that across customers. One of the customers was looking at seed plan generation as a problem. They spend weeks basically generating a seed plan. Built an agent just focused on seed plan as example. This is plug and play which can be solved for specific things and then can be deployed quickly instead of spending years or even months deploying. We're talking about 30-day deployments for specific problems. The reason we are able to do that is because we are hyper-focused on those pain points across customers. And then it also is built into your, your productivity platforms like Teams, Slack. So it just sits there. It's not a new platform. It's an agent which sits on your Teams or Slack or as a chat. So your planner can literally go in and say, hey, I'm thinking about doing rebalancing this week. How does it look like? Right. And then it just chats with you and we'll— I think the best way to kind of see some of the things instead of kind of going through our agents, we have certain agents we have built across the different lifecycle. But the best way to kind of go through it is probably to show you examples of some of our agents, because I think a lot of the presentations will look very similar. So we kind of, instead of kind of spending time on what the pitch is, we thought the best way to do it is take some agent examples. So let's see some agents in action. So this agent which you're looking at right now is coming out of the pain point that, hey, I want to generate seed plants. So whether it's a seed buy plan, seed merchant financial plan, the agent, you can have a conversation with the agent. It does all the work for you, kind of goes in, pulls LLI, LLI, does all the crunching of the numbers, builds the information for you. But then the subtle, you know, impact of this, the subtle difference in what you would see in some of the others is the ability to kind of look at that data and then really challenge the data, download a spreadsheet, work in spreadsheet if you want to. Push it back in, then the agent automatically figures out what the updates were. And then, or you can update it like a traditional update where you can go in and update the number by clicking on, on the column just like you would do in a dashboard. So it's like it gives you different ways of working with the platform, whether it is just updating a number or whether it is just a chat where you're saying, hey, I want to update the number from X to Y and it just goes and figures out where to update. So you'll see an example here. I'm just updating the, forecast right away and it updates the numbers through this stuff. So here literally you're not just generating the seed plan, but you're working with the agent to really get to kind of your outcome. So that's why from an adoption perspective, it's not a pain point because you are adopting it by design. You're actually working with it by design and you're able to kind of bring in whatever spreadsheet formats you have and then work with it if that's what you want to. I'll take an example of another agent just looking at time here. So this is an agent where you are actually bringing in newness. So you're trying to bring data from your PLM. We actually have some partners here too where we bring the data from some of the PLMs and then we are helping specifically with how do we forecast newness. So here the same thing, you can have a conversation with the agent if you want. You can actually go and see some of the AI-generated attributes for the new product which came in. You can click on a specific product if you want or just have a conversation. And then see how it created the newness forecast, right? So you're seeing this dress, for example, it is similar to these three dresses, and then the merchant can actually challenge it. That is the difference here, right? It's not a platform on the side, it's a conversational engine where the merchant is actually challenging it. Another quick example out here is in a similar area, but fabric forecasting. This is one of the issues one of the customers had. I'm not able to forecast my fabric demands. So here the agent specifically focuses on how to basically forecast the fabric which is needed based on basically the merchandise plan or based on the collection plan in this case. Now the good news is in the same way you can have a conversation, you can upload your previous plan, things like that. So my purpose here was to not basically just kind of show— sorry, just kind of show different examples of what the agents can be used in. And the way we do it is we basically build the agent and then it kind of is deployed across different customers based on the need. We are here, out here, happy to answer any questions if you have any. I know I have right there getting to zero in the timer, so I'll pass on to my next colleague out here who can present.

