Victoria's Secret & Co. (VSXY) Earnings Call Transcript & Summary

Hello. right here, on this side. I'm [indiscernible]. I'm from India, Capgemini. You spoke of generative AI, you spoke of digital experiences. And this is the best space probably to ask this question. Have you thought of biases and how generative AI solution could deal with biases, especially the product that you are in where even human biases are so strong and so difficult to contend with, right?

Right. Yes, exactly Capgemini, and we do work with you in India so nice to meet you. We're worried about that, and we're just learning about that along with everybody else. But I think whenever you're worried about something like that, it's sort of you get what you measure and so what you're going to have to do is measure the outcomes that you're getting from the technology and how different they might be from outcomes that you would desire or be proud of. And so the more you measure, it's like -- I know it's going to be in those edge cases that we don't like the outputs that we're getting, that we're going to have to train the models to do something else. And that's really -- I can only think about this philosophically at this moment because we're just kind of going into it ourselves. But I think it's a great question for everybody. We're all going to have to work on that together.

Bill Scott, I'm with [ VTAP ]. Can you tell me how does your loyalty program fit into or complement some of the other initiatives that you've talked about today?

Oh gosh, I'm glad you asked that. We -- I will say our company was late to the party with a loyalty program because we had such a great credit card program for so many years that we've built rewards into that we felt that, that was really the way we engaged with our core customer. But we came to realize that there were so many great benefits for having a relationship with customers outside of the credit card that this year, we launched a loyalty program. And we launched it in June. We have 24 million plus signed up members just in the 6 months that we've had it live, which I think attests to the power of the brand drawing customers in and so many customers that have shopped with us for a lot of years are just signing up for it very quickly. Now what does it do for us? Well, we can start thinking about how we use our loyalty programs and now the e-mail address and the other methods of communication to be very targeted about direct relationships with those who have signed up with us, and predicting what behavior we think is most natural that they might demonstrate next. I'll give you an example of this. When customers start buying with us to begin with, they're almost always buy mist or panties. They buy something that's less than $10 in our last decile of purchases. And so what we think about is how will we walk them up the loyalty ladder to -- can we get them to try 1 bra? Just try a bra with us and see if you love it. So using this information, we get about our customers through the loyalty program to then guide them on shopping experiences for things that we think that they're going to like based on their behaviors and their psychographic segments and all those kind of things. So at our last decile, people buy panties [indiscernible]. Our top decile, they buy everything in the store. So we're just going to figure out how we deepen our relationship with them over time based on this loyalty program that we'll be able to communicate with them actually as much as they want to, depending on how much they buy from us.

All right. Chris, I think we've gone out of time. Thank you, everyone, for listening. [indiscernible] thank you.

Hello, everyone. I'm Amy Eschliman. I'm Managing Director of Retail Strategy for Google Cloud. We have an amazing group of panelists with us today. I'm going to go through a little bit of context about what we maybe talking about, and then we'll go right into the questions. So to start, I want to address the elephant in the room, generative AI. You might have heard of it while you're wandering the floor of the internet. Retail has been through this before. These massive transformation is nothing new to retail. The Internet is a great example, change the way we shop, mobile phones, same thing, change where and how we shop. And now generative AI, which has the capability of transforming everything from the customer experience to the associate experience. It's got tremendous potential, and it's a really exciting time in retail because of this technology. Possibilities of generative AI in retail are really everywhere. It's the ability to synthesize and analyze information that we did not have the ability to do before. So think about handwritten forms coming from the stores, product imagery. We're able to use that unstructured data in a way that we, again, have not been able to do before. Then the ability to generate content, massive implications in the world of retail marketing. When you think about not only creating campaigns but then also being able to iterate on those campaigns, improve them with the insights that you got. And then creating and automating processes. Again, every part of the retail value chain has the ability to have generative AI effect and really change the way that we do things. And then finally, engaging through conversations. We're not in the same world before of robotic chatbots. These can be incredibly natural conversations that make a customer feel like they are having a very personalized, seamless experience. So the possibilities are really endless. This is some use cases. It's certainly not an exhaustive list of use cases. But if you look behind me, these are the use cases that we at Google feel like really have the ability to drive more immediate value within retail. And within retail, time to value has always mattered. When we talk to our retail customers and look at the retail industry, 2024 is going to be a year of action. Of this survey, we had 81% of retail decision makers say that they feel urgency to adopt generative AI. And 72%, a pretty massive number are looking at implementing it in 2024. So the time is now. We ask how they plan on using it, you can see a variety of use cases, everything from the customer experience to the associate experience and making processes more effective. So customer service, the primary use case [ enchant ]. Product descriptions, how do you speed the categorization of products, the description of a product? How do you personalize that to different sectors -- segments. Creative work is a huge opportunity, conversational commerce was listed. And then also interesting, almost equal amount of survey respondents mentioned the associate experience as well. So not just the customer experience, but the associate experience. I know many of you in the room are at different places in your AI journey. And I think one thing stands true, no matter what type of AI we're talking about, strong data foundations are absolutely critical. And before you can move on to those really innovative use cases, you need to have that data behind you in order to use it in an effective way. So with that, I'm going to actually turn it over to our panelists, which I am really excited to speak to. I'm going to have each of you introduce yourselves. Murali, I'll start with you, and we'll go down the line.

