Daydream On the Agentic AI Chat-To-Style Revolution

Voiceover:
Welcome to TD Cowen Insights, a space that brings leading thinkers together to share insights and ideas shaping the world around us. Join us as we converse with the top minds who are influencing our global sectors.
Oliver Chen:
Daydreaming about shopping, conversational commerce meets AI, meets retail. Really exciting new frontier here, and we're excited to explore this. Welcome to the Retail Visionary Podcast series, a podcast about visionary ideas and people. My name is Oliver Chen. I'm TD Cowen's new platforms, retail and luxury analyst. In this episode, we explore Daydream. Daydream's a next generation fashion tech company using Generative AI to transform how people discover and shop for style. Daydream is pioneering a new conversational discovery experience that helps users find the perfect item through natural language and image-based search. Today, we're thrilled to feature a legend here, Daydream's founder and CEO, Julie Bornstein. Prior to building Daydream, Julie founded and was CEO of THE YES, an AI-driven shopping platform acquired by Pinterest, and she was previously COO of Stitch Fix. Earlier, as CMO and Chief Digital Officer of Sephora, she launched Sephora.com and its iconic beauty insider program following leadership roles at Nordstrom and Urban Outfitters. Julie, it's a pleasure to be with you.
Julie Bornstein:
Thanks, Oliver. So nice to be here.
Oliver Chen:
I've followed your journey and you've always been one step ahead with what customer centricity really means. What inspired you to create Daydream, and what's the business problem you're solving for? We both love brands and fashion so much, I know.
Julie Bornstein:
Well, I am solving the problem that I've always want solved, which is help me find the thing I'm looking for, whether I know specifically what it is or I know generally what it is or I just have an event and I need something inspiring, and help me search across the entire web. I don't want to go to 20 sites and go through 23 pages to find something I'm looking for, bring it all to me. And so, I would say, the inspiration really started when I was a kid and I would get 17 magazine and go to the mall desperately looking for something that I saw and was inspired by. And as I've been in e-commerce since the late '90s, I've continued to watch technology evolve. When LLMs launched, I suddenly realized that the ability to combine natural language with the ability to search and find what you're looking for in the fashion domain was now possible. That is what inspired me to start Daydream.
Oliver Chen:
How is Daydream disrupting traditional online, and what's happening with the consumer's path to purchase? As we look at our research, LLMs are very powerful and scaling quickly, but there's still a modest percentage of total traffic.
Julie Bornstein:
Yeah, for sure. LLMs, I think overall, are not yet serving the fashion and shopping needs, but they're incredible and they're changing super rapidly, so anything is possible. But what I would say is that the experience overall has been pretty stagnant for the last 20 years. Having helped build Nordstrom.com and then looking at most of the sites today, there has been a dramatic lack of change, I think, and evolution in the way those work. I do think the big changes that have happened in the last five to 10 years is where people are getting inspiration. And so, if you look at TikTok and Instagram, obviously, those have become the primary sources of initial inspiration and probably a heavy source of what I would call impulse buying. But when you're looking to actually buy an outfit or wear something that matters to you, you're going to want to spend a little more time looking at your options and making those trade-offs and understanding what product to buy.
And so, where we think about daydream is in that center point. We're not handling checkout. Once you find what you want, you go to the retailer or to the brand and you can check out. But for us, we're really focused on once you have an idea, how do we help you find and narrow down the product that you want that's the right fit for you? And then, you can go ahead and buy it, and I think that has really become more and more overwhelming. The which and the where has become harder the more brands and the more sites exist. And so, it's interesting because early on when I was working at Nordstrom.com, we were very early in putting brands up on the webinar and there was a lot of fear around, "Oh no, do we want to sell online?" And now, the problem is very different. It's a problem of overwhelm. "How do I cut through all the noise and all the options to figure out the right thing for me?"
Oliver Chen:
Julie, could you give us an example? How does daydream work and what's the starting point? What's the most common/exciting use cases that you've seen so far?
Julie Bornstein:
Yeah. Well, you can start either by uploading a photo of something that you like or typing or voicing in a query. We'll take that second use case, where you basically can come into Daydream and you can ask for anything you want. I would say, the most consistent kind of query is I have an occasion. We see a lot of people coming in and going to a wedding and they tell us not just that they're going to a wedding, but where it is, it's in Austin and time of year, summer, and often daytime. And so, we'll end up getting about three to four different components in any given query, which is really interesting and a fun problem to solve. And then, we show you results, but we also ask you follow-up questions.
