AI in Merchandise Planning | Calvin Klein, Versace, Nordstrom, Mattress Firm | PI Apparel

So amazing. So literally, I will allow to kind of the Slido to start now. And if you would like to introduce yourselves, just a minute each to explain who you are and what you're doing in the world of fashion. Thank you.

I'll start. Good morning, everyone. I'm super happy to be here. I'm Jen Chaffee. I'm currently with Mattress Firm, but my career has been 25+ years in merchandise planning and the tech as well. I've worked for companies like L Brands, Victoria's Secret, Pink. I've also worked with Talbots and JCPenney. But currently I'm at Mattress Firm and I'm helping them modernize their platform. We're changing everything about our demand forecasting, our inventory planning, our supplier purchase planning and visibility. So just happy to be here.

I'm Sarah Riker. I'm at Nordstrom. I've been at Nordstrom since, I guess, college, so 15 years. I started in stores, really fell in love with the culture, wanted to go into merchandising and planning, and then decided to get my master's in analytics, which was a better fit for me. And now I help merchants plan and forecast and buy. So it's kind of a full circle, 'cause I get to do the job I wanted, but with more my skill set. And it's been really cool to see Nordstrom kind of evolve over the years. And we're in a big transition period to move all of our planning to one system. So helping them with demand forecasting and evaluating and getting them to really trust the data science models and all that fun stuff.

Hi everyone, I'm Shilpa Sachdev. I'm the senior director of planning at Versace. I started my journey like 20 years ago, came for college here from India to America following my dreams and worked for many fashion brands, Tommy Hilfiger, Lands' End, Lafayette 148, and I've been— I'm now at Versace leading the planning team. So I'm excited to be here and sharing my insights with everybody.

Amazing. Thank you. So let's kick in with the questions, which is our angles of discussion, shall we say. So question 1, where is AI delivering real value in merchandising planning today? And where is the hype— the hype still ahead of the reality?

Okay, great question. I think where the most maturity is, and I think a lot of you all are probably also would agree with this, is in the demand forecasting area. That has really matured over time. A lot of us are using different models, are maturing the models that we're using. I'd also say automated replenishment has also been around for a while, very mature. I think in the middle, the less mature areas would be truly localized assortments, but being able to operationalize those assortments when you look at customer segmentation and how do you actually operationalize those. I think that's getting into a little bit more complexity. And then I think where, where there's still a lot of hype and probably not as much maturity is in long-range, long-horizon demand forecasting. So it's very difficult to forecast out 9 months, right? But I think if AI can get there, I think that would be amazing for you all that you work with your product development and your materials and your components. I think that would be amazing. But I do think that is probably a few years out. Yep.

Cool.

Sarah, a question for you. What internal challenges have you faced in adopting AI, whether it is data readiness, cultural buy-in, or team trust? If you could give some insights on how you've fared there.

Yeah, it's definitely been a challenge to get adoption across our merchants. We have— I mean, especially at Nordstrom, it's a very legacy brand. We've been around for almost 125 years, and our merchants really, really love their gut math. So getting them to trust a machine learning model or AI to predict the future is really hard. We've had some examples in the past where we have a model that helps plan our clearance and when to mark things down and understand that. Cycle, and they, we projected a negative profit margin for a certain period. They didn't like it, made the plan positive, and then were shocked that it came in negative. And then tried to blame the forecasting model. So it's definitely, there's a lot of education, a lot of, just empathy and trust that we're trying to build across the teams. We have a lot of new data scientists or technology people that come in and are like, why are you even measuring adoption? Why is that a big metric that you're measuring? Because we've struggled so hard with getting people to actually use the models and trust them. Then they're saying, well, it's the model's fault where it's not working. You're not using it, you're not using it to make the decision. So it's been a really long journey to build that trust. And then I think also data integrity is such a big thing. If you have bad data feeding into the models, you're going to have bad data and bad plans coming out of it.