Okay. My name is Murali Sundararajan, I'm the Chief Information Officer at Victoria's Secret.

My name is Jessyn Katchera. I'm leading e-commerce for the Group Carrefour and the innovation as well.

Okay. My name is Chandhu Nair, SVP of Technology for Data, AI and Innovation at Lowe's, a home improvement company.

Thank you, guys very much for being here. So I'm going to actually ask the first question of you, Murali. You at Victoria's Secret are leveraging AI and generative AI in lots of exciting ways. I'd love you to talk about how you're leveraging generative AI in the digital commerce space. Can you share more details?

Yes. I think like any other retailer, Victoria's Secret is focused on going through the journey with the AI and generative AI, so typically, we focus on 3 categories. One is the customer experience for the customers and the associate experience, both in the store as well as for the corporate office and improving operational efficiency across various functional areas. So the use case is the one we started with last year was the focus on customer experience for the digital one. We wanted to bring the customer experience what the customer typically go through in the store, and we want to mimic the same operations in the digital, that's what we were focused on, how do we can able to fill the gap. The example is when the customer walks into the store and then if she is interested in buying a bra, if the bra, the one she has used it for 5 years before, she wanted to come back and say, "I want it to be the same bra, can you help me out?" So the associate who's in the store are experienced, they know what exactly the question they need to ask. They know how to exactly work through the customer to navigate exactly what she's looking for across all the 50,000 SKUs that we have. But if you want to simulate the same experience in the digital, we're just exploring how we can able to do. So that's when we had a conversation with Google and Google is helping us to say, can we leverage the virtual -- Visual AI on using their Vertex platform. So we started the journey sometime in March. The use cases came in early into March, and then we can able to work through the prototype. We put it in production in June of 2023. So the reason I'm mentioning the time line is because this is how we can able to quickly do the prototype, see what the value is and then put them into production and then get some feedback from the customer because AI or generative AI, the feedback is very, very critical. And as you talked about, data is also critical. We need to understand what kind of data we get, what kind of experience we are trying to solve. So this one, we could able to literally see -- we saw the use case and we can able to bring it to fusion pretty, very fast and then that created a lot of momentum in terms of how do you build the use cases. And now we are focusing a lot more on customer experience, a lot more on creating more productivity and efficiency for the associate so that the associate can reduce their mundane work in the store, focus more with the rest to the associate for the customer experience, and we can able to improve the work output.

Great. And you guys, I don't know how you did it, but you sat in order of the questions I wanted to ask you. So Jessyn, this next one is for you. I know Carrefour is doing some pretty amazing things with generative AI from marketing use cases, even HR use cases. For the purpose of today's conversation, I'd love to have you talk to us a bit about the marketing use cases you've started and any results that you've seen thus far would be great.

