Ep 31. Autonomous Marketing Is Coming - But Brand Still Wins, with Jamie Domenici, Klaviyo

Speaker A: Hello and welcome to another episode of the AI Edge podcast. We are here to help folks in the marketing and advertising industry keep up with all of the latest developments in AI and what it means for our space. We are here again this week with another awesome guest. We have Jamie from Klaviyo with us. Jamie, tell us a little bit about who you are and what you do.

Speaker B: Oh, uh, thanks. Well, thanks for having me. My name is Jamie Domenici. Hard last name, so I'll say it. And, uh, I'm the CMO over at Klaviyo.

Speaker A: Awesome. Um, and I need to know, I need the, like, 10 second Klaviyo elevator pitch for our listeners and for me, listeners, of course.

Speaker B: So at Klaviyo, we're a B2C CRM. Um, so we really help B2C companies connect with their customers across every channel, across every medium, and help them grow their business.

Speaker A: Awesome. Uh, well, we're thrilled to have you, Jamie. Thanks for joining us. We also have another special guest. Actually, he's not so special. His name's Miles. Hey, Miles, what's up?

Speaker C: If I not special or is my guest status not special?

Speaker A: No, it's not special. It's not special.

Speaker C: Uh, this is, uh, Jamie. Thank you for joining us. And I love the, uh, we're a B2C CRM. That's like, that's great. Single sentence. We get it.

Speaker B: Yep. Directed to the point.

Speaker A: Also combining two acronyms that everyone generally knows. You know, it's not like it's like two weirdo acronyms that no one's ever heard of.

Speaker B: Like, they go together. B2C and C. B to C, CRM.

Speaker A: Okay, uh, so let's jump into top news of the week. So, um, I want to kick this week's episode off with. You know, there's been a lot of, like, buzz in the last week about, uh, first of all, I had no clue what was going on. Like, all of a sudden my, like, social feed started filling up with this word called Kimmy. And I was like, who's are we talking about? Like, Kimmy Gibbler from Full House. Like, what is Kimmy? I don't even know what Kimmy is. Um, and then I had to, like, go and look it up. And apparently there's a new Chinese AI model that has taken the world by storm, um, called Kimm K3. Ah. It's a new model by a startup Chinese company which is funded by Alibaba. Uh, and the company is called Moonshot. Um, and the big hoopla around Kimi is, uh, that it's basically at the same level of performance as some of the frontier models from Anthropic and chatgpt. Um, and it's half the cost, right? That's the big headline, right? It's way cheaper. And as we know, over the last few months, the whole AI conversation has shifted from like, ooh, these AI models can do all these things. From like, how much do these AI models cost? Right? How many tokens am I burning? And so this became big news. Uh, over the past week, there was a couple of other things that kind of popped up adjacent to this news. So one is, um, Google, uh, launched three new Gemini models. So, you know, in some shape or some order. Uh, one is called Gemini 3.6 Flash, another one's called 3.5 Flashlight, and another one is called 3.5 Flash Cyber. Ask me what the difference is between any of them, I couldn't tell you. Um, one is faster, one is better at coding, one is whatever. Uh, but at the end of the day, I found the news to be interesting because they are starting to release models that are more so like, specialists for certain things, as opposed to, like, oh, just use the latest model to do everything right. And I think we've been seeing that more and more over the last year where, you know, they'll, they'll say like, oh, just toggle the switch in your AI if you want to achieve X or if you want to achieve Y. But I think that's been a little bit more meant for power users. And now they're like, I think, trying to make that concept more mainstream. Um, and so anyways, there was a bunch of other News. Like, apparently OpenAI got hacked. Uh, or like they, they hacked themselves and they went into some software thing called Hugging Face to steal some secrets. This AI stuff's like, just off the rails. It's crazy. This is like, yeah, I can't keep up. But anyways, uh, just kind of thinking about, like, the tokenization of AI and like, how that has become kind of the, the commodity that everyone's thinking more about. Um, thinking about how these Chinese open source models are like, blowing up again. And this happened with Deep Seek a year ago. Like, any thoughts on this, initial reactions to this, Jamie?

Speaker B: I mean, I think it's interesting, I think that the market, we're seeing some new players enter and, uh, it's going to become a different competition, which I think is actually good. I think it's never good when you just have one or two players now. I think it's commoditizing A little bit, uh, and around the fact that it's in China. I think it's interesting. I, uh, think, you know, this is where your legal team has to come into play and, you know, your engineering and make sure you're taking the right steps to protect your data. I think that's the most important. And the marketer in me, though, is thinking, all right, well, if, if we use, uh, outsourced model, or if any one of our customers are using outsourced model, you need to protect your data, you know. And so I start going into a comms mode where I think, well, what happens if something is shared? I would say, like, the CMO has to be ready to, uh, ensure that you're communicating about how your data is used, where it's used, how you're getting information. But, um, I also think these models are kind of exciting because I think they are going to bring down some of the costs, uh, as we start looking about how you're fueling AI, which everyone is looking at, as you said.