I think just to share, like, we probably have, if I really talk about it, maybe a 15% adoption rate when you look at like what AI would've done or not. But I think it's like mindset change, isn't it? Like with everything in the industry, like I think some brands were kind of leading the 3D digital design and like by this proto review we're gonna have 65% 3D design, save time, money, blah, blah, blah. And then the new leadership come in and the next minute it's like, can we swatch every fabric and put it on a board? And then I'd like to see the sketches 10 times. And it's like, Jesus Christ, like, that was like 1999 or something. So I kind of— I said this actually at work yesterday. I was in the New York office yesterday, by the way, working a little bit. And, um, and I said to them, it's a bit like—

it's—

I really truly believe in 50 years' time, once like the graduates are now doing their stuff and they become the next VPs, SVPs, or heads of, I feel like then there will be truly change. But while those like I don't wanna say it, are still there, they will start their kind of like egotistical way of working. Like, you're on the third proto, but oh, actually we now need to do this. And it's like, no, no consumer cares. I mean, there was a story I heard yesterday where literally the 3-pack underwear was, I think, maybe I'm wrong, but like blue, red, white. And someone said, but shouldn't it be white, red, blue, or something like this? Truly, like that was what they said. I mean, you're gonna die, right? Quite slowly and quickly. So I feel like it's those conversations, and we're all in those conversations, right? Like, we all go home and think, Jesus. But that's kind of where we are right now with AI adoption. Like, everyone's comfortable with what they know. And even though it's like sustainability, it's like, excuse me for saying it, but 90% talk about it and actually about 2% is actually happening. And when you look at like, if you go to your CSR department and look at your like your adoption of, I don't know, this initiative, that initiative, this water steward is, all these types of conversations, I feel like it's more of like a marketing campaign than it is actually something embedded in DNA. And I think, and that's where like, oh, but price is now more important than it is actually sustainable fabric. So I feel like it, that's the thing with AI, no one really knows about it, they're not comfortable so far, and it, So for me, like you, the adoption, it's mindset change. That's kind of how I see it.

Yeah, and yeah, AI is such a huge buzzword, but if you ask somebody to define it, or if you ask 5 different people to define it, you get 5 different answers. So I think it's really important to educate and meet the merchants where they are and understand their struggles, what's blocking them. and start small. Like, if you don't need the most complex model in the world, sometimes it's just making something simpler for them so they can remove a task that takes them a full day. It can be, you know, delivered to them in their inbox.

Cool. Um, so one question for Shilpa, and then we'll head to the Slido because there's a few questions coming in. Please fire your questions in the Slido because we're very happy to answer your questions as as well. So Shilpa, how do you decide when AI should guide planning decisions and when human judgment should take the lead? And how do you help teams stay confident, creative, and invested in that actual process?

That's a really good question for me because especially working in luxury retail, we just implemented an AI tool. And that's something we thought a lot about and went back and forth. As speed is important for us, we want to analyze quickly, but also we have the trend nuances and other creative needs that we need to balance. So for us, it was key to think about speed and also the nuances of our business, especially based on a trend. So again, what we decided was— and again, through our discussions, we felt like we have to let AI lead where it excels. So AI is great at looking at data at scale, working on it quickly. So the speed is very, very important and giving us results that is— that could— looking at granular data. So for example, if you're looking at a style, color, SKU level at, at, by door, and if you have to forecast sales, allocate and also rebalance inventory with all other variables like rate of sale, sell-throughs, stockouts, seasonal trends, promotional cadence. There's a lot of variables involved, and that's where AI really helps us with all these multiple variables and analyzing and processing the data at speed. So that's why we trust AI. We trust AI to lead us. Now, in terms of the human element, Being in the luxury business, it's very important that we support our designer's vision. We also, again, store feedback. For example, like, yes, we might be allocating by size based on historical data. That's what the AI algorithm will propose. But for us, listening to our customers is really important, hearing them out. So if there's a particular size need in Bali, Harbor and they've never sold those— that size before, that's something we'll respect and we need to follow. Say there's a design trend and it's all about oversized bags and that's not something we've done well before, but again, we are going to support the creative vision. So that's where we have the override function on our AI tool where the planner has the ultimate control on overriding AI suggestions. So maximizing on the space speed that AI offers and then using human judgment. And that's where we feel modern planning is thriving. For us, again, getting the teams involved and motivated and taking them through the journey was really important and instrumental. We got them involved very early in the process. We got their weigh-ins. We showed them how the AI tool worked. We did multiple parallel runs where we ran the data manually as well as through AI so the team can actually, could build the trust and trust and understand why AI was making certain recommendations. So that was very, very important for us. And we also had the message loud and clear, AI was not replacing our planners. AI was empowering our planners. So that was key for us.