Yes. And just to play back a little bit, I mean, similar to you, as soon as we heard about all the buzz that GenAI was carrying it around, we knew we wanted to be part of that journey. Why? Not because it's fun or because it's sexy but because we saw the potential it had to reinvent the way we operate to transform our operations. And so the first thing we did to come up with is what are those right use cases that we developed with Google for some of them. We started to talk with all of our operational teams to really understand what are the pain points today? What are the use cases we really need to solve and where GenAI can really play a role to transform those operations. And so by doing so, we realized there are 3 types of use cases that could be relevant. There are the everyday use cases like all the low value-added tasks that you can automate, that you can realize some small productivity gains, think about spending minutes automatically, following up on actions, smart composing your e-mails. And for all of them, to be fair, that's not where we wanted to focus because we thought there are a lot of brilliant players like Google and many of your competitors that will solve that for us. We just have to wait for the next release of those innovations in the tools we use on the day-to-day to bring that to the best of our employees. Then there was a second type of use cases, which was around where can we get a competitive edge because we're reinventing part of the experience where GenAI can help augment the way we serve our customers by being more personal, by being more tailored, by better empowering our employees to do the right thing. And so that's one of the use cases we developed there is around shopping assistant. And then the third thing, which I think is the truly crux of the issue for us is how can we reinvent end-to-end some big chance of functions that we have in our business and that's where marketing played -- entered in our reflection. We realize that there are a lot of functions that you can really radically transform by using GenAI to better empower your employees, to better serve our customers, to better propel our growth, and that's where we played in the -- in what we call the marketing studio. The goal there was really to think about how do we use GenAI to be smarter, faster generating assets for our marketing campaigns. Whether it's audio, whether it's text, whether it's visual. These are some of the use cases we're exploring. And why does it matter? It depends on the companies. For us, it's really about accelerating, the agility, the time to market. So instead of having to wait for many weeks to get a customer-ready product from a marketing standpoint, we can develop that in a matter of hours or a matter of days, which means you can certainly decide to expand and tailor the content you have to really feed the media you want to use to start to deploy local nuances to the content to develop globally because instead of having to wait for weeks or months to deliver the 1 asset at time, you can have done that in a few days. So you can start to really expand the reach and the personalization of your marketing campaign. And that's what we're really proud about and some of the use cases we developed with Google on that front and that are displayed in the booth are really exciting for us.

I've been in retail for 20 years. And I think we've been talking about personalization all that time. It's really exciting to see some of these use cases come to life because they truly change the way that we deliver personalized content. Okay, Chandhu, I'm going to send this to you next. You have been an early adopter of Google Retail search. I know you're using generative AI as some merchandising use cases. I'd love to hear your perspective on kind of what -- how it's going? Any results that you've seen thus far?