Speaker A: Yeah, yeah. And I think like, uh, if you're a marketer and you're like, you know, deep into it, you're building agents, you're using different platforms, and now all of a sudden you're starting to look at your AI bills, I think now more than ever, it's like, okay, well, there probably needs to be some process and scrutiny all the time right around like, okay, well, what's the latest model? Am I using the right model for the right agents? Right. So it's not just about like the, the cheapest company. It's like, oh, now we have specialized models that are better for different things. Am I using the right models for the right kind of workflows?

Speaker B: Right, yeah, you actually need permissions for models. So I, I think everybody ran out and deployed every piece of technology that you could, which was great. And I was like, build every agen. But now it's, uh, not that we're pulling back. I don't want to change that behavior. But yeah, engineers, they probably need a stronger model than the marketer to build copy or, um, some of the creative. I think more and more companies are having to be a little bit more cost conscious and then a little bit more strategic about how and what they're deploying, which is good. But, um, I think some companies are kind of backpedaling at this moment.

Speaker C: I'm just waiting for a return to the 90s. When we get the CPU under somebody's desk that says do not unplug. And that's where all the AI models are running because like we couldn't afford to keep sending it out to the cloud. We just had to buy our own GPUs. Do not unplug, please. The whole company runs on this box, right?

Speaker B: I love this podcast. We talk about Kimmy Gibler cpu. Yes, Rob, it's great.

Speaker A: Yes. Hey Miles, do you have any initial reactions to the Chinese models and the specialization of these models?

Speaker C: Um, M. Mine is one of um, there's an interesting, it's almost kind of like ah, a maybe three way market positioning sort of battle going on right now. One is say ultra cheap overseas AI is one uh, where like that's the value prop is like this is cheap. Like yeah, it's overseas but it's cheap. The other one is sort of what everybody has gravitated towards to date, which has been the perhaps overpowered models that are like extremely expensive. You know, performance just keeps going up and up and up and you know, eventually we're going to reach AGI and blah, blah, blah, you know, these really, really high powered, uh, you know, largely domestic models, anthropic, you know, Google, et cetera, uh, OpenAI and then. But the third one is, you know, where Google is investing, you know, into these lightweight specialized models. I'm like, well Google has to, they have like billions of global users. Them and Meta have to find a way to make it cost effective. Like they have to, they literally, it uh, will not work for them. And so what's interesting is that there's a little bit of kind of like a rug pull going on from the hyper sophisticated models of like. Well, they spent years convincing the marketplace that what you needed was more power, more power, more power. Then the bills come due and everybody's like, we'll start you looking at alternatives. So I think we're gonna, we'll see some interesting kind of marketing. I don't want to call it backpedaling but like adjustments made by Anthropic OpenAI to be like no, no, no, it's cool. Like you can be cost effective. With our platform too. We also offer products that are cost effective and downplaying kind of the sheer power aspect of it as the value prop.

Speaker B: Oh, I hope nobody if we do that, that's. I mean if I was in that seat, I would not move away from, I would not move to market on price. If I was anthropic in particular, I feel like they have a power move. They should. Yeah, that's not what you're saying. They're gonna have to have alternative Options, some tiering and such.

Speaker C: But it's an interesting question. It's like, where do you. Where do you go? Where do they go? Yeah, Shiv, I think you kind of called it out, like, specialization. You know, like, do we really. Do you really need to, like, fly your kids to school in a helicopter every day? Or can you just take a Honda Civic? Like, if that's the. If that is the job that is required.

Speaker B: Yeah, yeah, sorry.

Speaker A: Uh, go ahead, Jamie.

Speaker B: No, no, no, you go ahead. Exciting. Very exciting.

Speaker A: No, I was just, uh. I was. I was. I. You know, we'll go pop culture for a second. LeBron James is going to the Philadelphia 76ers, not, uh, the Golden State Warriors. Not the Golden State Warriors. I am, too, Jamie. I am, too. So I was hoping for that outcome.

Speaker B: Uh, call me out.

Speaker C: Hope.

Speaker A: Well, it's funny. I'm just. I'm tying it back to AI. Trust me, I'm getting there. I'm going there. So, uh, the whole, like, thing was he's like, why would anybody move when they're 42 years old and have a billion dollars from LA to Philadelphia? Uh, you know, he must really love basketball. Like, you know, what are you doing, LeBron? And everyone was like, oh, well, you know what he's gonna do? He's gonna live in New York. He's always wanted to live in New York. And he'll just take his helicopter from New York to Philadelphia. But he doesn't need to take a helicopter. Like Miles said, he can drive the Honda Civic down the 95 corridor and just go, new York to Philly. It's like.