Yeah, I fully agree with that. I think, yeah, we hear it at work all the time, like, so how many jobs are going to go with AI? And to be honest, literally a certain person in the organization said, how many people and hours can we save if we automate with AI? And I'm like, just relax. Like, it's not all about cost cutting and making everyone happy at the top shareholder level. It's like, let's talk about empowering people to support their decision-making, not career kill them. So yeah, I'm 100% there. One thing I'd like to say— who, who's controlling the Slido?— is the top 3 questions are there from the beginning, but anyone who answers after, they're being hidden, and only the latest question is being shown. So I just want to make sure that everyone who gave a question is equally being able to have their question answered. So I'm going to go to Slido, I'm going to say the top 3 And then I will go to the last question we can see, but I think there's 4 questions hidden that the audience has asked. So hopefully we can get back to those questions. So over to the audience, which is great. So first question we will answer. How is AI-based forecasting different than traditional mathematical-based algorithms? Who would like to? Sarah?

Yeah, I would love to. This is such a good question. And there's such a wide range of, like, you have your very simplistic rate of sale, moving average type of forecasting. And then there can go all the way to, like, deep learning, very complex forecasting. And it's some, I mean, honestly, from what I've seen, rate of sale can be great, just as good as a really complex AI model. The benefit from a lot of AI-based forecasting is you can add in a lot of different variables and information and feed the model thousands and thousands of different variables, and it'll learn off of it where a simple statistical model won't pick that stuff up, like trends, macroeconomic information, information about different items. So like we're doing item-level forecasting at Nordstrom now, and it's very hard to forecast a new item. We have no information about it, but we can feed it about product information and it can pick up information from other items and learn from it where a statistical model is not going to understand that. Yeah, that's the biggest difference.

Yeah. And also I would say, yeah, the attribution, using attribution in your forecasting is amazing too. And one of the things I wanted just to speak about real quick with— at Mattress Firm, we're a little different. If you've been in our stores, we're more of a showroom. But one of the ways that we're going to be using demand forecasting with AI is to look at the correlation of availability of product, how soon can we get it to you, and any type of lost sale. So that was your first choice of product. Getting that to you versus selling you maybe something as a substitute. So that should be interesting for us. And I don't think we'd be able to do that without any type of AI component to help us. So we're excited about that.

I agree with you, Sarah. We are doing the same. So lost sales is really key. It's really hard to go down to this size SKU level by door and figure out the stockouts, closures, missed sales. So that we are excited about as well. And hopefully Hopefully that'll help us get to better sell-throughs.

Cool. So the next question: How has accuracy been with your AI models? How do you measure it? I'll lead with this one just quickly. I think sometimes in our industry everyone's like, let's talk about retail price and wholesale price. It sounds massive, like this is great. But I was very conscious that I didn't want to over-egg, as you say in English, to sell too much. So for me, it was about fabric FOB price on liability fabrics and also how much you spend on SMS. So I won't go into huge amounts of detail, but whoever's going to join my presentation this afternoon, I will bore you massively all about it. But it's very much about a four-way approach. So what would Merch do or the planner? What would AI do by itself? Because to be honest, AI can just go off the charts, right? And I don't think that's a bad thing because in range planning it could say, actually, you should have invested 30% in knitwear, not what the human said, which may have been 10%. So it's about investing in the future growth. But then you could argue, but actually you're just looking at 10 years worth of history. But what does the future prediction say? And I will also go into that as well. So it's all about kind of how you can utilize your current data, AI, then AI linked to a target, and then what did you collaboratively do, human and AI? And for the last 7 seasons across Tommy Hilfiger as well as Calvin Klein, so what I do now overarches both global brands, um, is what did those 4 approaches end up at? And obviously it works more in a wholesale model for now in a sense, because you're looking at your selling, uh, market close Of course it works in retail, and I'll go to that later on. But it's then looking at what did those 4 approaches say and what is the accuracy versus 100% sell-in. So what did merch do versus sell-in? What did AI do? Blah, blah, blah. And then you see clearly AI is obviously a lot more accurate than the planner. So then you look at, okay, well, that's good news, but how much money did you save? And then it's a conversation about stock and oversold because that's what it's all about. Fabric buys, global buys. Buys, vendor capacity. So you look at, oh, okay, so this is the amount times FOB fabric price times consumption, because of course you want consumption. And to be honest, in outerwear you can have 11 different fabrics, right? Or in, in tailoring you can have your sleeve head rolls and all your technical elements. I don't go that far, but it's more about grouping the main fabric FOB times consumption, and then you see what is the actual price. And already that's like multi-millions in savings. So then if you aggregate to wholesale to retail and sell the dream, as a business you're making like a shitload of money. Excuse me, my language. But so I really feel like that's for me the points of like selling it. So if I'm like presenting to the CEO and he's like, what are you up to? And I'll be like, well, I've saved this much money in the last quarter. And that's all they— well, I won't say it, but money talks, right? We all, especially in our industry, because it's like a bloody yo-yo. So I feel like money talks. So if you have a structured process and an approach and you can talk about the leverages, because you also don't want to throw planners under a bus, but you also want to say, actually, this is how we can complement the planner and get here. And if I'm really honest, to carry on, is workload is a massive conversation internally, right? Like, don't worry about the accuracy. If we can automate it, that saves us days. And I will talk about it this afternoon. So if anyone else has a point on that?

