Brand Agents: Grow and Protect Your Brand in the New Era of Discovery

Anne T. Griffin: Hello, hello, everyone. Welcome. We're so excited to have you today. I see some people I recognize, some Huge alumni as well. Thank you so much for joining us today for Brand Agents: Grow and Protect Your Brand in the New Era of Discovery. So today we're going to be talking about how agents are changing the way people discover products, or more so, how are agents going to discover products on behalf of their people and make decisions and interact with brands. There's getting to be a lot of excitement, especially if you're on X and you're reading all the hype and obvious reasons also for concern in terms of security or like what does that do to a lot of these channels that maybe don't feel as impactful as they used to be, and now if somebody else's AI becomes the first place your customer or their agent goes to, how does your brand stay relevant in this era and deliver an experience that's both delightful and something that is trustworthy? So I'm Anne Griffin, and I am the founder and principal AI product consultant at Griffin Product Co, where I help companies build AI products that are trustworthy and I'll say actually drive results and growth. And today I'm also here with the wonderful Chris Stempky from Product, Inc., and Product, Inc. is a brand consultancy, and then we're also here with Tom Rudczynski from XTAL Search, which is a search and discovery company that basically focuses on how to get visibility and discoverability, especially in this age of AI. And so we're going to talk through that shift and make it concrete. You know, especially we're going to talk through a shopping example that we went through the other day, and I think is really interesting and fun. And we're going to also finish with: where do you start? What is actionable after this webinar? And feel free to put questions in the chat as we go. However, you know, we may see some of them as we go, and I may call some out. But we're going to have a great Q&A at the end. Also, this is being recorded, so you will have the opportunity to get this recording later. So, with that said, I actually have a question for you, Chris. And it's really behind this quote. So you know, Anthropic is like, oh, in three years, brand websites won't exist. And I'm on X, so I see all sorts of things about people saying I declare, you know, there's not going to be a UI, and everyone's agents are just going to use your data source. And like proprietary data sources are one of the only moats, so you don't need this kind of brand website. Is that true? What is this actually trying to say?

Chris Stempky: So, I mean, we know with Anthropic, there's a lot of predictions. Some of them are further out, you know, a little bit more crazy than others. I think this was one of those provocations where I think brand websites are going to exist, but they're going to exist in a completely different format, and it's really being led by customer expectations. And we've seen through a lot of our digital transformation periods in time that it's really customers' expectation led by technology that are driving that. So will they be gone in three years? I don't think that's a reliable deadline. But I do think that we're already seeing that shift into that new interaction model between brands and consumers. And you know even just what a couple weeks ago with Meta launching Muse, you know now we have AI agents that can research products and they can compare prices and help you actually make the purchase. You know and we're also seeing some of that pushback where Amazon is blocking Muse, right, where Shopify is welcoming it in. So I think the website and brand website, there's still going to be a need. What they look like, I think, is going to evolve and probably more significantly than they have in the past. And I think that's kind of you know why we're here today.

Tom Rudczynski: Yeah. And when I heard this from the Anthropic employee, it was pretty shocking. The concept that this thing that we've worked in the last 20 years plus to establish seem pretty you know ingrained into the way that we use the internet, shop, and kind of live our lives. But like as we continue this conversation, I want to talk through like why I think that's the case. Why you know what changes are actually happening on the ground that are going to enable a world where brand websites have to transition into a different format because of the expectations that consumers are bringing, but also because of the fluidity and friction-free nature of driving everything through an agent that already has context about who you are and what you want, cuts out a lot of kind of the work in between that we used to have to do manually, and I think that's the tipping point that leads someone who's there every day, like this Anthropic employee, to say, I could completely imagine, you know, that this is gone and we've hit this pivot point. But I think that's optimistic, and as we're going to talk about a little bit later, it doesn't give the regard to the relationship that consumers have with brands today. And with that, I'm gonna you know just pass it back to Anne because I think that's the interesting tension that we're going to be looking at.

Anne T. Griffin: Yeah, and I think this gets into a really interesting piece is that part of brand discovery in the past is like I'll just insert a little bit of lore and history for those who are joining us today is Tom, Chris, and I used to work at this little agency called Huge, and one of the things that we always talked about was like removing the friction, removing the friction, but in reality, you know there is a little bit of friction in discovering. Part of how you discover a new brand is that little bit of friction. But also, if you are an existing brand, you want to actually reduce as much friction as possible, so that you're like, great, they don't have time to find somebody else. Like they're happy here, they stay here, but Chris, like especially as someone who really like you know consults with brands and like really understands them inside out, can you actually walk us through like this history of kind of like removing friction and like what does that mean in this agentic era today?