You can refine the results based on either if you see an item that you love, but it's not quite right, you can go on that image and you can say, "I like this, but I want it less low cut or solid colors," or whatever it is that you want. Or you can chat with the agent who is asking you follow-up questions. The agent may ask you, "Do you want something that's mini or more below your knee?" Or they might ask, "Do you want something that's dressier or more festive?" And so, we're trying to find out how this first set of results works and then where you want to go from there. Rarely do you find the perfect product in your first turn, but what we've done is we've given you a bunch of tools to help you refine and get to the right product that you want when you're starting at a broad search.
The other kind of query is I'm looking for a certain kind of things. I've been looking for, let's call it black block heeled boots that are one to two inch heel and under $500. And so again, those are the kinds of things we do really well at because the user already has a whole set of filters that are already in their mind, but you can't just go to a normal site and get there. You have to go through many steps because search on typical e-commerce sites don't work that way. And so, it's really fun to be able to try all these different terms that you want. We're working with so many retailers and brands that the ability to sort of get this incredible cross-section of product all available in one place that fits your need, and then you can start to refine from there is one of the big advantages.
Oliver Chen:
Part of the core of AI is building the right training set as well and iterating the training model with the LLMs based on the customer interactions, as well as the product and the data you're using. What have you done in terms of iteration of the LLMs and/or how has that impacted likely happiness of search and what have been really surprising to you and how people are using the conversational elements?
Julie Bornstein:
That's such a good question, and it's so core to how we think about building and improving the product. We launched in beta at the end of June of this year, and part of the reason that we launched is because we really wanted to start understanding what people were looking for and getting a sense of what is that edge case across the board? What are people asking for? How are they responding? Are they doing follow-up questions with the agent? Are they seeing and using what we call pivots where you can refine based on an image? It's been really interesting and very useful for us in a number of ways. I would say, the biggest takeaway for us was the range of questions is very broad, and in order to be able to answer these questions very well, we realized that we needed to move from a single system, where there was a single call to OpenAI, and we would basically get the search result call back and then go into our products to a system, where we have a whole series of mini models, where each model is kind of its own expert.
And so, initially, our search system was this single open call and we have now basically spent the last three months rewriting the architecture so that we have this series of mini models. One is, for example, around color, one is around neck type, one is around silhouette, location and every possible ... We're going through and training each of these mini models to become true experts at what their domain is. And then, when you get a question that comes in, we understand the query, we know which expert to call upon, and we have this combination that works faster because they're working in parallel and is much more refined and accurate because we now have trained our models to understand everything that matters in fashion. And so, the data that we've been able to get from users and understand what people are looking for has helped us to basically both refine the way our models are built, and then, refine the way we measure the success of those models and the success of the search results themselves.
Oliver Chen:
Yeah. So almost like a bit of decentralization and plus neural networks or parallel processing. What about the reality of like perplexity relative to OpenAI and the many choices? It's a bit of a race to have a better LLM. How are you thinking about that innovation opportunity?
Julie Bornstein:
We work with all of the LLMs, and we're constantly testing which ones work the best and the fastest. You're always trading off speed for accuracy. What we do is with each of ... we use LLMs in so many parts of our business, but if we're taking, for example, the mini model construct that I mentioned, we test each one against about four different LLMs and we see which ones give us the best results and which ones come in the fastest. And for the less complex things we're trying to solve for, i.e. color, we do the quickest one because it's an easier thing, whereas the more complex ones sometimes we'll lose a little on time to gain a little on its understanding of the space. And so, we are, I would say, constantly testing the latest models and we are very model agnostic.
Oliver Chen:
Yeah, you're looking for the best. I guess on that question, what is the best relative to the fastest and how are you judging best?
Julie Bornstein:
We actually judge best a couple of ways. We use LLMs to judge best. You can basically run models against a specific subject, so let's call it neckline, and you can say, "Here are the necklines." You can put into the model, "Here is what you should be getting." And then, you can have an LLM judge which one of those models get the most accurate and the highest number, for example, of neckline types. And then, you also use a human. We always have humans in the loop, kind of across the board in everything we do. We do spot checking and for the mini models, we can absolutely go in and we can run the results, run the query against the models, and then have a human look at the results, eyeball them, and see if they agree or disagree with the LLMs vote on which results were the best.