Yeah, I think accuracy is such an interesting thing. I'm in charge of our forecast accuracy for our new planning system. And I love that you're putting it into context of money because If you look at item-level forecast at a store level, it's like 40-something percent. It's not great. It doesn't sound great, but it actually in reality is relatively okay. And but like I think a lot of times people think forecasting should be, you know, as close to accurate as possible. And obviously the lower level forecasting you get, the worse the accuracy is going to look. So, definitely putting it into business context is really important, and then also it's like the time save and, you know, putting it as a baseline of compared to what we were doing before, we're actually doing much better, but we don't actually measure that in terms of accuracy. So, I think that one of my biggest, like, pushes as I'm working on this is that education piece and, like, don't focus on the accuracy. It's a guide. The forecast is telling you what you think you can sell, but then you're making all of these decisions by the time it gets to the store. So how accurate can you make that or make that correlation?

Yeah, just quickly, and also I think the human, which sounds really sad, but that element's really important because if it's a key look or a mandatory buy or a marketing campaign, the AI might say, oh, That shirt. But actually it's a marketing campaign. It's part of the color palette. The DNA is part of the windows. So it's almost like it will never take your job. It's just supporting your job because that's what it's all about, right? To efficiency, streamlining, and for me, cost saving.

Yeah, absolutely. I always say, because even in the tech space, people are worried that AI is going to take their job. AI is not going to take your job. People using AI are going to take your job because they're going to be more productive.

Yeah. Okay. Next question. How have AI agents augmented the work of your teams? If we talk about like a product development sense, there's obviously the whole 3D element of AI, right? Where it can generate automated key looks or like even catwalk campaign, like catwalk shows with product that actually look more more realistic than what our kind of showrooms are actually photographing. So I feel like in a sense of an AI agent, in that term, um, it actually automates a hell of a lot of work because you're not getting salesman samples, you're not going to the show, to the studios, you're not photographing, you're not sending salesman samples to the showrooms. It's literally like, this is the sketch almost, and this is actually how it can look in a flat shot, 5-angle, an actual catwalk image, an actual 3D live catwalk show. And of course you have to invest millions to get there, but it is— you do save money in the long run, that's for sure. So I feel, um, that is one way. And then in the other way, just automating and demand forecasting— automating a demand forecast in PLM is also like a massive win in a very quick way.

Yeah, we have a really cool example. I think something that we're going to be using AI for is in our inventory management. So right now we have 63 distribution centers across the country, and then each distribution center we have about 7,000 products. And then we today we segment those products, so A, B, C, D, and, you know, different, different letters mean different things. But a person today goes in every week and looks at every 63 DCs, all of those products, and they basically turn those dials. And those dials are based on weeks of supply. And we dial up and dial down the inventory. And we are so excited to move to AI-managed inventory where those parameters aren't something— we'll set them to start, but we won't have to have our users go in and just be in that minutia. I'm super excited about that.