Chris Stempky: Yeah, and the nice thing is that sort of I've lived through all of these. I started my career in the mid '90s in San Francisco during the dot-com boom. So I've sort of been through this. And so when I first started out, I was you know a scrappy young 22-year-old, and I was working in direct marketing for a little agency called Times Direct Marketing, where we did a lot of mailers, brochures, mailers, white papers, and we were always asking for people to register for an event. Let's say, like come to our event, or you know, give us your information. We'll send you a white paper. And if you can imagine, the friction there was that we would mail this out. It'd be a physical mailer. They'd receive it in their mailbox, whatever that was, work or home, and then they asked. They had to take a pen, fill out a BRC, a business reply card, put a stamp on it, or sometimes just drop it in the mail. We would have to get that. Somebody would have to then manually enter in the database. That'd be shifted. So that first transition of the web, where we started actually printing, you know, printing the web address on the mailer, so people could just go onto their computer, enter the information. Just imagine how much time that saved that manual entry or having to make a trip to the post office so that friction went from you know what maybe took 30 minutes to an hour to like three minutes. Right? So then you have the mid-2000s, 2005 or so, when web platforms really became much more products and services, and that's where we saw things like at Huge, for instance, we saw the inclusion of the booking engine on the homepage of JetBlue, which was revolutionary at the time. Right? People were traveling, still using travel agents to plan travel, or they go to American Airlines or another website, and they'd have to hunt and find and look for itineraries and routes.

Anne T. Griffin: It was innovative. It was like you know, and it was something that really changed you know conversion for so many of the travel companies.

Chris Stempky: It did, right? And even beyond the travel companies, right? It sort of sets these expectations. So a consumer suddenly they can go to a website and book their own travel. Now they want to be able to do that same thing with every category of product that they have, right? Another thing we saw at Huge was HBO Go, right? Where we thought about revolutionizing how people get their content. Before, that was the Netflix days where you got the red envelope. HBO Go was like, I want this content on my phone. I want it on my tablet. I want it on my computer screen. So now we're living in a time where any time you want a piece of content, it's almost instantaneous. You can find it anywhere. So that's what we mean by kind of getting rid of friction. It's going from what used to take multiple steps, multiple you know a long period of time to now instantaneous, and so I think what we've seen now over the last almost 15 years, from you know when the iPhone was released, when we can do everything in our pocket, to now where we're talking about the agentic era, you know we're taking out all those steps in between. I know I want a product now. I have to go search for that product. I have to visit multiple websites for that product. I have to like re-review sites for that product. To now, I just ask the agent the question: Like, I want this. Help me find it and buy it.

Anne T. Griffin: Yeah. It sounds amazing, right?

Tom Rudczynski: Like the ability to just type a few words and get a travel itinerary planned. That sounds pretty good, and this is what Muse is promising, right? To go book the travel, send you your tickets, you know, you're cruising, right? There, all of the work that went into an interaction that we've made more convenient for the individual to do on the web, agents are now making convenient for the individual to do without having to do anything else. They have the thought, and from thought to action, it moves really quickly. That's the ideal, right? The thing that we're also looking at is these agents, ChatGPT, Claude. They're unreliable as well. So even though they might make it easier or more fluid, they don't always nail what the user wants, what the user would expect, or even what, for instance, like what a brand would expect the LLM to show about the brand, right? So the danger here is that we're moving into a friction-free era, but that brands were able to kind of form relationships with their customers and you know provide offers like loyalty and convenience and an elevated level of service. So the risk is that ChatGPT can bypass some of the major avenues that brands have already used to establish their you know where they are in the marketplace and kind of what promise they can make to consumers. But the issue is that ChatGPT might get you to an answer faster, but ultimately, if that's not the right answer for you, you'll be burned. And we saw that with a couple of the initial agent rollouts, where you know they tried shopping on ChatGPT and they roll that back almost immediately. So the concern is like, should we cede everything over to these agents and kind of let them decide? Their decisions aren't always going to be accurate, right? They need guardrails and they need help to get to like the same place we're at right now with the web, within the agentic kind of ecosystem.

Chris Stempky: Right? And I think it's actually interesting. I saw a stat from McKinsey not too long ago where it says I think 50% of people that they surveyed start their search through AI-powered means. So that could be ChatGPT, it could be Claude, it could be Gemini, right? And we've all been using Google. You search for something, and the first result is basically an AI-generated response. And your point, it's not always right, and it's not always what the brand wants to be out there, and I think that's another risk that we want to overcome by having these brand agent discussions.

Tom Rudczynski: Because it's not even just what the brand wants. It may not be what you want, right? Like we were talking before the show, right? And we were talking about what are some interesting agentic examples, right? So like choosing a new phone plan has always been a headache for me, and we were talking about like what that would look like on Verizon, but I think you had a really interesting insight. What does that look like on Google right now? And so if the agent doesn't know what your current preferences are and how you're, you know, navigating life, it's going to give you a suggestion that for you might be completely irrelevant, even though that might be you know what the agent says is the best option. And I think from a brand point of view, you definitely want the agent to know this person already has an account with us.

Anne T. Griffin: Yeah, and going back to your guys's Muse example, one of the interesting things is we all know Meta is the king of ads. They they're the best at it, and now we have an agent that's going to be doing brand discovery that's tied to one of the best ad platforms. But like when you throw that in the mix, it's this new way of thinking about well, like how much control do I have in that relationship? This is not Meta, but I talked to someone who works at a major food delivery company, and they said they tried to use some of the original kind of ChatGPT apps so that they could, you know, have a way for people to order via ChatGPT if they wanted to, and they said one of the frustrating things was even when they had their own app through ChatGPT, they felt like they were ceding some level of control, and so now you basically are tying this world of, yeah, Muse is gonna, you know, you're gonna be able to interact with different brands, or like your brand, right? People are going to be able to interact with your brand through it, but it also is incentivized to recommend people other brands because it's tied to ad revenue.

Chris Stempky: Correct. And I guess just like search is now, and even Meta is now, right? Which brands are going to get priority in that model, right?