Speaker C: And, Shiv, you just lost us our entire Philly listenership here.

Speaker A: Sorry, Philly.

Speaker C: All of our, like, 100,000 listeners we have in Philly, they're gone now.

Speaker A: Wait, I'm going to. I'm going to one up and make it worse. I'm actually cheering for Philly this year. Go, Philly. Go Sixers. Because I hate the Knicks. So now I just abandoned all of our New York listenership. Sorry, sorry, New York City. Okay, let's, um, keep moving. Next, uh, story. So this one is awesome. Uh, I'm going to talk about it because it's so awesome. At the risk of just overdoing the whole AI Voice thing, because on our last episode, we spent a lot of time with Sam Bloom, who's head of partnerships at pmg. We talked a lot about how, like, AI Voice is going to be a huge unlock for the AI space and for consumer devices. And we just got to, like, figure out how to get it right. A, uh, couple of weeks ago, OpenAI had released their new, I think it was called like GPT Live new voice model, which can actually like talk to you in a more natural way without like interrupting you and stuff all the time. And then last week there was a rumor that OpenAI is working on the new device. I feel like everyone's known this has been coming for a while now, especially when they paid like some billion x billion of dollars for Joni I've's, you know, non existent company. Um, and so, uh, now there's a rumor that the device is coming soon, apparently. So we have no real details, but the reporting is that the, the device will have no screen. It'll have sensors, like motion sensors, it'll have a camera, it'll have a microphone and speakers, but no screen. And apparently the device will be able to move. So for those people that are listening, I'm doing quotations because I don't know what that means. Um, but anyways, you know, like, again, I think this has been coming for a while now. I think we had some fits and starts with devices like the Humana pin, if you remember that. That was a huge bomb, like, but that was the same idea. It's like, oh, have something on your person that you can talk to that can be in your ear that could be a personal assistant for you. Tap into all of your files and your drive and your calendar and this and that and kind of just like be with you all the time. So any, um, I think voice is just going to be this massive unlock for again, consumers. But then ultimately, if it's an unlock for consumers, it's going to be a massive unlock for marketing, hopefully. So, um, Miles, let's start with you on this one. Any quick reactions to the reported OpenAI device?

Speaker C: Does anybody else picture the little box on a roller skate that's going around the Death Star in Star wars as being the device? That's what I'm picturing here. No, literally. And, but then you said like that it moves and you used air quotes and it was like, well, actually, yeah, movement could mean multiple different things. Maybe they'll, it'll be something different. I, uh, mean, that's interesting. My main reaction is I'm, I'm very intrigued that it's not going to have a screen. I think especially with consumer AI, I, uh, would actually be worried, more worried if it did have a screen. So I'm like, if you can't talk to it the way you could talk to a personal, to A human being who you're asking to help you with something or perform a task. If I can't just talk to you, uh, then it's not perhaps particularly useful. So I really like that. Maybe it doesn't have a screen. But Jamie, I have a question for you. Uh, maybe if you can give your reaction to this. And I would love to know kind of from Klaviyo's perspective. Like Klaviyo, I would expect kind of both sort of in terms of leading and trailing indicators. Like you see a lot of the traffic of consumer interactions and so, you know, like SMS and email, voice, whatever it might be, app, you know, push notifications, all that kind of stuff. How's Voice trending in Klaviyo world?

Speaker B: Yeah, I mean, I don't think Voice has really disrupt. Well, I don't think it's really changed, uh, the market, the landscape that much yet. Uh, but this, this kind of, this innovation is not the hardware itself, which I don't know if it's a. You went to a box on, uh, the Death Star. I was like thinking microchips. So I don't know this is going to be interesting. Um, but I think if you think about marketing, we've been optimized for screens, email, tv, websites, apps, you know, social ads, everything is a visual representation of our brand. So while I said like, Voice has come into play in some areas, but I think this could be really disruptive. Right? And disruptive maybe in a good way. But I think marketers will have to adapt now to how they think about attracting via Voice. Like I, I think that'll be a new medium. So I think, um, I think it's interesting, I think it'll be different. I don't think we've seen it yet, but at Klaviyo too, I think there's gonna be many more, you know, many more mediums. And so even with the introduction of LLMs, like that totally changed the way that you market how retailers in particular, like agentic commerce definitely shifted some of the playbook. So I'm just thinking it's another channel I'm going to have to adapt. But I need to capture that information and figure out how I bring it back in to understand the buyer and what's going on with them.