Right, good morning everybody. My body's at the point of the morning where it's not quite sure if it's had too much or too little caffeine, so please bear with me. So I'm just gonna introduce myself. I'm the CEO and co-founder of ProfitMind. A little bit about my background. So I'm a 25-year retailer. I led every function of a retailer other than accounting and stores. Stores, and I was a head of planning inventory management for 25 years across the logos that you see there. I've also got a doctorate in applied analytics, which I specifically researched application into retail, but given my, my industry of choice, and also how to use AI to create more profit for retailers, customer satisfaction, but also how to transform organizations using artificial intelligence. And I got frustrated as a retail operator, and I'll explain why in a second. But I left my job as a retailer and I said, there's a solution missing in retail that nobody's developed. And I felt like I was uniquely positioned as a retail executive with a background in AI to go out to Silicon Valley, build the solution with leaders in AI, and then come back to retail with the solution that I felt like we needed. So I went out. Silicon Valley right at the end of COVID in 2021. I partnered with Andrew Ng. He's the founder of AI at Google. If you've ever used Google Maps, you're already using his tech. He's also the founder and chairman of Coursera, which is the largest online learning platform. His mission has been to bring AI to the, to the world outside of just these big consumer internet companies. And so actually, I think 20 million people have taken his AI courses on Coursera, and I highly recommend those if you're trying to learn about AI, does everything from basics to, to advanced. He's recently joined the board of Amazon. And you'll hear a lot today about AI agents and agentic AI, and we're going to be no different. I think the one difference about us is Andrew is actually widely credited as the person who invented those terms, and in particular agentic AI. And in large part, it's because we built the first ever agentic AI platform. We deployed it 2 years ago. and that's what I'm going to talk about today. So the reason I left inline retail was this. And actually, this look you see on the face of this person was the look I saw on all of my teams every Monday morning as we tried to kind of go through the 2 petabytes of data that every retailer has, the thousands of reports that we get bombarded by trying to figure out like, what are we doing? What are we going to do differently? What's happening in the marketplace, how do I navigate? I saw a stat recently that the average retailer has between 10,000 and 25,000 active spreadsheets. And it's because as retailers, we're dropping this data into Excel to create the rounded view of what's happening in the marketplace, what's happening with the business. And there are no real solutions that do it. They're all siloed. And that's a massive problem. And then we're all wrestling with inflation and tariffs and social media trends and just trying to keep up. And it's an impossible task. And so this is why I left retail and said, let's go build a solution that's going to help the planners and the merchants of the world do this much better, faster, more intelligently, and, and actually use their talents for the right purpose, which isn't in spreadsheets. It's actually to figure out how to use that data to go and deliver amazing products, amazing experience. To customers. So what we built is, is ProfitMind. And the way to think about it is a kind of a 360-degree analysis of the business. It looks at pricing, inventory, assortment, looks at the competition in real time. It will go out and collect data on your competition, match their products to yours. You have real-time visibility to their assortments, their prices, analyzes marketing. And also analyzes the merchandise plan. So it goes through the plans. It will help you adjust the open to buy, find opportunities where you can have rebalance your plans. It'll look at every metric of the plans and give you guidance and advice. And we like to say it's not software because it's truly intelligent. So the way it works is you feed in your business goals in language. What am I trying to achieve? How am I trying to win? We bring in your business strategy and your unique business rules. And then your data. So we bring in all of your available data. And then what ProfitMind does is it goes out to the marketplace and it collects data across your competition and matches that. And then it takes that and turns it into recommendations for you. Prioritized, quantified, finding sales, profit, cash flow that's tied up. Gives you real-time insights on your business. So strategy, pricing, assortment planning, inventory, and marketing, as I said. So the idea is instead of pulling reports and data on a Sunday afternoon and spending all day Monday going through your business, you come in at 9 o'clock on a Monday morning and ProfitMind's already done that legwork for you and now you're picking and choosing the opportunities that you want to take for your business. So it's truly an AI intelligence layer and I've kind of covered the scope of it. And I actually, as I was designing the UI, so it was a lot of kind of rigorous research and building, as we were on the first to build this kind of agentic AI approach, I went out to all my merchant friends and I said, I know you're gonna be the busiest users of this tool, so design the UI for me. I do not want this to be another report that you feel like is just an and. This needs to be a replacement. So what we basically did is we went and sat down with merchants and we said, okay, how do you want to digest this information? How does it, um, how can you use it in, in terms of basically shortcutting your day? And so the way to think about it is you come in at 9 o'clock on a Monday morning, it's already gone through a strategic evaluation, it understands your business goals, and it's going to tell you where you're executing well, where you've got opportunities against your business goals. It goes through and actually identifies, uh, the actions that you should take and quantifies every single action so you understand what the profit, the sales and the cash flow impact will be, and then it prioritizes everything. So one of the, the merchant team said to me was, give me this landscape of the opportunities, but then tell me the 5 things I should go and do today in priority order. And ProfitMind literally will do that and it will tell you like, here's the first thing you should do, here's the second thing, here's the third thing that's going to move the needle on your business. So we're not spending time on the 80% of things that drive 20% of the value. Come in in the morning and we're focused on things that are really going to move the business. As a result, the results have been incredibly strong. So we rolled it out with 8 enterprise-size retailers last year. Average growth was 20% using the platform in sales, 14% profit growth. And it's largely because those retailers couldn't see these opportunities because they're just buried in data. And our system was able to sort through, get them to these opportunities much faster. We're starting now to get some reward— awards and recognition for what we've built. And I'm almost out of time, so would love to show you more. I don't have a booth here today, but would love to follow up with you and feel free to send me an email or grab me today and I'd love to show you a demo of how it works. Thank you very much.