Productive-culottes.

I also have an example for us, especially related to allocations and rebalancing of inventory. So before we had the AI tool, say we have a huge weekend Memorial Day sales and we come back to work, our— my allocator goes in, analyzes the data, figures out, of course, the sale— we have huge sales that past weekend and how do we quickly allocate and react to those needs by the store. So it will take her a couple of days to analyze and come up with a plan by SKU, by door, by size. So probably 3 days. And then rebalance as well because we've sold inventory better in certain stores and there's opportunity. So by the time she gets done with the whole analysis, most of the week is done. And when we send it to the warehouse, it doesn't ship until Friday or the following Monday. And the stores don't get it. So the East Coast doors will get it probably the next week and then the West Coast doors end of second week. Here we have lost two weeks where we are stocked out of inventory of our best sellers. So with our agent, we come in on Monday mornings and literally it's now a 15-minute task for my team. We can override it another 15, like where they go through and make the changes, and it's out to the warehouse by 10 AM. With all the internal processes, and it shipped the same day. So that's the big win for us, to be able to cater to our customer needs and fulfill the inventory needs for all of our stores.

Thanks. So next question, we still have over 15 minutes, so we're all good with the questions. So this is for Mattress Firm. How do you identify sales substitutes and include that in your forecast?

[FOREIGN LANGUAGE] Yes, so this is something we're working on now. And I would say product attribution, location attribution, all of the different things about the products that you would go to buy and the comfort, the feel, the price point, obviously the size. So we will look at all those attributes, probably regionalize it based on where you're shopping, and then understand what types of substitute generated the most revenue in the past for that type of combination. So, again, something very hard for an end user to do. You know, we have 2,300 stores. So, with AI, you'll be able to look at those attribution and basically those bundles of attributes. And then just what is the most profitable combination based on what we have available to sell and then what has been profitable in the past.

Thanks. So there's quite some questions coming in, so, which is great. So this is from Jennifer, I think I said that correct. If merchants' performance is benchmarked to last year metrics, sales margins, turn, adoption of creative solutions will be slow. How can you overcome these barriers? Anyone want to grab that?

Zara?

Yeah, Nordstrom loves to benchmark off of last year and it kind of drives me crazy. There's so many different factors that go into it. It kind of goes back to that gut math I mentioned before where they like, well, we did this last year and that sizing or clearance model I was talking about, like they were like, well, last year we weren't negative. And so I do think it does kind of hinder growth and understanding the possibilities of what else are out there. And like I said, for us, it's been a slow process. I've been on the analytics data science side at Nordstrom for 8 years now. And we've worked very hard to teach and educate the merchants and start to trust more data-driven models and also like meeting them on their terms, not— we were working on an in-season recommendation which helps them decide what actions to take in season. And we decided to take a really heuristic approach before diving right into like a machine learning model. So they understand that their feedback is getting looped into the recommendations. So we're building that trust to then overcome that, like, just based off of what we did last year or kind of those kind of like standard metrics that they're using.

Cool. Next question. What new skills are most critical for today's merchandise planners as AI tools become more embedded in decision-making? I think from my side, as crazy as it sounds, it's not kind of new skills. It's more like accepting. I feel like they don't need a skill as such. It's more like let them do their job. Well, to be honest, every merchandiser I work with is like, "I hate forecasting 'cause it's not accurate. I can't do it. And then I get a kick in the head for being wrong with the stock and oversell position because I gave a bad forecast. And then the vendors aren't happy because the capacity's not there, the quantities aren't there." You know how it goes. So I kind of feel like, With trust of the results and the value saving, they are like all over me, like absolutely just quickly put it in PLM, let me just say yes, and then we're done. So I feel like it's about roles and responsibilities in the new organizational structure, which is getting a bit deep, but it's more about like the merchandiser in a sense of where we work is overarching quite a lot, right? And the planner is more more just planning. So if you focus on a product merchandiser who has to do everything, I mean, they just want to work with product, right? Like, you know how it is, like the talk and the product and this and that. And I feel like— and then they don't want to sit in their seat and go, let's forecast every option. And in our organization, we have 650,000 different options a year. I mean, we're a massive company, Tommy and Kelvin under PVH. But they're all about like, oh Jesus, we now have to forecast. And if we're wrong, like, you know how it goes. So I feel like it's not about new skill sets, it's about how can AI support decision making, which I think is kind of now the, the running conversation. Yeah, come.