Speaker A: So yeah, and I like, I think the, obviously the consumer shift needs to happen before the marketing shift. Right. And like, I just, I think about this from the consumer lens and I don't think we've ever quite figured out the right kind of, uh, the voice device. Yet, like, I think we've, you know, and it all kind of over intersects, right? With like AR and VR and the glasses and like the Humana pin and you know, and miles. It's funny, when you think of the, the Star wars thing, my brain always goes to the movie her and the creepy one. That's a creepy one.

Speaker C: Sorry.

Speaker A: Sorry for getting creepy, I guess.

Speaker B: Um, but like, was creepy. You're not creepy.

Speaker A: Thank you.

Speaker C: Well, wait, actually I have. I have a pointless anecdote to add to that too. What? I didn't. I didn't see her, so hopefully m.

Speaker B: Well, now you have to.

Speaker C: Great movie. This. This build is. Is at least in, in the ballpark is. I, uh, spent, uh, the weekend. Jamie. I just got back from a little bit of a road trip. I spent the weekend with some older in laws. Uh, and the topic of hearing aids came up. And the topic of like hearing aids today, I guess are like Bluetooth enabled. And I uh, was like, well, of course they would be. I guess now that I'm thinking about it, why wouldn't they be? And then I'm like, oh man, that means like, not only might AI just be like an earbud or glasses or something, it's literally implanted in your ear.

Speaker B: Yeah, yeah, yeah, yeah, yeah.

Speaker A: Like, I mean. Sorry. Go ahead, Jamie.

Speaker B: No, I really wonder if that's where they're going to go. I mean, kind of the George the Jetsons, right? Like it becomes part of your being, which is scary but exciting. Same time.

Speaker C: Yeah.

Speaker A: I mean the AirPods. The AirPod is kind of an on ramp to that, right? Like I have a. I have a 15 year old nephew and we'll always be, you know, wherever we're hanging out, holidays or whatever, we'll just be hanging out. I'll be talking to him and he's got the thing in his ear. He's always got.

Speaker C: Oh yeah, the kids. But they never take the headphones off.

Speaker A: I'm like, are you listening to something while you're talking to me? Like, what's happening?

Speaker B: Yeah.

Speaker A: And he. Sometimes he is. He's like, oh yeah, get some music playing or something. Like, that's really weird. Really weird.

Speaker B: That's how my children. I have two teenagers. That's how they do their homework, which I really don't understand. Like, wait a minute. Like what's going on in there? You know? Um, but have you used the voice? Because I actually think the voice, uh, part of OpenAI has absolutely. I love it. I. That's my commute now. I'm talking to my chat GPT. She has a British voice. It's quite lovely.

Speaker C: Nice.

Speaker B: Um, but I find it to be game changer how I'm interacting with the technology and I can always be on, so it's kind of interesting. So the idea of a device makes a lot of sense, but you've mentioned a couple that have had some failed attempts. Right. So I also wonder for OpenAI, like, this is a new diversification of resources internally. So I'll be interested to see how they invest and if they continue to invest to drive a chain. Right.

Speaker C: Oh, that, that, that's a whole other podcast episode of how, how thinly can open AI spread itself.

Speaker B: I just had to say it like I, I'm watching, you know, and it's exciting, but we've seen this movie before, so.

Speaker A: Yeah, well, I reason Voice. One of the big reasons Voice hasn't taken off first is because the voice technology was not good enough. Now, to your point, Jamie, the voice technology is good enough. It's, it's really damn good. And so the next thing that needs to happen is like, well, the consumers have to understand the use cases for it. There have to be some killer use cases. And like, the, the form, the form of it has to be conducive to consumers being able to use it. And like, I think putting voice on a screen. So right now we have ChatGPT on a screen. It lives in a browser or it lives in an app. And I think the problem with that is users are programmed to type on those devices. They're not programmed to talk to those devices because, I mean, and even they actually have PTSD from talking to those devices because Siri and Alexa were so bad and so they're like, I'm not talking to this thing. Like, but then if we have a new type of device that literally has no screen where you can't type, and if it has a few killer use cases, I think that's where like you might start seeing that consumer shift happening. And that's where maybe there's an opportunity for it to take off.

Speaker C: Right.

Speaker A: So, you know, we'll see, we'll see where it goes.

Speaker B: Great. And now you're a 15 year old and my kids are going to be walking around talking to themselves if this does. So, I don't know.