Yes. Thank you. Thank you.

Okay. Good morning. My name is Anthony Volpe. I'm with Sisio. Sisio assists you in executing at a local level. And why do I have this pink hanger?

That must be—

most important question, like you're all thinking right now. The reality is that no matter how good we are at planning, the success or failure of that plan happens where? Right, at the moment in time, at the point, the location, right, online, where that customer, where she comes in and she needs that product she wants, where supply meets demand, where execution of our strategy actually really, really matters. In fact, over the last 20 years or so, I've been working on this and similar problems, and the best reminder of this came from one of the legends of your industry, our industry, Mickey Drexler. Has anyone worked for or with Mickey? Do you remember how he would come on the PA system like it was we were in high school, like the principal? I am at a store in Columbus, Ohio. That's where he was, right? And I'm here with, who are you? Catherine, Catherine. Catherine's here to look for this skirt. This is a beautiful skirt. I don't know who designed this. This is beautiful. I'm so glad it's in our assortment. But we don't have the right color. We don't have the right size. Somebody do something about Catherine. I'm gonna give you somebody's phone number. It's really, that's a true story, right? That's what he did. Day after day, a reminder of all of the things we do in planning, no matter how good we are, have to get executed at local levels. Now more than ever, you've heard this over and over, this is the drumbeat, we're under pressure. We have less inventory, right? We have less open to buy. We can no longer do what planners have done for a long time, which is those execution guys aren't always that great. We're going to have a little extra. Think about the sell-throughs you planned. Think about the sell-throughs you planned for years. That's what's going away now. We have to be more precise and precision in localized supply, the local execution of our broad assortments is what it takes. And unfortunately, too often the planner's role ends when we get what? Here are the products, here are the quantities we're buying, and our belief around distribution. These are going to A stores, or they're going to hot stores, or they're going to hot A stores. right? That was the end of it, right? And somehow execution happens. And guess what? Sometimes your plan was right and sometimes execution messed it up. So what we do, did I go the right way? Sorry. So what we do at Sizewell is we take the plan, whether it's from any of the vendors that are here who are doing exceptionally well giving you AI to make smarter plans, or those who are not here, when you get to the end of the plan, what happens? Generally speaking, we've seen that people throw it over the wall to some junior allocator, assistant allocator, somebody is providing the details to their supply team, their sourcing team, their allocation team, whoever it might be. Place the order, here's how much we want, and you guys figure it out. Is that true? And guess what? I think Steve did a great job early in the morning talking about that's what planners have sort of had a legacy of doing and that's what they want to be a part of. But there's a whole lot more that happens after that that they have no control over today. How does presentation quantities affect that plan? Right? That's hard. How do size ranges— are we able to have different size ranges in different locations? Are we stranding inventory in the wrong DCs in the wrong sizes or wrong colors? That all happens on the execution side. You might all think to yourself, well, allocation will get it right. Isn't that what we kind of have relied on allocation for for years? But the reality is allocation can't do a great job if you didn't buy it better. Right? I think Jennifer, you said earlier today, localization is where AI really hasn't made as big an impact. And that's ironic because that's where there's a lot of hard work. Why don't we go down the size in our plans? Combinatorial explosion. It's just huge. It's too big. I can't get to store size, but we can execute there and that's where AI should play a role. So walk with me for one second. How do we go from the left side to the right side? What does Sizeo do? We take that plan and we take your vision for distribution. Hot A stores, 3,000 units. We figure out how would you actually execute that ideally? What stores will it go to knowing you had Hot A? What are the sizes required? What is the presentation vision you had? What do you have strategically around customer experience and customer service? Do you always have to have one in stock? How do you think about deliveries and flows, replenishment? We think about all of that. We divide it up according to deliveries. We figure out the sizes. We figure out even if you have a packing strategy, how to pack it so that when the store opens that box, they're floor ready. You execute the vision. We can adjust your buy volume and say, you know what, that's a number not quite right because you wanted to achieve all of this. Can we adjust up or can we adjust down? We do all of that. We could do bulk alongside of pack across different deliveries. All of that is thought through at the time of buy-in. So we could correct that. You have a chance to weigh in on it still, and you have a greater control, if you will, over execution. This hangar where it happens, you have a bigger say-so in that now. So what does this mean? Though we're a startup, we've been around 2 or 3 years now. Again, we've been working on this problem for a very long time. We have 10 enterprise customers and we're seeing amazing results. Why? Because no matter how good you get at planning, no matter how much your forecast improves at the item level, that's not what's on here. This is a size, this is a color, this is at a location. And so when we change what would otherwise be a markdown because we have it wrong and we put it the right product on this hanger, you get a full sale. We're doing more with less. That's inventory efficiency, inventory optimization. We're seeing gigantic increases there. Right? And that goes right to the bottom line. You already invested in the inventory. You invested in the inventory. Let's get a full price sale. Let's serve a customer. Let's meet her demand. Right? With the same level of inventory. And then of course for those of you who also are comped on or measured on the supply chain side of this, I know this is always debatable whether that— how that affects planners and buyers and everyone else. We're dramatically improving supply chain operations and costs all at the same time. So that's the story of Sisio in terms of how it works. In terms of where we're going, the biggest opportunity we have is to push insights from downstream execution to upstream. That's what we mean by AI-enhanced planning teams. and someone said that this morning, I think Sarah maybe, I don't recall, how we need not to replace planners, to enhance them, to contribute to their decision support. So the more you know about execution downstream, the smarter decision you'll make upstream. We also wanna automate those things that you don't care about, that are hard, difficult, time-consuming. So come visit us, we have a booth here. I'm one of the founders. We have Gray, our CEO, we have Jameson. We'd love to talk more and hopefully get you on Sisio. Thank you.