Oliver Chen:
That's going to be interesting and constant and changing. There's definitely human loops. Okay. The other question I'm always thinking about brand versus retailer and partnerships, because every hour we're watching new partnerships and we're a big fan of the speed and testing of what Walmart is doing. But deals with OpenAI, who signs up with you? Who's your customer between the brand versus the retailer? How do you make money?
Julie Bornstein:
We are both, we have retailers and we have brands on our platform, and we think about ourselves more in the space of like Google shopping. The truth is that there's quite a different assortment on a retail brand than there is on a brand direct, and we found that having both is great, both for having the biggest assortment possible and for helping the ecosystem as much as possible. We work directly with probably about 20 multi-branded retailers and then hundreds of brands direct. We have about 10,000 brands on the platform today as a result of those relationships with multi-branded retailers. Our business model is that we get a commission when we drive a sale, so retailers and brands only have to pay when a sale is transacted. Our integration is quite easy, because we are very aware having been on their side in the past, how many technical projects you have on your roadmap as a retailer and a brand. And so, we really do the heavy lifting for the brands and retailers.
The model is pretty straightforward. We can take a feed, we can take an affiliate feed or a marketing feed. We can also ingest the catalog from the brand and that gives us a really big selection. We add brands and retailers every week because to do personalization well, which is our goal, you need a very big assortment. And when you have a very big assortment, you better be good at personalization or also gets very overwhelming as a shopper. And so, we are working with both. We love having the catalogs of both and the bigger the catalog, really, the better we can be for users.
Oliver Chen:
Yeah. That's interesting. Julie, on your platform, the products that are shown are only related to the partners you have. Is that a true statement?
Julie Bornstein:
That is a true statement.
Oliver Chen:
Okay. And then, that begs another question, like what do you see, because I can ask this person anything. What do you see evolving with your assortment between everything you've done in your life like beauty, skincare, wellness, dresses, Walmart has hundreds of thousands of products.
Julie Bornstein:
I know. I think that to do this well, my belief is that you have to really focus deeply on the domain. I think that a Walmart and an Amazon and even an OpenAI, who is trying to build commerce that serves everyone, will do more of the spec-based product well, because if you can read reviews and you can understand the size or the dimensions of a product, then it's easier to compare them and that information is generally available. I think taste-based verticals like fashion and maybe home decor are a little more subjective and they require a deeper understanding of both the way the user asks for things and the nuances in that category and what someone might want to see when they're asking a certain way. And so, my bet is that to do fashion well in this kind of search, you need to go very deep.
And so, we are going very deep in fashion and sticking to fashion until we get this nailed. And to be clear, I don't know that this is ever totally nailed, but we have, I would say, many months ahead of us of tuning before we feel like we've gotten this nailed. It's taken such a deep focus on the fashion category that it's hard for me to envision anyone that's trying to take a broad sort of approach to making commerce work that'll work well for more subjective-based, nuance-based vertical.
Oliver Chen:
Yeah. You found that interesting challenge too with creating mini models too, specialization across these accesses. What do you mean by fashion though? How are you defining fashion?
Julie Bornstein:
Well, I think that fashion sometimes is a name that turns people off who feel like they care about what they wear, but they're not necessarily trendy. But the truth is fashion embodies high-end stuff. There's lots of people who care about fashion, but can only afford Zara, H&M, and they buy a lot of vintage. We, obviously, have high fashion. We are trying to cross price point, but we are actually focusing on brands. And so, we're different than I would say a Google, for example, is that Google is flooded with secondhand, it's flooded with I think Etsy kind of product and it's flooded with Amazon Fink brands and type of fake brands.
Oliver Chen:
Kind of a free for all, yes.
Julie Bornstein:
It's a free for all, and it's a mess and it's really hard to find what you want. And so, while we will certainly expand our definition over time, where we're focused to start is apparel and accessories from brands that are established brands. Whether they're entry point brands or they're luxury brands, they're trustworthy, they stand for something, and we feel when the shopper lands on, finds a product that they love, they can know that they will be able to order, they will get a brand that will send them a confirmation, will ship on time, is returnable, all the things that sometimes there's a lot of sketchy stuff out there, it's not always the case.