Speaker A: Great, great. This is a future we're all very excited about. Um, okay, let's get into quick hits of the week. So, uh, Jamie, I'm going to go through these pretty fast. If you have commentary or something to add, please jump in if not, uh, we'll just keep rolling. So, um, first one, uh, Google Images is doing a big revamp. Uh, I thought this was cool. I haven't seen it hit my Google Images per se yet. Uh, I'm guessing it will soon, but it kind of feels like they're taking Google Images towards a Pinterest direction where they're going to have a discovery kind of angle of it. They're gonna have like a feed you're gonna be able to save or like pin essentially like different types of images. You're also going to be able to create. Right. Imagery. Um, and then save that imagery. And, uh, it's going to be more personalized. It's all going to be about like personalization and discovery. Um, I thought this was really cool. Also you're going to be able to now, uh, create imagery using Nano Banana, like the Google Image model in AI mode, just natively in like Google Search. So I think this is a cool direction. Again, I have yet to see it play out. Um, but I thought this was interesting to highlight.

Speaker B: Yeah.

Speaker A: Okay.

Speaker B: One thing I'll say it's going to be interesting for B2C companies because is if it's print Pinterest, like you own the algorithm, but can and will that influence shopping behavior? So how will recommendations externally start filtering in there? I don't think they're there yet, but I was, that's where my mind went, was like, oh, product recommendations, here we go. How can I influence that? So I was thinking about that from the brand side. As a consumer, I think it's awesome. I love Google essentials.

Speaker C: Jamie, I'm curious to just hear more about like what you're imagining. You're imagining Google surfacing more product recommendations in images, results.

Speaker B: Well, I think that they're going to be building out more personalized, uh, recommendations to you. Right. So to your interest based on your behaviors. But, um, if I'm typing in yellow, uh, dress. Right. It's going to show me yellow dresses based on my search history. But my mind, again, I'm always thinking about my customers. Here I was thinking about how could I influence that algorithm to put my yellow dress closer to the front.

Speaker C: Yeah, yeah, yeah, yeah. Uh, well, I mean, like to what extent is Klaviyo sort of bumping up against perhaps this need? Well, it really speaks to like the multimodality of whatever digital marketing. And another thing. Yeah. Is like images, video, sound, a lot of stuff was text. Up until several years ago it was largely text based.

Speaker B: Right.

Speaker C: So is that changing? Klaviyo's world right now.

Speaker B: Um, it's changing my world, our world in the fact that there's just more, more, more places, more places to interact with your customer. Right. So I think text, email, on site, in store, um, LLMs. So while images. Everybody spent millions of dollars optimizing SEO for your Google rankings. Right. But like I never thought about optimizing for my image search. Yeah, that's kind of where I'm going is um, it would be another place for Klaviyo to capture information, but it could be another buying signal that we haven't really been using. So that's kind of where I was thinking, oh, go same way as search. Yeah, yeah, no, I mean it's optimized for, in a different way.

Speaker C: Yeah, I could totally see Google kind of turning this into, well, perhaps what Pinterest has certainly always aspired to be is kind of a mood board for consumer intent. But obviously uh, Google has a much more closed loop system to make that happen versus a Pinterest. Yeah, yeah, yeah.

Speaker A: Like doesn't at some point won't Google shopping kind of intersect with Google images in that way?

Speaker B: You would think. I mean, we'll see. That's my prediction. You heard it here.

Speaker A: I mean that might be annoying. Like I'm just thinking about how I use Google images.

Speaker B: Ah.

Speaker A: And I'm like going to look for memes and stuff, you know, or like going to, going to like find a silly image to drop in an email to my friends. And so all of a sudden if I start getting product recommendations in there, I might be annoyed. But yeah, who knows? Um, you better believe Google wants to monetize the hell out of every, every surface they have. So um, possibly. Okay, uh, okay, next quick hit of the week. So kind of a small one but I think this is important and uh, you know, uh, something to talk about. So chatgpt now, um, you know they, they've been evolving their ad product over the last four or five, six months and they just added a capability where you can upload your client list, you know, just an Excel file of all the emails, uh, apparently without a clean room or any, any sort of like data, data collaboration platform. Just Upload your list into Chat GPT, they'll do a data match and uh, then you can target your users on ChatGPT. So the very simple use case is retargeting. Okay, well I have you know, 100,000 people that have gone to my site, uh, that have set up an account, maybe bought some products and now all of a sudden I can retarget them on ChatGPT which you know, again it's like, it feels like a small, simple thing and it is for the most part like you can do this with any major ad platform today. But I think it's interesting because of the AI experience, right? Like the fact that people are not like doom scrolling on AI like they do on Meta where it's like yeah, if I retarget somebody on Meta, that's really, really valuable, right? Or on Instagram like that's extremely valuable. I could bring them back, they could then purchase, great, everybody wins. But in AI, like there's all this context history about the user, the discovery and um, and um, like the funnel kind of collapses in AI. So what does that mean from like a first party data targeting perspective? I find it somewhat interesting to think about and go down that rabbit hole. Any thoughts on that one?