Sure. Thanks, Amy. Absolutely. But I'll start with kind of the common theme that's there. I think generative AI is not kind of the differentiator by itself, your data is, right? So that's -- I think what -- how I think of generative AI as to how it will unlock value for retail, especially is, you got to think about now data that we could not process before. Think about all of your operating procedure documents, the PDFs and things like that, that is out there that is typically hard for you to process and understand and really kind of build out experiences for your associates, for your customers or kind of help improve in generally your productivity as a company, right? So I think that's kind of the broader premise by which we started looking across the entire retail value chain and see where generative AI use case could apply. So -- and merchandising was definitely one of the areas that we picked on. And the reason I picked on is if you look at home improvement, we sell anything from appliances, you sell appliances very differently from lumber to paint. I mean, people curate. I mean, if you're a merchant, you curate those assortments very, very differently compared to like your next category of assortment, right, that you need to bring to the table, right? So what we started was with very basic stuff. We don't -- a growing problem that we had was when it come to product onboarding like onboarding a SKU or onboarding a vendor, it was a laborious process, very manual process. A lot of the data quality that is involved in capturing them upfront, it was not there. So it was kind of exposed to the customer or the store associate to explain that. So it was not that we never tried to solve that problem. We have applied a lot of AI models before to solve that problem. But now, with generative AI, we almost saw about 60% reduction in like the amount of manual labor that is needed to upload these initial product descriptions. Think about all of the attachments that needs to -- the images all that, that needs to go into that, right? So that's one part. Then we thought about, okay, how do we productize this. And that's a theme that we're working on because -- any other thing, and it does require a different way to think about productizing generative AI products. But now think about how you shop for a vanity in the bathroom, right? So if you go to a store, you kind of see a curation of things that actually tells you how it could look like. The backsplash, the faucet. I mean everything that is in there. How do you kind of bring that? How do you curate that in a very dynamic way. Usually, it is all very manual. There's somebody is -- a marketer or a merchant is working together to kind of pull all of that together. But with generative AI, you can actually curate that with a lot of -- not just with generative AI, I should say, we use generative AI alongside a lot of our model because it has to understand our data. We bring that together to curate what the vanity could look like, what are the options, right? So -- and we are in that [indiscernible] kind of [indiscernible] and that will be a customer-facing experience. So we have both associate-facing experiences like helping onboard products with better data quality, better images, better graphics to kind of help sell, help kind of help tell the message to the customer the right way. At the same time, we also want to kind of have the right experiences to our customers with curated collections of -- that can actually enable sales digitally or in-store, right? In-store or associates are doing that. So that's been the part. Now I'll kind of share some of the learnings though, as far as to how we applied it. To me, like, I think large language models, all the buses around generative AI, large language models are, in a way, commoditized. I mean there's a lot of options open source or otherwise to go and get different large language models. Like I said, the data is foundation. I think my job from a technology perspective is to look for what is the best and the most accurate model at the lowest price, right? Because these are expensive ones to run and you have to kind of figure out what are the right use cases to apply. So you kind of have to think about what is the most accurate model at the lowest price. To, like you said, your data is your foundation. And you have to have an ecosystem where your models will work in tandem with the LLMs to kind of give the right experience that is in there. The third learning is really in terms of when we talked about -- we are a product [indiscernible] structure in terms of our technology organization. But when we started to bring in generative AI, we have to kind of disrupt our model to kind of make sure that we can build the right experiences for the customer or our associates. The reason I say that is these models are not good at giving you definitive answers, right? They are 80% good, 20%, they can do things that we kind of don't know or they are not accurate. So you have to now work with -- you have the change in engineer's mind to work with something that is not very definite, but it's 80% there, right? So how do you bake that into your process of rolling out experiences to your customers. Now you have to also make sure that the customer is not exposed to that stuff. So those are some of the things, both from how we rolled out the different products from the use cases to different products and how we thought and continue to learn as we evolve this -- evolve into this journey, if that makes sense.

Great. You each talked a bit about prioritization and as you were describing those use cases, which is great. Thank you. I'd love to hear a little bit more like any advice you have for prioritization across an organization, how you get the organization on board with your -- the use cases that you've decided? I know you kind of addressed each question, but I'd love to go a bit deeper into that question of all the use cases you can do, how do you pick the one that you want to start with? Murali, I'll start with you again.

I think the right way to do is what you said is correct. I think we have to understand the use case scenario is what makes more sense. I think the reason we picked up customer experience is a critical one for Victoria's Secret is because that's where the value for what we are trying to do, and that's where we can be able to learn a lot of the insights of the customer, learn from there for certain experiences. But the other experiences, we are focused on, associate focusing. Because specifically when you go to the store, there are multiple situations where we can able to have an impact of the associates because in retail, the labor at the store is always the premium. How much have they been able to reduce their work behind the curtain, put them into the store, work with the customer to make them processing easier. That's want to make the life of the customer much, much easier. So we're looking for the business value, we're looking for where all we can add value back to the customer. So that's the priority that takes precedence over the other one. Doesn't mean there are other areas like supply chain or merchandising or finance, it doesn't make some sense. But we do them, but it's going to be a little different priorities. But for all of them, the focus is going to be the data and the experience that is going to have a value.

How about you, Jessyn?