Oliver Chen:
Yeah. I was thinking when I was using it, what is your approach to value, and if somebody has a budget relative to not, and also just executing to offering the consumer a form of value, because that's so important as you know, like permanently important.
Julie Bornstein:
Yeah. Yep. I think longer term, our hope is to be able to incorporate all the promotional pricing that the brands and retailers are offering, so that that is incorporated into the search experience. What we're doing today is we are basically, we show product all the places that you can buy it on a single product page. And so, if two outlets are selling a product for different prices, you can see the different prices and you can find the one that's the best price for you. Over time, we want to include shipping price, we want to include shipping speed, because we know those things matter also, and value can be based on anything from, "I need to get it by tomorrow and so who can ship it to me the soonest to where can I find at the lowest price?"
We also can see if you tend to be more of a sale shopper and you want to get notified when things get marked down versus you're more of like a new shopper and you want to see as soon as the new things are released from a certain brand. What we're finding is people have different things that they're optimizing for, and so understanding that and being able to serve that up is important.
Oliver Chen:
Yeah. What about your competition, Julie? Who is your competition? A form of it is Google Shopping, but I think you're playing a really innovative role in discovery in terms of being a point of discovery.
Julie Bornstein:
I think that over time we will be able to incorporate more elements that really help with inspirational discovery. I think, for now, where we play is you have something in mind and then we help you discover options that could work for that. I think it's a combination of like search that exists today, so Google certainly is top of mind for me when I think about Google Shopping and where you go, where you can sort of see everything.
I think it's certainly there's elements of, you go to Instagram and you go to TikTok, search is just not great on Instagram and search on TikTok is great if you like low price stuff. It's all the cheap stuff. It's very hard to find higher end stuff. But certainly, I think those are becoming places where people start their journey as well. I think that the LLMs and people are starting to play with what happens when I search on ChatGPT for something. Today, I would argue it doesn't work at all, but certainly, we're talking to them and I'm sure they're thinking about this, so that will change over time too, and I think who becomes our competition in the future will probably change.
Oliver Chen:
Will you have a loyalty program too?
Julie Bornstein:
I don't think so.
Oliver Chen:
Basically, I think the promise holds, as AI gets to know Oliver Chen better, it should have a really strong view of me over time and all the things I'm dreaming about. But yeah, how will loyalty to your platform be handled? It sounds like you don't think so.
Julie Bornstein:
I don't think so, because we're not actually trying to interfere with the relationship between the shopper and the brand. I think that, at the end of the day, we're never going to hold inventory. You're always going to be getting your product from the brand's inventory, the retailer's inventory, and they have wonderful programs. I think what we can do is, if we know that you love Nordstrom's loyalty program, we can make sure to incorporate that in the search experience, so that you can see what you get from that and we can make it really easy to access that. I think we have some ideas around, you can have all the offers that are available to you, say in your inbox, you can just find them and they're automatically incorporated into the search on Daydream. And so, you have all that information, so we know you, we know what loyalty programs you're a part of, where you'd prefer to buy, what might be a good discovery for you, as well as what events are coming up for you and what things you should be thinking about in your future.
Oliver Chen:
Yeah. Julie, I've known you for a decade or more, I think you've been thinking about this problem for a long time. Why hasn't it been solved?
Julie Bornstein:
Well, I think that until the LLMs launched, the idea of being able to apply language to images has been impossible. And so, there, literally, wasn't the ability to say, take this conversation and then translate it to product images. And if you even look at the first wave of ChatGPT, it wasn't showing images on anything and it's still like showing just a few images. To understand the fashion vertical by the product and the image and description of the product takes a lot of training, and that training is leveraging LLMs. What we do is when we get the catalog from a brand, we don't just take the copy that they've written. We use an LLM to then understand everything about that product and we can take the language that consumers are asking for and we can put that into the products, so that the products actually show up in the right time when a customer is asking for something.