Anne T. Griffin: Yeah.

Chris Stempky: You know, is Muse going to recommend products based on you or based on who's paying Meta more?

Anne T. Griffin: Yeah.

Tom Rudczynski: I mean, it's the Wild West right now. We don't know. I mean we're talking specifically about Muse, but Muse is only one of a set of agents, including Dot from ChatGPT. I don't think Claude has one yet, but you know Claude is conquering the world in other ways, right? They're connecting agents to agents on phones and websites and your local computer as you use tools like Claude Code or Claude Cowork, and that's bringing the world a little closer together. But the issue is like there's no way for a brand to right now represent itself to each of those agents as far as like, hey, this is already one of my customers. This is what they have to make the agent more helpful. And the vacuum from that is how the LLMs work, right? They're going to fill that with information, whether it's the correct information for the user or the sponsored information. It's going to be biased in one way or another. And the challenge I think for brands right now is how do you address each of these different agents who all do things slightly differently?

Anne T. Griffin: Yeah, so I would really love to ground this in an example. And Chris, I didn't know you did wakeboarding or wakesurfing, but we were talking about this the other day, and you were talking about kind of all the different like kind of levels of like, hey, you have to know this if you're buying something like a wakeboard, and so I would love to talk through how something like even just finding a wakeboard, trying to buy one of those, how would that work even with an agent? And then I'll let you talk about it. But I know Tom has something that allows people to actually get more of a visual as well on what this would be like. But we'll start with Chris.

Tom Rudczynski: Yeah, give us the anecdote, Chris, because then I'll show how the agent actually could activate something like what you're looking for.

Chris Stempky: And shameless plug, this is me riding a wakesurf board, so I had to get that photo in here somehow. It's an up-and-coming sport, so people are still learning about it. I'm still learning about it. I didn't know a lot about it when my friends dragged me out into the boat. And so when it's time to buy your first wakesurf board, there's a lot of questions that I have going into this, and that first one I buy is very important, because this is where you're learning. This is how you're gonna get good at the sport. This is gonna be how you impress your friends, and if you have the wrong board, you know all of that. All of that is lost. And so my search was like, I'm looking for a wakesurf board, and all of a sudden I've been taken to like six different product websites. I'm going through and I'm looking at all these things, and I'm just completely lost. I start watching videos of wakesurfers and reviews, and it took me days before I could find something for myself. But what if I'm able to walk into a store with a very knowledgeable salesperson and then say, "Hi, I'm Chris. I'm new at wakesurfing. I'm looking for something that I can learn on, and I want to spend about $500. Can you help me out?" A knowledgeable salesperson is not going to say, "Yep, I got the board for you." They're going to ask other questions that are meaningful in that purchase process, and they're going to customize it to me. Like, well, how tall are you? How much do you weigh? What do you want to do on this wakesurf board? Are you looking for just you know riding at a nice calm speed? Are you looking to do crazy tricks, which I can't do yet? Is this for you, or to share with your family? All these questions that go into a purchase. Right now, a website, product website, doesn't ask you, so you're left trying to figure all this out on your own, or reading reviews, or leaving, and all of that breaks again. All that coming back to this idea of friction. But what if I can just come to a brand website, I can type in, I'm looking for a wakesurf board, and they're like, great, tell us a little bit more. And I don't want to get too far into the story because I think this is where Tom is expecting to show how this can work.

Tom Rudczynski: I want to add and build on that because right before I show the demo, I think the concept of having a salesperson, someone who knows exactly what the product category is, what the consumer wants, it's something that actually has been missing on the web. The you know dynamic content being able to like respond in real time just didn't exist, and there's a lot of categories of products like wakesurfing, or even photography or anywhere where you're like looking to compare and get inside information. Those insights are missing from the brand website, or they're very much buried in different pieces of content that are not present for you. So you know what we're going to show you with this website is like how things look today, and we'll conclude with what will happen immediately after this because we're at such a period of velocity that even the things I'm showing you today are the tip of the iceberg compared to where agents and full agentic integration could go. So my team put together a wakeboard shop. This is a real product catalog. It is named Anne's Board Shop for privacy purposes. We're going to search for a wakesurf board. My product handles semantic search, and it can get a variety of results for a variety of different search terms, but in this case, we're going to start with where Chris is starting, right? So I'm searching for a wakesurf board. When we put an agent on the page, the agent can understand what's happening on your web experience, but then it can begin a conversation with you to help you identify these things as you narrow down the information that you would bring. You might not know the questions to ask, so the advisor works like a salesperson to ask you the questions that'll help you formulate your purchase decision. So, you know, here we have a set of examples, and if you guys are familiar with Amazon, Amazon is doing something very similar to this today. What we're saying is that any brand can do this, right? You just need to have the right kind of tooling connect with the right kind of data, and you'll quickly be able to put together an experience for your consumers that actually will marry those two surfaces, right? The self-service speed and the you know informed advisor. So, Chris, it's asking you what kind of style you want, right?

Chris Stempky: Yeah. So I like the skim style.

Tom Rudczynski: Okay. Built for spins and tricks. I can't wait to see this.

Chris Stempky: I'm trying. I'm trying to learn.

Tom Rudczynski: Okay. So we're refining the search results here, and now we've got a few more questions for you. What's your riding weight? The board size, the one you're actually going to pick, depends on how much you weigh. So we've got a couple of examples that are skim style, but we can fine tune these here.