I mean for us, it was -- I don't think prioritization for GenAI is very different to any form of prioritization of projects that you have. We approach it in a very pragmatic and humble approach. We don't know everything. That's okay. And at the end of the day, it was really about what is the size of the trial, like what is the total amount of value that you can address with your GenAI project [ or theme ]? How much savings do you anticipate? What is the technical feasibility given your own readiness, your talent pool, your ability to embrace that challenge. And the only additional lever is how advanced is the AI technology itself to serve that need today. Is it fully ready? Is it still explanatory? Is it like very early stage. I think you need to be very realistic about it because it changes every day. But it's not true that today, GenAI can do anything. I mean you can do a lot of things, but can it solve any problem to the same level of precision, same level of scale, same level of effectiveness from a cost savings perspective? No. So you need to be very particular about this. And when I said we wanted to be pragmatic and humble, it was really around not being paralyzed because you don't know everything. I mean GenAI is an area where you need to test fast, you need to -- you can test fast, you can fail fast and that's okay. And so you need to have that to embrace that mode of thinking into your organization. One of the things that our CEO helped us to do is to circumvent a lot of the traditional decision-making process to test our toes in the water so we can actually develop use cases in a matter of 4, 5 weeks for some of them, really see the results. And if we have to make adjustments, so be it, we're learning, and that's okay. On the opposite side, you need to be very disciplined because you don't want to start to be -- to think that GenAI is going to change everything they want and to have 100 use cases that you can't afford, you can't be disciplined about. And you need to be able to say, that doesn't work, that's okay, I'm going to change it. I'm going to kill that again, I'm going to move to the next one. And so I think that's really important to be ruthless in your prioritization every day. And you think about where do you have enough proof of concepts, so you really want to scale to the next approach. And I think the last thing regarding your question about learning is really about how do you embrace that into your ecosystem. And that's why we wanted for us to enable all of our people, whether they are on the frontline or whether they are at the headquarters to be part of the solution of identifying what use case we want to tackle because if you do that with just an innovation team and not connected to the business, you're going to move faster, you're not going to be able to scale. If you do that just with embracing everybody in the organization, you might be slow. And so you need to find that balance. So when you want to scale, people feel they're part of the ideation in the first place. So you have the right success factors to be able to scale. And what I truly believe, and that's where I'll end is there is a question how do you move from a moment to movements in GenAI. It's very easy to have a lot of wins that are moments of innovation. But what you really need to do is to create a movement in your organization. So everybody thinks at their level, what can I do with GenAI? You have the right expert that says, yes, it's realistic today. No, it's not realistic. Yes, we need to prioritize that because the size of the prize warrants that effort and then being very restless at picking a few use cases to go at scale. So then you can fund your next wave of innovation, and you can really build the movement into your organization that's going to be lasting. So that's the way we're trying to transform function by function at Carrefour on our end with the support of our CEO.

Great. Thank you. Chandhu?

Yes, I'll go on some of the themes that we shared here. But I think generative AI is one of those things where it's like a [ death ] by 1,000 use cases. Like every engineer, every product manager, everybody in the business team has a use case that they think can apply, you can apply generative AI. When we started off -- and Google held this with the Hackathon event, and we've got a few Hackathon events across to kind of get ideas around it. And soon we realized, obviously, like just I was saying, it's a massive list of things. How do we go after it, right? And the approach that we took is we looked not just from a generative AI, from a productivity play standpoint, we looked at both from a sales enablement, productivity and experience standpoint, all 3 things in the mix. Then we created a value to risk framework. So there is value in terms of like the financial outcome or other outcomes that it can drive. And then what is the risk? Risk, in meaning, is the technology ready for that particular use case? Two, is there an adoption risk from the users because of concerns around the technology? There are other risks around brand risk. You've heard a lot of the bad PR that can come around. It is from a brand perspective, how do we protect that. So we mapped out that across the value to risk framework. And then what we did is we identified core areas like marketing, merchandising, et cetera, and then we look to productize these use cases. So a very definite set of high-value, lower-risk use cases, for example, an associate-facing use case lower risk that we can pilot it before we put something in front of a massive set of customers, right? So and then we mapped those functional areas and created products like what would a Lowes.com AI product look like to support all of the digital activities and map the top uses around it. So that way you can manage it like throughout a road map, again, aligning that to the risk and the value it creates. So that was kind of how we are [ doubling ]. And we're continuing to fine-tune and adjust to it because we're all on it.

Great. Thank you. So in our last almost 5 minutes, you guys are each poised -- your companies are each poised to drive a lot of value on generative AI in 2024. Any advice for the audience on how they can get started and really put generative AI to work with that short time frame?