Speaker B: I think it's always scary to give away your first party data. So I double triple check that, you know. Uh, sure, but I mean clearly and we were early adopters at ah, Klaviyo of advertising, um, and Chat GPT. So I've kind of been, you know, have my toe in the water there and yeah, clearly it's, it makes sense. Uh, very competitive with Meta. I don't think they're there yet but I understand the path that they're on. Um, and if I think about for my customers again, um, retail agentic commerce, like you want the ability to connect and find those relationships. So um, interesting. But man, I hate giving away my first party data. So I cautious, cautiously optimistic maybe is how I would say it.

Speaker A: Well, so I played out like a scenario in my head a little bit Jamie, where I was like, okay, how is this different from like just retargeting on uh, meta, right? And so the, the, the little example I came up with was like if I am trying to, if I'm like looking around for shoes, I go to a sneaker site, I drop some shoes in my shopping cart and then I kind of forget about her. I'm not sure I want to buy them. In that moment I move on, right? Two, three weeks go by and then like I'm on ChatGPT, let's say a few weeks later and I'm like doing some research about a trip I'm about to go on. And you know, people tend to buy things before they go on trips, right? Oh, I want new shoes, I want to look nice on, buy new apparel, I want to look nice, whatever on my tr. That to me is like a really Interesting moment to surface. Like that shoe ad again.

Speaker C: Yes.

Speaker A: And like Google and Meta would just never have. I mean obviously Google would through Gemini now. So let's put Google aside. But like Meta wouldn't have that kind of signal. Right. It would just be like blunt target, blunt retargeting. Um, and so I just find like the, the idea of like context.

Speaker B: Yeah.

Speaker A: Within ChatGPT and like how context can be used alongside data. Right. To surface ads at better times or surface different types of creative or like all the other different variables you have to pull with ads. Um, anyways, that's kind of where I was going with it.

Speaker B: Um, but for me, and I'm totally biased because at Klaviyo we do, we believe that you should be able to look in Meta. Uh, you should look on your website. You're going to look in the LLM. How do you capture all those signals so that you can be ready? Well, because you can know your buyer and then you can be ready to offer when they're ready. So I totally coming from the place of like, okay, more is more, surfaces are good, signals are good. The way people buy and the way people search, the way people learn is changing. And so I think for, I always am advising our customers, you got to capture all that. That's the most important thing. So having integrations with LLMs with you know, shop commerce agents, wherever they are, this is becoming more complex. I think it's not as one and done. Um, and that's why, yes, I'm advertising in ChatGPT and Meta and, and, and, and more and more places.

Speaker A: Um, okay, I love that. Okay, last quick hit of the week. So, um, Meta has tons and tons of AI tools for marketers, including like creative generation tools, video generation tools, optimizers, you know, performance plus. Wait, what is the advantage? Plus is their thing. Um, so apparently Meta's AI tools have been bungling, uh, a lot of the marketing campaigns in different ways. So like the one big example that kept coming out was the uh, bike with two handlebars. So it's got handlebars in the front, it's got handlebars in the back. And apparently Meta's like, well, sorry, it's on you guys, I'm Mark Zuckerberg. Screw you. So, um, that's going on and obviously that's like a huge pitfall of anything AI that you do. Human in the loop is really, really, uh. But yeah, that's, that's a thing. That's, that's happening at Meta and, and I'm sure like that's happening everywhere. So, uh, with that, I want to transition Jamie into your work at Klaviyo and what you guys are doing and what you're seeing. So, um, you know, I, I did, I did a little bit of my research, a little bit of my homework. Like, I think it's fascinating to see the journey of like, you know, marketing automation and how it's gone like pre AI to post AI. Right. And you guys, I think are leading the pack in this respect in a lot of different ways of like, okay, pre, it was very rules based, right? So it's rule, rules based marketing automation. Hey, we're going to set some rules triggers, things like that. Um, but it was all very deterministic for all intents and purposes. And now post AI now you can kind of build agents in these platforms that can access the data, access all the different channels and be more probabilistic, right. And be a little bit more autonomous and make smart decisions based on, you know, maybe some triggers. But like, the triggers can be way more vast, right? There can be all sorts of different permutations of triggers or they don't have to be triggers at all. And so anyways, I would love to hear kind of like your strategy, you know, as it pertains to AI, uh, how you have seen that shift with Klaviyo specifically. And then we could kind of evolve the conversation from there.

Speaker B: Well, I mean, I think it's an exciting time. You said autonomous. I love that because we're the B2C CRM, but we're really, we see the world. In the future, we'll be autonomous. So there will be a time when agents and everything's running our world. We're not there yet. Clearly. Exactly why you need a human in the loop to make sure you don't have two handlebars, all that good stuff. Yay. I still have a job, but, um, you know, I think about at Klaviyo, uh, this isn't new, but you know, we've been focusing on sort of that personalization, uh, down to the individual consumer for a long, long time. But things like send time optimization, you know, back to triggers like, which is built into our platform where you could just click a button. And now it will look at your user's behavior when they like to open an email, when they like to read a text and it will optimize your send for you. Like that's.