Chris Stempky: Yep, I weigh about 200 pounds.

Tom Rudczynski: Okay. So as the advisor thinks, we're reviewing and you know recreating the grid. So this is one of the cool things about using agentic technology is not only we're giving you a chat experience, we're modifying on the fly what the user experience is. So what was typically standard in e-commerce now can be very dynamic and curated to the conversation, not even to the kind of use case. And you know we're exploring different paradigms between like, are we happy searching in chat? Or would we prefer to see a grid with a large variety of products, what we're thinking about is like you know when these things start to converge, when the experiences start to become one and the same. How are we connecting between a chat and a search and a you know self-driven experience? So here we have some Skim style boards. Looks like a variety of price ranges. I think you mentioned you wanted under 500, right?

Chris Stempky: I did.

Tom Rudczynski: Yeah. So these are pricey.

Chris Stempky: They are pricey. That's why you want to get it right.

Anne T. Griffin: And I assume that this type of thing would be something where if someone's using Muse, you could even have a Muse agent connect to one of these brand agents, so it can help people walk through these steps if, like, Muse is not familiar with all the intricacies of buying a wakeboard.

Chris Stempky: I think you put $400, and I put 400.

Tom Rudczynski: Should we up it?

Chris Stempky: Yeah, raise my budget a bit.

Tom Rudczynski: All right, let's see what comes back. Okay, we've got one or two key choices here. Looks like the Connelly and the Hyperlite are your preference. So, which one of these would you choose? Let's choose the Connelly just to keep the demo flowing. We're going to choose the size as recommended. So 4'7", 4'8".

Chris Stempky: Yep.

Tom Rudczynski: And add it to cart. So now with just a couple clicks and you know a conversation, I have a product in cart that's very much what I'd like to get. So from here, you know, we've now had the interaction at the level of the salesperson, right? They're bringing information to you. They're helping educate you about how you might use these things, and they're making it easy to add to cart and check out.

Chris Stempky: Yep. And I know we're going to get to this later, but I think it's important to point out that all those questions I just answered about my weight, my style, you know, my level, my preferences, all that is information that the brand could not collect before on me, and now it is collecting. You can collect that, and again, that just really goes back to the level of personalization, I know we're going to talk about personalization a little bit more later, but I just wanted to call out here because not only was it my product preference, but it was all the nuance behind the selection of that product that was now captured by the brand.

Tom Rudczynski: I was thinking about this on the way over, right? We're beginning to form profiles, not just of like what consumers buy, but what they ask about and what kind of interests they have before they buy, right? So that data can then inform how you're structuring your product catalog, how you're communicating with LLMs. Like, hey, our consumers are interested in you know these kinds of use cases. So when they ask about these kinds of use cases, here we are. We have the right information for you.

Anne T. Griffin: Yeah.

Tom Rudczynski: And brands can start to build these portfolios of their customers, whether it's on an anonymous basis, just an aggregate, like what are people talking about. But I think more importantly, on a personal basis, right? So after you convert and after you check out, we know some things about you that next time you come back, the agent will then pick up, right? Like, hey, congrats on the Connelly. Can I get you any accessories to contribute to that? So you might be on a different part of your purchase journey, but brands are going to be able to like much more accurately target what consumers are looking for, potentially looking for, and not in the kind of rote way that we see from Amazon, right? Like if you add a shower curtain to the cart and you check out with the shower curtain, you know, I personally will get shower curtains recommended. Hey, is it time for a new shower curtain, right? That kind of like monotonous, over and over, kind of dumb personalization will get replaced by these dynamic experiences that I'm trying to demonstrate here.

Anne T. Griffin: One question I have for you: I noticed, you know, we talked a lot about removing friction, and one of the things is I noticed you have the add button or add to cart directly in here. What needs to actually be hooked in to make that work well?

Tom Rudczynski: In this situation, this is a Shopify plugin that we're developing, but for the most part, brands need to start thinking about opening up their ecosystems to allow agents to do things dynamically. You see some of the like commerce protocol that Shopify is pushing. Shopify is all in on this, by the way. They're making it as easy as possible for agents to drive checkout. But being able to expose your tooling via API, being able to allow transactions to occur, connecting to your data sources is another key element because connecting to your data sources is what's going to give the information to the agents to be accurate and clear in your own experience. Right? Elsewhere, they're kind of guessing a lot of the time. They're using world knowledge, but when you own the experience, you can also more curate what the advisor is going to be doing.

Anne T. Griffin: Yeah, and my follow up question here is actually, if the agent is doing the shopping, why does brand still matter?

Tom Rudczynski: This is a really good question.