So I, like you said, it's all about having that open mindset and not fearing AI, being open and accepting to the new solutions, working with the solutions, challenging the solutions. Don't always accept what AI is just giving you. Challenging and making it better. And I think that's the way to go and basically adopt it, have that open mindset.

Yeah.

Yeah, I think, agreed, just being open and curious to what else is out there. I think people just kind of get stuck in what they know and it sounds scary. But definitely curiosity. I don't think you need to, you know, fully understand. I think it's good to understand, like, how it works, but not, like, the nitty-gritty. We spent a lot of time teaching our merchants SQL so they could, like, get data they needed. And I— that was really helpful to, like, stop the, like, one-off requests, but it also, like, gave them, confidence and also understanding of what actually the data looks like and how it's structured. They don't need to be experts, but it helps break the barrier a little bit to see behind the curtain a little bit.

Yeah, and I was just going to comment something very similar. So back when I used to be a merchandise planner, you always had to defend your plan, right? You own that plan. You had your binder, for those of you that are as old as me, and you'd go to your Monday morning meetings and you'd have to defend your plan. I think with AI now, it's yes, having a view into the data and the why and being able to explain the why, because you'll still have to defend your plan, but you also need to speak about, you know, how and the why. So it's not just, it was a black box, right? And now I have my demand forecasting. We can't answer that way. So I think it's just understanding the why and the data structure. I love that.

Yeah. So next question, is anyone concerned with sustainability issues with the increased use of AI in terms of the amount of resources used for AI? Um, I think that's quite a wide question in a sense of, is, is anyone— it was anonymous, but is anyone brave enough to call their question out?

I do have an answer.

Cool.

Like I think about, I, I use ChatGPT every day as my like sounding board. It, I have trouble paying attention in meetings, so I throw the transcript in, ask questions, and I'm like, am I, you know, killing the world with all the resources and the water it's running so I can remember what happened in a meeting that I was in. So it is something that I am concerned about and probably should be a little bit more aware. Not all AI systems are gonna use as much resources as, you know, the big large language models, but it is something to consider in the impact it has on the environment.

Yeah, I think also the reason why I say it's a wide question, 'cause I also look at it in a different angle about systems and resources in a sense of people because, when we use AI, we use everything that's already in the business within our systems. So it's not like we do anything different. We look at maybe 25 datasets, 10 years of history, and it's all what we already have in our business and the people. So it's a bit like, yeah, I mean, if, if an ERP is a sustainability issue, yes, but we already have it within our business, within our structure. We're just literally leveraging that we already have in multitude of systems. So sustainability-wise, we are kind of using less people, we're using less time, we are automating based on everything that we already internally manage in our systems anyway. And honestly, you could argue 47 people do it manually and one data scientist automates the AI, which is more accurate. So in a sense of sustainability, Maybe I'm wrong, but I feel like it is more sustainable and quicker. But of course, if you look at then macroeconomics, or you want to start bringing internal things into the conversation, yes, maybe that's a sustainability issue for the future.

Yeah, it's interesting how you took that differently, but I agree with your context within a business, it's less resources, and they're more productive. But yeah, globally.

Also elevated roles because of that. 100% that yes, it is more sustainable in terms of applying AI in your company because like you said, the data is the same. People can utilize that time to be more strategic. I have my allocator now being promoted to a planner, and that's a great story. And we are backfilling her role, but this is also growth within the company and doing better projects. And learning, and they're excited about that. So it works out.

Yeah. One thing when I think about sustainability, yes, we're using more computing power to do all of our prompts and analysis. But if you think about how we're using this technology and the savings that we could get on just manufacturing and sampling fabrics and just all of the cycles that that takes, And all of the great technology that's probably in this room right now can help tell you is your— what your trend will be like, how well received it will be to your customer, the components that you need and the sizes that you need and where you need those sizes. So I think it's— yes, you're using more resources on one end of the spectrum, but then you're saving probably so much more over time. Just with the output of what you're generating.