Speaker A: Yeah.

Speaker B: If you go back in time 10 years ago, like, come on, that was such a nightmare. I was sending emails to APAC in The middle of the night because it was just all you could do. And so I think the future being autonomous is where we're going. However, I think there's like these obvious steps that are already available, whether it's send time optimization or a B testing, um, auto testing headlines for you. And now like we just launched our own agent, um, which is actually building emails for you just based on your website, just based on your looking at your, you know, your kind of public facing information. We can build an email, we can use Nano Banana, you can remix your image. Like I don't know, I think, I think there's just a lot at your disposal today that allows marketers to be way more strategic about driving that personalized message. Which is kind of our end goal. Right. Always.

Speaker A: Yeah.

Speaker C: Jamie to Shiv's points or this whatever news about Meta's, some of Meta's AI tools going wrong. AI tools go wrong wherever. Uh, what's Klaviyo's like evolving approach to guardrails? I mean like Klaviyo clearly operates at like massive volume, you know, across clients, their clients. Like I imagine you have like hundreds and hundreds of millions of individuals kind of who are communications being transacted all the time. So like what's how, yeah, how are guardrails shaping up?

Speaker B: We have like 200,000 customers, eight ah, billion messages a day. So we got a lot. Yeah, yeah, we got a couple things going on. Yeah. Um, so guardrails are especially for the marketer brand guardrails are the most important thing. So I think that um, we have two agents. One's external customer facing and one's internal. You know, more productivity around campaign creation and um, having the right brand guardrails is so important. So that's built in in both. And then two, I think we think a lot about how you train the agent. So for our external agent, like um, Naked Wardrobe, they use our customer agent and the agent is acting on behalf of them. It's an extension of their brand. Whether you're asking to return something or uh, you know, making a product recommendation, it has to be in your tone, in your voice. So we've invested a lot. And the agent training, the agent skills and then the ability to control your brand, um, your brand guidelines and that's all built in default from day one. So I think that's another place where people get it wrong. They're like, get an agent, they turn it on, they let it rip and then it does crazy stuff. Well no, you gotta train it. You gotta really put the time in up front. Uh, so we've invested a lot in that sort of front end onboarding experience of the agent to really train it up.

Speaker C: Yeah, yeah, yeah. Cool. Jamie.

Speaker A: Uh, qu. So like let's. I want to play off that a little bit. So you know, you're a marketer. You mentioned you have empathy for the marketer keeping their job. Right. And like, um, we want, we want everybody to still have jobs. And so like, you know, as you guys, ah, as Klaviyo, push more towards AI, more towards identifying things, like you're making things easier, you're making them more autonomous for marketers. Um, obviously the work that people do, uh, will change, right? Um, and some work will get reduced, some work will get reallocated, et cetera, et cetera. What I'm interested to hear your perspective on is like, you know, let's just talk about current state, like the current state of your platform today. Where do humans still have like the most impact in working within an environment like Klaviyo, Um, versus like humans that are using it, maybe, and not like where, where are the marketers kind of creating the differentiation just within the context of like Klaviyo. That, that's.

Speaker B: Yeah.

Speaker A: It's something I'd be interested to hear your thoughts on.

Speaker B: There's still so many of our, As I mentioned, 200,000 customers. A lot of them are still trying to figure out AI period. Like, okay, I got a lot of going on. AI sounds cool and all, but like, I gotta sell this, you know, dress today. So, um, one, we've been really thinking about how do we embed AI into the workflow that they're already using. So something that I m mentioned, like real time personalization that's in the. Where they work right in the day to day. So I think that's been really important to how we're thinking about it. And then two, um, I think people are less feared about their job being replaced, but their job changing. So in the spirit of Klaviyo, there used to be somebody who would wake up every day and they'd go build a campaign. They'd build the brief, they'd uh, add the images, they'd size the email. That job is no longer there. Because now you can come in, type a prompt and say, build my campaign and it will build it for you. But I still need a person, um, who can look at the objectives, help understand, like I said, train the agent and um, input what are the business challenges of today and work with the agent to get to the right conclusion. So there's still that interaction. And then I need someone to oversee it. I'll show you two emails, but like, you decide which is right. And as you train the model, it's going to get better and better and smarter and smarter. Um, but I think the companies who are embracing that, they're really freeing up. They're building more, they're building faster, they're personalizing a lot better than the competition. Um, and then they're also freeing up more time. Right. So that they can spend time in other places. And people say, oh, the brand's gonna go. You're not gonna need a brand team anymore. I'm like, oh, you're not. Brand is what's going to differentiate us. Right. And yes, I could do an LLM and they can give me pictures, probably with two handles, but that'll get better. Um, but the brand will still need somebody to build in that differentiation. Right. And make some judgment calls. So I think, um, that what you do the day to day is shifting. The companies who are embracing it are going to move faster. And we're already seeing that, um, there's still some really key areas in marketing in general that are going to be really important to the future when things do become more autonomized.