Search is a very, very hard problem to solve. Most of the search engines that have been built as third-party products for retailers are based on the keyword that the brand has put into the product and it's very hard to scale that kind of product search experience. What LLMs allowed us to do is ... And by the way, when I was at Sephora, my team ... well, some of my team, many of whom are still there, will remember, I got obsessed with this idea of understand the product and the data around it, so that when you're searching, it shows up. One of the merchandisers for the site and I worked together, we took every single skincare product and she basically tagged every single item with all of the attributes that matter when you're searching for skincare. What kind of skin you have?
Oliver Chen:
Yeah, she was like human AI, because she was like a human training like-
Julie Bornstein:
Exactly. She was a human AI and [inaudible 00:26:13].
Oliver Chen:
... I think you're going to ask this.
Julie Bornstein:
That was what we did. That was in, probably 2000-and god, what year was that? Like 14 or something, 2013. I realized back then, understanding all of the data points of a product, and not just the objective ones, but the subjective one, is like good for a formal event, is good for a casual event, all of those things are so critical to understand if you're going to know when to recommend them. What LLMs have allowed us to do is understand language, translate it, be able to build out these incredible models at scale and be able to look beyond the words to understand how to return search results and how to let people explain what they're looking for in more than two words.
Oliver Chen:
Yeah. The transformer model has changed the game in terms of language, and furthermore, you can do what you were doing in the 2000s at scale very fast with a few labels. What about the magic of people and you being a leader and having ... how many humans versus computers in terms of data science versus merchants and humans do you employ? Yeah.
Julie Bornstein:
We have about 25 engineers who are, some of them are machine learning engineers and some of them are just terrific engineers. They are, basically, building out the system. We have about 10, what I would call in-house stylists, who are helping with training the models. We do things where we will take a whole series of trend terms and we will build out collections of products to train the machine to understand what these trends look like. We actually come at trends in multiple ways. We use LLMs. We also use our humans to assign products. And then the visual really makes it easier for the machine to learn. And then we also do a lot of testing. We have, every week, we look at a whole series of queries and we do spot checking on both sides, both the search team and the stylist team does these tests, so we see what the results are coming out as, and we make adapt the searches where we see problems and we reinforce where it's going well.
We use, really, a big blend of human and technical to make this work. One of the things that I've always loved from a leadership standpoint is working with both fashion people and with technical people, and I think that that has been the cornerstone of my career, is being able to translate back and forth between the two. I would say, having a startup, where we have a combination of people who have spent their whole life on the brand and retail side, and we have people who have spent their whole life in the technical world, that combination is really, really powerful. One of the most exciting things for me was when I met the woman, who's now our CTO, and she has a deep search background. She's extremely technical, but she also loves to shop and she's a woman, and I said to her, "I didn't think you existed. I've never met someone like you out there." She's been an amazing partner.
Oliver Chen:
You found your dream. You found some optimized dream. You're like, "Where were you?"
Julie Bornstein:
Exactly. Exactly.
Oliver Chen:
Okay. That's great, so you guys have fun late at night a lot.
Julie Bornstein:
The truth is this world is, it's so experimental right now and we are learning so many things as we go. Our work is, it's very non-deterministic, which is hard for people who like answers, but we are constantly testing new things and gearing more towards the human, gearing more towards the machine. This is really a journey that has been super interesting and sometimes frustrating. But it's part of what is, I think, why this new world is so interesting is that it's very hard to know where it's going, but you need to build tools to start to learn and then you can refine as you start to learn.
Oliver Chen:
Yeah, we agree. I teach that class at Columbia business school called New Frontiers, where students of mine work in your organization. But it's a little hard with no answers or formulas plus the evolution of magic and logic, meaning both the pipes, but this whole notion of what experiential means and also cultural relevance, yet having very responsiveness and great supply chains and speed.
Julie Bornstein:
So true.
Oliver Chen:
What are your core competencies? I think I have an idea, but what would you say is your secret sauce at Daydream?
Julie Bornstein:
I would say that it is obsession with the vertical and being really, really understanding how nuanced shopping is and knowing that we can't just build a generic search engine that works for our product category. It is like, at the end of the day, marrying great technologists with people who really understand fashion shopping and being able to continue to think about and refine and make this product work to the point where exacting shoppers like me and probably you feel like, "Wow, this thing actually works. It actually understands what I'm asking. It's actually going to help me find what I want."