Chris Stempky: So let me start, Tom, and you can jump in if you want. But what we sometimes forget is the brand is more than just the product. It's the experience around that product. And I was thinking about this, and I have a really good example where I don't know how many years ago this was Bonobos, the clothing brand that started with men's clothing. Like everyone, I bought a pair of pants for work, and you know it was fine. It was just another pair of pants I had in my closet, that I could throw on. It wasn't until a pair of those pants ripped, and I contacted customer service, and it was probably the best experience I had ever had with the brand, just the tone of the email, the way that the consumer rep kind of chatted back to me, told me what to do with them, and you know they had some, I forget what it was, kind of like a kind of witty response about what I could do with a pair of pants. So they sent me a new pair of pants, and that moment there, I was like, I love this brand, and so we have to remember that the service that you're offering, the value you're offering your customers is more than just the product. It's everything around that. So when we're thinking about agents, and the example here about the wakesurf board right, it's like when I just went here and it was just so easy. I'm going to get that board. It's going to be you know specifically for me. It was customized to my goals, my why, and I'm going to get that, and I'm going to love it. And so suddenly, that relationship is more than just the board. It's like, where did I get that? And it's like, oh my God, I got it from here, and it was so great. They knew exactly what I wanted, and they made a recommendation based on that. And so we have to think about the brand as the experience, the larger experience now. And I think those agents let your brand have those personal experiences on a one-to-one basis.

Tom Rudczynski: The level of service definitely makes your brand shine in this example. But I think more broadly, there's this concern. I think that if you let agents do what agents want to do, they're going to drive to lowest cost. Or you know, in the Muse case, wherever the revenue is coming from, that's where you're going to direct it, but that's not necessarily what consumers want, right? You don't always want to buy the cheapest thing. Consumers develop identity based on their relationships to brands. Brands can broadcast elements of your identity and how you want to be viewed in the world and associated. And I think LLMs, for all their wisdom and knowledge, they don't understand how people want to be identified, and kind of the elements that people curate from brands to make that their own, right? And so, by bringing these kind of like friction-free experiences to your brand website, you get to own that relationship further, right? You're not ceding it or abandoning it to the ultra technology, but really, like you're taking that technology and bringing it to where you want your users to be, and your users want to be, right? This is why, when we talk about brand websites going away, I'm skeptical. I can imagine a world where things are navigated without the web being the primary presence, but at least for brands establishing great customer service, great customer experiences, and great, you know, great experiences that represent the brand. You know, you can have an agent talk in a specific way to a specific kind of consumer, and in this era, that's going to last and be stickier a lot longer. In a similar way to Chris, like finding those great salespeople.

Chris Stempky: Yep.

Anne T. Griffin: So, one question I have here is also, you know, what exactly is a brand agent? Because we've been talking about it, but like, if we had to give it a definition and like, what makes this different than just like a typical e-commerce website chatbot?

Tom Rudczynski: Yeah, it's not a chatbot, and I think that's the important thing. And this is kind of what's changing now that we have these personal assistants, right? The brand now has a relationship with you via the agent that it didn't have before. Taking this information from this chat and extending it to future interactions is kind of a starting point. Extending this to all relationships that you have, say you can log in to your version of Claude and have this experience, so that your version of Claude remembers what's happening here as well, means that that relationship with the brand and the expression of interest then carries in through the rest of your kind of custom experience using these agents to drive a lot of what you're doing, and if that's the case, then those experiences become much more robust and personal, and that's the true friction removal, right? You're not struggling to find the things that you like. You're leveraging what you've already expressed to grow, you know, to become, you know more of the wakesurfer person that you will then go recommend surfboards to, or you'll even recommend the brand experiences to because they represented you so well.

Anne T. Griffin: Yeah. So one of the things we've been talking about a lot is the personalization aspect as well. And Chris, you brought up this really great example in our conversation the other day about there was this like when you were working at this agency called AKQA, walking down the street and like there are these shops you used to pop in pretty regularly and they just like knew you. So can you talk about like how we're kind of translating some of the personalization and delight that you got from these in-person experiences that were kind of a little bit absent from the website era, because it just has a different set of capabilities in agents and how that's going to translate into agents.

Chris Stempky: Yeah, thanks for bringing up my embarrassing shopping stories. I thought it was very endearing personally. So I mean, I lived and worked in SoHo in Manhattan, and you know, and I have to walk by, and you walk through all these stores, and so inevitably you pop into a store, and you're like, oh, I'm gonna check out this or that, and you start to have your favorites. And what I was telling Anne the other day is that one of them was Ben Sherman. They were a British brand, and I'd walk in, and pretty soon, you know, all the salespeople get to know you. So when you walk in, they say, "Hi, Chris, how are you?" And then it gets to a point where like, "Hey, Chris, we got some new shirts, and we think you're gonna like these. We put some aside for you because we knew you'd be in." Or they even went to the point where I knew them so well, they would say, "Hey, we're gonna set these aside until Thursday because we're having a big sale on Thursday, so you don't need to buy at full price." So it was like the relationship was so personal that they knew my name, they knew my style, and they even were like kind of looking out for my best interest as well as their brand. And I feel like you know when shopping moved kind of in the digital space, you get the personalization has gone from that really hyper focused individual personalization to what Tom mentioned, where he buys a shower curtain once. If you buy, and you've probably all been through this. You buy a shower curtain once on Amazon, and suddenly you're just recommended every shower curtain under the sun for the next however many months. It's not real personalization, and it's not understanding the person who's doing the buying. And that's where I think like we're moving to this idea, this opportunity where brands can have these one-on-one hyper personalized relationships because they know you, they know your motivations, they know the why, and so if they know all that about you, not only just what you bought, that can lead to those conversations being much more beneficial for both parties down the road.

Tom Rudczynski: They can build on the experiences that you're having, in addition to like, they can look at what you've purchased and then guess what the next step after that is, right? And we didn't have that level of intelligence before to map out these journeys of like product engagement over time. And I think the really cool thing about LLMs is that it allows us to do that dynamically with you know just a core set of rules, but that personalization level actually relates to human experience.