Things start with the data. So where the data is -- so wherever you have more confident the data is because AI model is common. I'm saying, AI model, you can use the model whichever you want, whether it's a traditional AI, generative AI, depending upon the use case but the data is not the clean one, then what they all communicate out of the generative AI is not going to be the one you want to test, right? So focus on the data, focus on the learning. When I say the learning, you have to keep training the model. The model which you generate on day 1 is not going to be perfect. You have to feed the data back and so that the model learns more and more and then you come to a point where they're just going to be more reliable. So that's what you have to do. So it's the iterative process. That's why it's just going to be more focused on the data, the process and the evolution.

Jessyn, how about you?

I would say don't get paralyzed because you don't know everything, just get started. The good thing is you can. In a matter of weeks, you can see the results of what you wanted to develop. At the same point of time, stay focused. So start with data, I agree with you and start with true pain points you want to solve. I used to be in the start-up community. And it's always about solving a problem. Don't try to do GenAI because it's fun or because it's cool, it doesn't matter. If it has an impact, it matters, if it doesn't have an impact, do something else. And so really think about what are the foundational pain points you really have to solve and whether GenAI is the response that is the more suitable for what you want to do. And maybe the third thing to add to just an additional dimension is be responsible because GenAI is a fantastic machine, and it opens a wide range of possibilities. But at the same point of time, we all have a corporate responsibility. We have a responsibility for our customers, we have a responsibility for our employees. And so it's important to now in advance, like what are the red lines you don't want to cross? And how do you animate that into your organization? Because at some point, it's going to be a spread movement. And so it's important to have guidelines to our guardrails to be compliant with GDPR, to think about data privacy, to think about also like what are the HR implications down the road because we all know that GenAI, in a way, if you use it for associates processes, it's going to make performance converge, right? And so how are you going to reinvent the way you do performance management to distinguish the top players, the lower-performing players? How do you make sure that your employees get smarter at using GenAI, not only by using better prompts because that's going to be automated at some point. But also, how do you make sure that they have the self -- they're able to step back to understand the technology to make sure that on the 20% cases where GenAI is not the best answer or where GenAI [indiscernible], they can identify it by self-criticism. They're being able to self reflect and say, [ shares ] there might be something I need to do or to adjust because otherwise, you're going to -- you're going to be much more powerful than the 80%, but you're going to be less powerful on the 20%. So some of those problems are going to be -- you don't have to say it [indiscernible] day 1, but you need to think about them so you can find the right answers for the company and build the right processes around it.

Great. And we'll finish with you, Chandhu.

Yes. Now I'll -- obviously, we cover data and the guardrails that needs to come into play. Those are fundamentals. In my opinion, really, data is a differentiator. The only 2 adds that I would do is in terms of driving -- to me, it's like a fast race on a tight rope for driving adoption is going to be super, super key in a technology that we are still trying to understand, right? So bringing in your business partners who are going to be your biggest change management agents into any of the initiatives that you're trying to drive is going to be super, super critical. That is one. Two, is involved a human in the loop. It's like the first time you get to -- back in the days when elevators kind of came into systems, there was a person inside who was pressing the bottom to get to up and down, just to make you comfortable in that. So because there could be things that could go wrong. And that is the human and the loop part of how you instrument this technology is good. Over a period of time, that human will not be there in that elevator to kind of bring you up and down. There's a voice that tells you going up, going down over a period of time. Even that doesn't exist, you can walk out and you can just press a button and you walk in, you trust it. But it is very important to get kind of that tribal knowledge and the support system that is there in a lot of the employees and the associates that work in. It's not a technology that has that view because it is not trained on tribal knowledges. It's trained on things that it can read of the Internet or what you're telling them as to what it is. So having a human in the loop and really thinking about change management and adoption right upfront, that's the only way I believe value realization will happen out of this technology, right? So yes, those are some of the -- some of our learnings and still evolving.

Thank you, each of you. We're out of time. I really appreciate all of you spending the time with us. You gave great advice and great insights into what you guys are each doing, what your companies are doing. So thank you very much.

Thank you.