And I think that there's so many examples of people just putting lipstick on a pig and you go and then it breaks down pretty quickly. We are so deeply committed to understanding the needs of our users and making this thing work really well in this domain that I think that I just can't imagine anyone else in the world spending the hours that we spend around the table working on the things that we're working on. I would say an obsession over the details of making this very narrow experience work.
Oliver Chen:
You have some awesome partnerships and such a great heritage in the sector. Should every company or brand partner with you? Why would some choose not to? What's happening with the nature of adding?
Julie Bornstein:
No one should not work with us. We are truly here to help the industry thrive. My earliest memories are my love of certain brands. I would say that pretty much everyone that we've talked to has signed on with us with the exception of maybe two or three companies that I can think of. The reasons that they have not yet signed on with us was either they don't work with anyone, so it's the idea of having any kind of partnership is new and so they have to get comfortable with it, or they had other things they wanted to test on their own before they started testing with us and they didn't want to be first with us. But I think that we should, just like Google Shopping, everyone is sort of visible through Google Shopping.
What we didn't want to do is we didn't want to go scrape the web. We really wanted to create partnerships so that, number one, we could have a model where we could charge when there was a sale as opposed to advertising, because we think that limits the quality of the results for users. And number two, we want to build trusting relationships, because over time, we think we can serve additional needs that these brands and retailers may have as it comes to AI. We're also excited to share the data back with them, so we see that partnership as important for the longevity of what we're building.
Oliver Chen:
Very interesting. Topically, Julie, we're seeing this every morning or every day, another company signing a deal with OpenAI. What do you see happening there and how might it relate to Daydream? What are your views on this as we watch the news evolve quickly?
Julie Bornstein:
Here's what I would say. I would say they are a force unlike anything out there right now and like some of the very big forces that have changed the way the world works, but probably moving faster because the pace of change is so fast, it's crazy. They are definitely a force to be reckoned with. I think that we have really enjoyed working with them so far. We've worked with them mostly from the model side so far, but we are definitely having conversations about some other ways that we could work with them, so stay tuned. I think that their new approach to apps is really interesting. They launched a couple of weeks ago with an announcement really around working with 11 companies in different spaces, Spotify, Canva, Expedia, and so where they can build apps within ChatGPT, I think that's going to be an exciting, interesting area to watch.
They've just launched their new browser. That's going to be super interesting, so we're excited to continue to work with them. I think that they are definitely going to be a relevant player for, really, everyone in every industry. I also think it's a little too soon to know exactly where they're going next, because I think in some ways, they're also doing a lot of experimentation.
Oliver Chen:
Yeah. Testing and reacting and evolving. This also begs the question Amazon and Amazon Saks too. Amazon, clearly, has been making strides with adding brands and being a lover of the idea of having luxury on their platform and beauty. What do you see evolving there? You compete quite differently, but everybody's competition.
Julie Bornstein:
Yeah. I don't think of Amazon is competition in the fashion world, because the Amazon shopping experience is so not luxury and it's also just not set up to optimize for this vertical. I think it's an amazing place to go for a basic, although you have to sift through all their weird multi-consonant fake brands to find something that is maybe trustworthy. I think they've lost a lot of trust by the way that their search results work in the apparel category. I think it is a very frustrating experience when you see 10 of the same product with different fake brand names and all of them feel like they're just trying to keyword optimize to show up for what you're looking for. I think that Amazon has not really built themselves as a way that you could envision shopping for fashion, and I think that's very much echoed by every consumer we talk to who's engaged in the category.
I would say that I have huge admiration, I'm a huge Amazon customer. They've done so many things brilliantly, and I am worried, for the beauty industry, what's happening with so many of the beauty brands selling through Amazon, because it's a different vertical and it works differently on the Amazon platform than I think fashion and clothing do. I think, if you look at what's happened with the Saks deal, a lot of the luxury brands who could afford to say, "We don't want to be on Amazon," have said, "We don't want to be on Amazon." I think that reinforces the fact that fashion doesn't want to have to rely on Amazon to sell. It's just a subpar experience as a way to discover a brand and get inspired. I see a lot of opportunities still in fashion, where I don't worry about Amazon. I think Amazon is a force across the world on almost everything, so it's maybe a little arrogant to say that, but it's mostly my experience as a consumer and my experience talking to so many consumers and talking to brands.