Anne T. Griffin: Yeah. So I mean, part of that, like, that's partly how the relationship forms, right? It feels very personal. But how does this differ from... there was a data science project I did years and years ago that was part of this website called Kaggle, where you could do fun little data science projects with Airbnb data, where it was like you take Airbnb data, whatever like they could get about like your age, gender, everything else, and they were like, "Help, help me figure out where the first place someone is going to most likely book a trip, and show them homes in that place." So, like, what is different between that? Like, how does that evolve? How do agents evolve that relationship from web into the agentic era?

Tom Rudczynski: So, I think for me that machine learning data model that you just described is about predicting based on most likely to happen across a broad population. What the agents are letting us do is discuss based on your individual needs, and that's where like it goes from like we guess that most people do this after that action to let's find out, let's engage and actually create the right option next for this specific customer.

Anne T. Griffin: Awesome, and so I think one other question I have is we talked about this Amazon example of like I bought a shower curtain. It shows me a ton of shower curtains, right? What is that ideal post-purchase experience with an agent? Like the agent helped you buy a shower curtain. Now what?

Tom Rudczynski: Yeah, I mean, I think we have an example for this. But you know, as we were talking about, like the data that you're giving to the agent then can inform the rest of your customer lifecycle journey, right? So we have all these tools for active engagement with customers, sending them texts, sending them emails, and the best we can do is say, like, you know, you've done this category, so you might be, you know, interested in the category again. But an agent can actually create an ultra personalized email based on your previous interactions, right, and follow through and continue the conversation the way a good, but maybe persistent, salesperson would. Right? Like, "Hey, I learned this about you." Or Chris's example with the pants: "These pants just came in, and I know you bought three pairs last time, so let me know if you're interested and we can show a few products to you." So it's that level of: we're all the way down the funnel to the last step, and we're able to present that as you know when it's important to that customer. And that is to me, it's a completely different thing than getting a coupon for women's clothing because you shopped for women's clothing before.

Chris Stempky: Right. I I'll add to that. I think it's sometimes it's not just selling you more product either. Right. In my case, right. It was that relationship. I built a relationship, and I was saying this, Anne, the other day. I built a relationship with the salespeople because I felt like they knew me as a person, and this is kind of weird, I guess, looking back at it. But I, you know, we became Facebook friends with salespeople at these stores I used to go into. But if we think about these agents, and if the follow up isn't just I want to sell you more, but I want to see how you're doing. Like, if I take the wakesurf example, I want to see did the board meet your needs? Are you able to do tricks? Hey, by the way, here's instructions on how to do a couple tricks. Suddenly, that relationship feels much more meaningful and deep, and then that goes all the way back to the brand expression that we talked about the importance of brand. In that, I'm not getting that with you know just my current shopping experience or even the LLM, it's like that follow up and the opportunity to kind of deepen the relationship just because you know as a brand I care about your experience with my product.

Anne T. Griffin: Yeah. So we are getting close to Q&A time. There's at least one other question I want to ask, but I think we'll make it two. So I'm going to ask this next question about like what is the risk of a brand becoming invisible, and I'll just share like you know this is not necessarily you're becoming invisible, but you're competing also with what is drawing the attention of the agent, right?

Tom Rudczynski: Yeah, I love this example. This came through in August. An influencer decided to create a fake deodorant brand using Claude Code, and you know posted it one day, and it began to get traction in LLMs to the degree to which it was ranking like second or third in alternative deodorant brands, it was a completely made up brand, couldn't even purchase, but they had managed to get ChatGPT to suggest it as a regular brand. The reason I like this example is because brands have spent a ton of time invested in brand equity for things like SEO, you know, paid social, social. But in this situation, the LLM, the ranking mechanisms and the way that it found and informed the agent about what to display isn't really that clear. So the issue for brands is if you're not moving in this space, someone else is. That's one. And two, you're never sure what the agent, or ChatGPT to be specific, is saying about you. Right? They're liable to make things up if they don't have a canonical source of truth about your brand to represent, like you know, the specifics about it, it's liable for consumers to get confused, get misinformation from the brand, and there's no way to enforce the LLM is telling the truth every time it recommends things. The only way we know right now how to do it is to publish well documented data about your products and services so that it has something to pull from to combat things like this, where there's you know less historical presence and to take a step towards regaining that traction in these spaces.

Chris Stempky: Yeah, and I think going back to your other question, like when brands become invisible, I mean, then I think we've seen this in every other digital transformation pivot that we talked about. If brands don't evolve, they become invisible and by nature obsolete, or they get beat out by competitors who are staying more relevant through technology. We've seen it in brick and mortar stores. We've seen it in digital stores. We've seen this time and time again. Every time there's a pivot, you're gonna see new players kind of win at that. You're gonna see players who've been around sometimes decades or longer sort of go away, and I think that's the risk there as well.

Tom Rudczynski: And I want to reiterate, sometimes it seems like we're talking about two different things, right? Like we're talking about what the agent experience is like in the ChatGPT interface and what the agent experience is like on your website, but we're hitting a moment where those things will begin to converge. You know, just in the last week or so, I mean, first of all, Muse being the number one app on the App Store, I think is like opening everyone's eyes to hey, we need to have a solution for when this thing starts to talk to us.