So if I'd like to welcome the panelists back on the stage. Thank you. So super interesting. I think you can actually see what's going on out there in the world that can actually support the conversations that we've been having, which is great. So now we're going to open up Slido. Please scan the Slido with your questions and please reference them to the company that you would like them to be answered from, and then we will start the Q&A for the relevant companies. So please list the company first for who you would like the question to be answered from. Thank you. Okay, so there's two questions without the company, but maybe Someone's brave enough to shout who you would like that to be answered from. So how long have you been developing the machine learning models, question mark? Was that for a particular company? Okay, I'll open it to the floor. Please.

I can, uh, is my mic on? Yeah, I can go, uh, go ahead here. So we have been, uh, developing our ML models from the beginning. So actually we started out working on developing ML models and not really having like the software layer on top. But just, you know, crunching all the data and then providing back the results. We started out with two specific brands and it was like they thought they did a really good job on their Never Out of Stock program. So they said like, what's really hard for us is when we have new product introductions. So we want you guys to try to take a look at that. And yeah, that's where we started. So we did a lot of things around that. Did a lot of like matching products in between, but also looking at what wholesale customers are buying into this, what signals can we get from that, and which, how all of the different seasons are performing in the past. And we were able to, I see there's another question here on the forecasting accuracy. We were able to do a reduction in forecasting accuracy on 60% for male clothing and 44% for female clothing. Female clothing, more colors, more categories and stuff, a little bit more complex to forecast, especially new products. But we're really, really happy with that.

Yeah.

Yeah.

So similar, like we started about 2 years ago. The key difference from what we did was we actually started as an agentic, basically demand forecasting model. The good news was we were pretty new, so we were able to do— and there's a difference. Only difference between the previous models and what the agentic model does is it can reason itself as compared to Then we had the statistical model when we had the machine learning models and then the agentic model, which kind of can reason with itself. So we kind of had that benefit. And to that point, how do you measure it? You kind of look at MAP and, you know, mean average precision and kind of look at that and see how the new model is performing against whatever the customer is already using, right? So core of everything is that model. And then what we did is built agents on top of the model, which kind of solves the specific issues.

I'll just add, so for ProfitMind, we started building the models in August of 2021 and immediately went in and started working with Walmart on their data to build out the models that we have under the covers in the system.

Great. So we have 6 and a half minutes left, so we'll kind of fire around quickly. So ProfitMind and Hakio, what is your forecast accuracy?

Well, actually, I just did an analysis of this with one of our retail partners. So we analyze the merchandise plan and we were trying to look at our accuracy. And so we were able to achieve between about a 15% and 20% accuracy on a fashion category, which we found is on par or slightly better than the planners were able to do. But now we're able to automate the forecasting process for the planners.

Yeah, we, we measure actually error instead of accuracy because accuracy is often like, like a little bit harder to measure and like it can be discussed how is the right measure of accuracy. So when we— and error can be as well, but we look at the absolute percentage error. And when we looked on new products 6 months in advance, so that's for our customers the timing that they had to make decisions 6 months in advance, we were able to get the forecasting error down to 27% compared with 85% on their own.