Oliver Chen:
Yeah. You're also a beauty icon too. It's hard not to ask you. What do you think is going to happen in the beauty industry? We could probably do an hour on this, but what are some highlights about beauty?
Julie Bornstein:
Well, I have to say that I'm pretty proud of my former employer, Sephora, for this new idea of a marketplace. I think that it's a really interesting add-on for that business. I actually haven't even talked to anyone there yet about exactly how they're thinking about it, so I speak only having read the public stuff that they've written about it. But I think that beauty has changed probably more than any vertical over the last decade. When I was there, Instagram was, let's see, just getting going really as a course of discovery. And then, the fact that the brand universe has changed so much as a result of Instagram and all these brands around influencers have become so dominant.
It's a different field they're working in. And so, I think Sephora's done a great job of both staying up on who those important brands are and making sure to incorporate and pull them into their universe, as well as thinking about the fact that there are so many small brands out there, and consumers love discovery of beauty and the fact that they're opening up this place to discover above and beyond who they're able to sell in the store is a great idea.
I think that the weight of the influencer and the growth of those brands has been unbelievable to watch, to build billion dollar brands behind someone who just says, "I love this space and I'm going to build a brand," is pretty incredible. It'll be really interesting to see how long those brands survive and thrive. When I hear Hailey Bieber, for example, talking about Road, she sounds deeply committed to that brand and I think this length and success of that brand over time will depend 100% on her continued investment and engagement in the brand because it is so person-led. I think that's kind of always been the case with beauty founders, but a lot of the brands that were originally important brands were not as, some of them were makeup artist brands like Charlotte Tilbury, but some of them were brands where behind the scenes, there were very committed people like NARS. I think of some of the earlier brands are Make Up For Ever, but different people can come in and continue running that brand without the consumer necessarily knowing it. I think that space has just changed a lot as a result of that.
Oliver Chen:
Oh yeah, it's evolving. But then, there's new people too at different brands, but I see what you're saying. Okay. Fast-forward five years, what will Daydream look like and what are you most excited about? And final question, a lot of people listening to my podcast want to get smarter on AI. What do you recommend they do?
Julie Bornstein:
To answer the first question, what will we look like in five years? If I knew I would be worth more than I'm worth. However, I will tell you a guess of what I think we could become, which is really kind of a personal stylist in your pocket. My goal is for us to understand each user well enough that we can give you a few suggestions for whatever it is you need and those suggestions are either awesome or really close to what you want and you can have a quick dialogue to get to finding the thing that you want without feeling a sense of FOMO that there may be that perfect thing if you just scan 23 pages and look at that 24th page of results on a website.
We should know basically the different use cases that you may have. You may go to the gym and dress very casually, work from home, have events that you have to dress up for at night. We should know what's on your calendar to help you figure out what it is you need. It should all happen from your phone. It should be a few taps or a few conversations at most. Again, it should be like a personal stylist in your pocket. That's, I think, the vision that we have today. Obviously, we're starting with search because you need to really understand the product catalog and the universe of products and you need to understand the user well, and those are the things we're focused on learning today.
To answer your question on how do people get smart on AI, I would say number one, start using it. If you don't already, try it for absolutely everything. I think whether it's ChatGPT, it's SORA, it's starting to create some of your own, whether it is once Canva connects, create your own presentations. Today, you can start to talk to these chatbots and tell them what you want to build and they can build stuff for you. I think there are a number of amazing YouTube videos that show you how to do things. Don't be afraid. You can just ask ChatGPT, "How can I learn about using X, Y, and Z?" And it will tell you.
I think dedicating some time to really learning, grab a kid in your life, ask them what they're using it for, watch them show you. I think all of those things are really amazing and the learning curve is so fast. It just takes seeing it to understand it, and there are so many ways you can see it. I would say dig in, don't be afraid, and just start playing with it yourself.
Oliver Chen:
Well, Julie, it was such a pleasure to be with you. We covered a lot from really LLMs pioneering change, also how you think about the future and your curation expertise and what Daydream brings to the table to the future and these announcements and all the flexibility that people and shopping of the future is really going to look different with this conversational commerce. Thanks for your time.
Julie Bornstein:
Thank you. Thanks for having me. I really appreciate it.
Voiceover:
Thanks for joining us. Stay tuned for the next episode of TD Cowen Insights.