Anne T. Griffin: It got beat out by Dave's Hot Chicken, actually, yesterday.

Tom Rudczynski: Congrats to Dave. That's awesome.

Anne T. Griffin: Someone asked what models they use.

Tom Rudczynski: Oh, where was I?

Anne T. Griffin: You were talking about Muse and those sorts of things, and convergence.

Tom Rudczynski: Yeah, they're convergent, right? So I just wanted to share this anecdote, because Claude Code is something I use every day, and over the last two months, they've introduced this idea of remote control, where now everything that you do on your laptop, on the website, anywhere else that you can log in with Claude Code understands the context of your previous interactions. And so, following on the heels of Muse was Sierra, from one of the guys from Google Labs [Clay Bavor], which just came out with something called the personal... What is it called? Something personal

Anne T. Griffin: Agent protocol.

Tom Rudczynski: Agent protocol. Yeah. So that's going to allow users to kind of control more what information gets passed to which agentic channel. And in the case where the agent is making decisions on your behalf, that's going to be really critical. But because that exists, I think it's opening the doors to creating these fluid experiences where you can move between ChatGPT and a website and preserve that context for everything you do. You know, through these agents going forward, we're looking at the future right now.

Anne T. Griffin: Yeah, absolutely. I know we have so many more things we can talk about, but we may have to share that for like a part two. I want to make sure we do get time for the Q&A questions. There's a really great question from [an attendee] in the chat, who works at the Associated Press, and she says, "I'm curious whether you think agents can be useful on a news website, and what that might actually look like for readers." Which I'll start off with. I think one of the trickiest things about this is that so much of media had to start like traditional news had to start competing with like the clickbait and everyone who's just trying to drive a bunch of numbers, which obviously like flipped a lot of the traditional publishing for that type of media on its head. And so there's a lot of things when you think about: Are the agents going to funnel people to the clickbait, or are they gonna funnel people to certain things? And how do people set those examples? And how, you know, does the agent know what they are interested in? Is it because they've connected their Financial Times to some sort of special Financial Times agent or some other thing. So I think this is going to be one of the tricky ones to make sure it's not just a topic. It's also like, what are the publications that this person tends to actually take action on? Tom, I can see you're excited.

Tom Rudczynski: Oh, I was so excited because on LinkedIn this morning, I just saw one of our former colleagues is now working at Bloomberg, and they just put together for Bloomberg Media an agentic search and chat, right? So when we talk about news organizations, the use case is okay. I read an article, right? How do I drill deeper into this? If the purpose of news organizations is to disseminate information and make it understandable for a broader audience, an agent can help users not just read an article or find an article, but research the whys behind it and kind of form a much broader perspective that they can then, you know, leave with that great relationship with the news brand itself, saying, "Hey, this is where you can go to not just like hear a story, but like dive deep into the story and they be guided to similar information that might be relevant to you." Without the kind of like the way we're doing it today, with like you know, this article was correctly tagged or not correctly tagged. Those are all like manual interventions that brands have to do. One of the nice things about agentic is we can load these things up with the right amount of context, so that crossing between articles, experiences, and eras in history is a lot more streamlined from an operational perspective.

Chris Stempky: But the other thing that I think the opportunity is obviously with Associated Press and any news site, the amount of information and data that you can collect on any individual person. If you think about how you use that data, right? Not only can you use it for better recommendations on stories, you can see what your people are coming for. It helps you inform future editorial decisions. It can also help you like build deeper, stronger relationships. So people keep coming back to AP for their news source, but versus just going wherever clickbait takes them. And then we can think about it from a revenue perspective, right? If I know Tom and I know Tom loves sports and he loves you know world news and NASCAR or whatever. Like I'm building all that knowledge base, not dissimilar to what Meta does right now. You have the opportunity to build those profiles, so that way when you're using advertising, and I believe there's advertising on the AP website or any publisher that has advertising, those ads are going to be much more targeted and valuable to Tom, and in turn much more valuable to AP to sell that ad revenue, and as we think about agentic advertising, where that's going to go, and we'll save that for another topic. But thinking about again, the personalization of the product recommendations and ads that I can give my readers, it's much more. And again, it's building that deeper, more meaningful relationship with your readers through the editorial, the ads, the engagements they have on your website.

Tom Rudczynski: But the data moat itself becomes like a win, like you were saying, right? You now have much richer profiles of what your consumers are interested in, how they're interacting, and that obviously is going to lead from an ad targeting perspective to some much better results. I would imagine.

Anne T. Griffin: Yeah, I think if I were a media company, I would be focused on like what, especially if like I don't know if AP has a login, but if I think about other brands like New York Times, Wall Street Journal, Financial Times, I would probably be focused on like because a lot of you have to log in, right? I'd be focused on like what do I need to build so that, even if I use my Muse, or there's some sort of special brand agent, like "this is the Wall Street Journal agent," I can actually do a lot more personalization to know okay yeah this might be the front page but there might be certain things where it's like other things that it wants to share with me, like I can give them much more personalized news, both in the newsletters that I send to that person, but also maybe there's like a more agentic driven page, those sort of things.

Chris Stempky: So a page just built for Anne, so you know what stories Anne likes when you open up apnews.com. All those stories have been curated for you as a reader.