Cool. So we just have a— we just have over 4 minutes left and there's quite some questions. So we're going to do a bit of a fire round. So one question per person. So is anyone using generative AI solutions in their planning business, or are we all still more focused on application of ML models when we are talking AI? So Zara, quickly to you.

More machine learning models. We're using GenAI more on our like digital customer-facing like recommending looks and everything on our website versus in our planning.

Cool. Shilpa, how are supply chain hiccups impacting the benefits of using AI in merch planning allocation? Has your org adopted AI in supply chain operations?

So I'm not involved in the supply chain aspect of things, but Definitely what has helped for us is when we have quantified our lost sales and figured out the true demand, we are able to get inventory which is well stocked and covering the needs. And that's helped with our safety stock and our supply chain in the back working and getting that for us. So—

Great. Jennifer, have you incorporated AI into managing the growing returns inventory and QC processes in an omnichannel approach?

This is actually one of the use cases that we're just looking at now. So yes, we do have what we consider outlet product. So as that outlet product as a percent of our inventory grows, we have to understand what supercenters to send it to. So again, a lot of attribution, size of product, region of location. So yes, we'll be using it for that.

Great. Curious what the specific AI tools that you have used are. Would love to hear the ones the panelists have used. Thanks! Exclamation mark. So from my side, we developed internally. There was a lot of external, internal conversations, build or buy, which is a new phrase I found out recently. So yeah, so I feel like if you have in-house capabilities, great. But I really believe, like, if you're a new brand or a new startup, like, there's phenomenal things that you can buy nowadays. And, and I think the biggest conversation is— and that's how you have to sign budgets off, right? Is that, oh, but that's so much money. But it's like, but have you seen how much money you spend on liabilities? Have you seen how much money you spend on air freight? Have you seen how much money you spend on salesman samples? Um, did you know that Sharon just canceled 60% of the line and got it redesigned. Like, that's multi-millions in our business. So when someone in budget level questions, oh, but why would you spend $10 million? It's like, because they'll save you hundreds of millions over like a 5-year CAGR. So I feel like you got to come to the table with like a real business case and plan. Otherwise the old school, as I refer, is they're not going to get it. And I kind of feel like there's so much out there nowadays is that, yeah, if you don't have in-house capabilities, I mean, you can buy it. And yes, it's money, but you're gonna save in the long run. There's absolutely no doubt. And I think we're all firsthand going to tell you that that's the case for us as well. Okay, last question with 1 minute and 5 seconds, which is amazing. Any advice for retailers that are even further behind from a process and technology standpoint on how to begin shifting an organization into AI adoption. So I kind of feel like we roughly just minimize on that one, which is build a business case on— yeah, right. On like, I sat there literally and planned like for a full year '22, what did we lose on stock, oversold, liability, air freight, and just bad decision making. And you'll be super shocked if you can get those results. And be like, and then, I mean, look today, right? And I'm not selling, I hate that, but look today, you can speak to these providers, right? Who will literally be able to save you millions. It's just like, you gotta make the move, right? And that's the best business case way.

Yeah, definitely, I agree with you. It's all about solving a real business case. What is your pain point? What will make a difference to your team? And starting there. Yes, AI is cool. Everyone's talking about it. All the execs want to use AI, but it's about solving real problems that can make a difference for your team. So this will help them in part of— it'll be part of the adoption process where they would want it rather than us forcing it down. We all have consultants coming into our office telling us what to do, and it's really hard, whereas here it's It's about bringing the change together, championing for it, and getting that done. Also starting small. I would recommend take a small project, a couple of small projects, and then implement and test it. And that's what we did before going big and keep working on a bigger roadmap, but start small so you can be successful there. And of course, bring your people along. That's big.

I know we're out of time, but also Data Foundation is very important, so making sure you have accessible, clean data that is— will be able to be fed into these systems because without that, it'll be a real rough transition.

Jennifer, really quickly.

Kind of my thought, the last thought here is make sure— maybe go after things that are smaller that you can get quick wins. And when you get those quick wins, you can gain momentum with your users. And then those economic buyers, and you can really start to grow that momentum for your program.

Amazing, thank you so much. I hope you enjoyed that.

Thank you.

Thank you so much.

Thank you.