And I was quoting error, by the way, as well.

Yeah, and it's about the continuous process also, right? How to improve the, reduce the error percentage.

Just to add on that, I think MAE, like mean absolute error, is one of the hardest things to sell to the company, just to share, because you have WAP, MAPE, MEA, right? And I think MEA is the hardest one to sell because they're like, but what does that percentage error mean? But of course, if you're looking at option level and you're looking at, say, demand forecasting, that, that one option you predicted 1,000, that one option you predicted zero, and it was actually the opposite, then your MEA is 50%. So then it's like, but you were 50% okay, but it's 100% wrong on both because you're stuck in oversold position. So I find the MEA conversation really interesting that Some people like to see WAPE, but then it's an average, like to see MAPE. So I think it's interesting where to, to financially calculate that, um, mean absolute error is also an interesting one as well, right? So just to add from my, my side, um, so next question: how does intello.ai connect to data sources? We have Databricks and standalone data in Qlik and YPLM.

Yeah, so that's the kind of one of the reasons the deployment time is faster. We are able to connect to data sources through APIs. We have interestingly, I don't know who called this out, so we have Databricks instance internally so we can connect to literally any of the data sources which are out there. We build connectors. So as part of building the agent, we also build connectors, for example, connecting to Oracle Retail or connecting to actually the PLM solution and things like that so that easy for the customer to just start in because that's typically the, the thing which takes the most time is how to get the data into the agent, right? So we try to kind of simplify it by building these connectors out.

Cool.

Next question, intello.ai again. Do you build agents outside of the planning process to help connect other areas like e-commerce, supply chain, et cetera?

Yeah, great question. When we started the company, our vision was to build AI agents in retail, right? And then we kind of quickly realized that There are too many problems to solve in retail. And we were like, we need to double down. And where we found the biggest pain points were in merchandise planning. So do our agents extend to digital merchandising? Do our agents extend to other pieces? Yes. But if you look at all our customers right now, our success has been in merchandise planning, including some of the customers who talked this morning where we talked about actual success with agents.

Great. I appreciate there's quite some questions coming through that are kind of hidden, but the one for the next one for Interlo, have you thought— okay, that's gone. Profit Mind, what is your implementation process like?

Yeah, we focused on this a lot. So my last job in retail was as a CIO, and I knew this kind of data challenge was going to be the, the make or break of the company. And so we actually apply agents in the data onboarding process. Um, and as a result, we're able to get up and running usually in 4 to 8 weeks, uh, not only with data ingested but all of the modeling done and first insights provided. So the, the focus is very much on a, uh, fast time to value.

Cool.

Um, two questions, generic ones. Who would like to pick that up? How do you help with revenue growth?

I'll take that one.

Sure.

Um, we help with revenue growth by ensuring that every unit of inventory actually is maximized in terms of its return. So we are looking to put the right item, that unit of inventory, at the right store, in the right size, at the right time. And so obviously, that's going to increase revenue because we're getting that full price sale rather than markdown. Very straightforward.

Yeah, our measure is pretty simple, sell-through rate, which literally eventually impacts revenue. So it's important to have leading indicators, not the lagging ones. Then we also look at out-of-stock. Those two measures then eventually impact revenue because it's very difficult to say, "I grew sales by 6%," because then it's like, "What part of your solution impacted sales as compared to everything else?" Great.

We have one minute left, so I'm going to ask one question for all of you. I think it's amazing, literally the questions are firing in left, right, and center. This is your opportunity when we finish after this, when we move to the networking one-to-one. Please go to the relevant companies. They would love to answer your question. So the last question for all: how much revenue does a customer need to make for it to make sense to use your product? So we have 1 minute if each of you would like to make—

uh, for us, we work with the large brands like Versace to basically smaller brands like Canyon Ranch. So from a revenue perspective, it, it doesn't matter. The key is to ensure that I think we all talk about it. More data you can feed in, more intelligent it is. So if I can get LY, if I can get LLY, beautiful.

Great.

Yeah. All our system does is focus on incremental sales, incremental profit, incremental cash flow. We have not had a customer that didn't pay for the entire annual subscription of the platform who didn't pay for it basically in the first month it was live.

So I'll do it fast from $20 million and up.

Right. And I would agree with that. We cater to the large and the very small.

Amazing. So thank everyone, to the panelists and the startups. Thank you so much.