Anne T. Griffin: Yeah, and you said loyalty has become central to AP in the last year, and I do think about that as like a big thing of like you know I still read and I understand like right now people keep talking about this world of like oh personalized UI we're not quite there yet, only because LLMs still need a lot of driving behind. Like, well, what? How does that look? And maybe news. It's not news. It's maybe more of a standard layout. But I do think of there being a lot of kind of dynamic things where it can add to a whole nother level of personalization. Where it's just like if we already know that like Tom or Chris wants these specific things, like how do we make sure that they get to them? Because I know, like I recently subscribed to the Financial Times. The thing is they ask me for my interest, and I'm like, I gave them my interest, but then it's like after that, like the personalization kind of stops. I have the little page for like cybersecurity and AI and like all those other things, but there's no page that's like for Anne, which is actually really interesting because people even though it was controversial at first, X is obviously not like a news site, but they are in the attention economy, which you know like the AP is also in the attention economy, and people hated that For You column. And then once they kind of fixed the algorithm there, people stopped complaining about it. And there's a lot of things where people interact with people they otherwise would have never met or maybe found. And there's not really that concept of it in what you talk about, like loyalty and like personalization in kind of media right now on like their own domain. Anything else we should talk about? Because I have some other follow-up questions, or if anyone else has questions in the chat, I have some seeded questions I want to ask Tom and Chris. But otherwise, Tom and Chris, is there anything we didn't get to cover today that you want to make sure we talk about?

Chris Stempky: The only thing I would say is, I think this conversation has been fruitful, and anyone listening is probably thinking, okay, where do I start, and do I have to? Does everything that I've done with my business for the last 30 years matter still, or is this essentially just going to replace everything I've done? And I think the best answer is that everything you've done over the last 30 years, or you know, I say 30 because that's how long I've been doing this, but that's the foundation of building a good brand agent, right? All those systems you put in place, all that learning you've done to help understand behaviors of folks on your website, all the infrastructure you built, the data infrastructure, the relationship management infrastructure, the e-commerce infrastructure, that's still the foundation for this. Again, we got to remember, and I always say this because you know AI can be very complicated. There's so many nuances, and you start to like go into this almost like people start to tune out because like oh it's too complicated. AI is just software, and so at the end of the day, a website is software. AI is the software, so it's just now the evolution of that software. So the foundation and everything that the software is built on, that still matters, and everything. The other thing is part of that. Everything that you've sort of built to build your brand narrative, your brand DNA, that has to go in there now. So it's now just a combination of connecting those systems together. So I think this, you know, to start, you know, I think you just have to ask yourself, like, hey, is my infrastructure ready? Can I get it ready? And then maybe pick a pilot or pick an area to tackle first. Don't boil the ocean, so to speak.

Tom Rudczynski: Yeah, exactly. We don't have to solve the entire problems of the business and all your users, but I think where agentic flows really help is like to sort out confusion that a consumer might bring. So, what's the stickiest problem in your business? If you're looking for a place to start, figure out how you can document the information about those experiences and bring that into an agentic experience, because at the very least, you're going to be hitting consumers in a way that they haven't had an opportunity to work with your brand before. It may not work perfectly the first time. I'll say that. Like as I mentioned before, this is the Wild West. So as the situation is developing, there's going to be more alternative and interesting opportunities as they emerge. But I'm here to represent also like the reason to push you towards that is if you're not taking those actions, someone else is, right? There's going to be a land grab in terms of friction-free experiences. So this is similar to you know those pivot moments that Chris talked about where the gaps get filled in by new entrants because they're willing to take those risks. So it's both a great customer experience, but it's also like a protective brand experience to start engaging in these kinds of ideas.

Anne T. Griffin: Awesome. Well,

Chris Stempky: You said you had a question, Anne.

Anne T. Griffin: Yeah, I'll make it quick. So one of the things that we talked about this a little bit, but like, you know, how can adding other data sources to these brand agents, like getting information from like whether it's connecting to email to learn your preferences or other places, like how can that help learn customer preferences?

Tom Rudczynski: Yeah, so you're not just like taking static data sources, right? You're going to be building libraries about your consumers to say like these are the tendencies among our groups. These are the specific preferences, and because an LLM can make inferences and draw connections between data points that a human would otherwise have to do. You can do this at scale, right? You can do this very rapidly for a large number of consumers using only an LLM to kind of drive that more personalized messaging. You have to put rails around it. You have to evaluate what it's doing. But once you get started, that creates that kind of more personal touch. And I can only say, by the success of Muse, that people are interested in that.

Anne T. Griffin: Yeah, absolutely. I think that's the key part. This is, I think, the big opportunity: how do we make it so that these are things that people are interested in, people want? But in a world where it's again, you have this proxy doing it. So I want to just say a big thank you to everyone who joined us today. I want to say thank you to Tom and Chris. This has been so insightful, and I loved [the attendee's] question, and I just really love kind of I think the interesting space that brand agents are taking us. Thank you so much for your time.

Tom Rudczynski: Thank you very much, Anne. I'll tell you what, I'm excited to keep this discussion going. We can meet again and talk about what shopping looks like next year.

Anne T. Griffin: Yeah, because it could be completely different. Even just like infrastructure, what do people need to do to prepare?

Tom Rudczynski: Absolutely.

Anne T. Griffin: Yeah, amazing. Well, thank you, everyone. Have a great day.

Chris Stempky: Thanks, Anne. Bye.