Speaker C: Yeah. If your brand could be reduced to a prompt, maybe you didn't really have much of a brand to begin with. Uh, so I have a question, Jamie. You have Digital is of course a training company, and you were talking about how there's segments among your customers. I imagine there is a relatively small but growing percentage of innovators who are racing out ahead of the pack. A bunch of people in the middle of the pack, maybe a few laggards who are like trying to catch up. And you certainly want them to catch up because you want them to get like the full value of the platform. Um, uh, and I don't know if, uh, customer education is like under your remit as cmo, but just would like be curious as to where Klaviyo is right now with respect to up leveling your customers, given that in the last couple of years the entire tool set is just being turned upside down and their job roles. To your point.

Speaker B: Well, and also this, uh, is such a fun question. I'm glad you asked because also the idea of releases, like having three releases a year, that is gone. Those days are gone. So the way that we would product market five years ago, oh my gosh, forget about it. And um, to your point, I think with things like OpenAI and Anthropic now, they're releasing a new model every day. Right. So you have this. Consumers have built a tolerance for um, change I think, which is good. But I've had to really adapt, um, how we communicate with our customers. Hey, you know what? Isn't that we used AI actually. So now we've actually built an entire end to end workflow starting from linear. So now when our engineers actually release a feature, we have a whole workflow that automates that update in product product on the website, on our what's new page. It sends a slack to our internal teams that need enablement. Right. And also because I'm a marketer, it creates a blog post, uh, and pushes it out into the world. Uh, so that's a place uh, where AI, I feel like has really unlocked our ability to help uh, inform our customers with what's going on. Now our customers still have to consume that. So I think like 12 bug releases, not that we have all that, but you know what I mean, if that's what's coming out that day, they might not want all that. So now we're trying to build in the controls of what do you need to know and when do you know it? Um, but yeah, it's kind, it's kind of an interesting time. I would say the traditional release has been really disrupted.

Speaker C: Yeah. Yeah, that's, that's really interesting because yeah. You spent time at Salesforce I was looking at. Yeah. So yes, your, your career goes back far enough that like you. Yes, it has very much changed in the last couple of years in terms of pace.

Speaker B: Yeah. Oh yeah. I was at Salesforce for a decade and we had three releases a year which was cool because you could plan for it. But I also ran our adoption at scale. So that was when I first started, um, thinking about how we're communicating post sale changes to our customers. And it's come a long way. I mean back then we were doing like nurture track which those, you know, those are fun. But now it's more about this like personalized communication, um, and trying to make something that's always on so that consumers can actually consume it. Otherwise, I mean we have partners who are coming, we're innovating so fast, which is a great problem. But sometimes our partners, our customers, like wait, what happened? What is this? Where did this come from? I didn't know about this. Yeah, it's the speed of innovation which is exciting.

Speaker C: Yeah. Like our mindset, a lot of it at UF Digital right now because we've got our own version of that being actively built out behind the scenes at U of Digital. We're not a software company, so it's a little bit different, but we're also working to streamline these things of like, well, drafting the blog post should no longer be the bottleneck. It's still an important thing. But like, those things used to take weeks, sometimes months to draft those things. It's excruciating. We really can't afford to do that anymore. Um, and then just automating or at least trying to streamline all of these touch points to get a prospect or a customer or a partner up to the point of like that interaction that can't be automated. Like a meeting or a webinar or podcast, something like that, where it's like, yeah, let's set aside all that excruciating stuff and really focus on that whatever last yard that really matters in terms of that customer interaction.

Speaker B: Yeah. I mean, I also think that one to one or the human interactions are going to be more at a premium, which I'm excited about. Like that.

Speaker C: That's right.

Speaker B: I'd rather talk to people, you know, so automating some of that sort of, um, important work. Right. But that scale work so that I can divert more resources to field marketing or brand marketing. Like, I'm, I'm excited. I think that's gonna be a positive shift.

Speaker C: Yeah.

Speaker A: Uh, well, Miles, you kind of stole my last question. Which was. Which was. No, no, it's great. I think you answered. I mean, the answer was awesome. Right? Like I was gonna ask you. Jamie, I know you're CMO for a B2C marketing company, but you are the CMO of a B2B company. It's like confusing and weird and meta in a weird way. But like, uh, you know, how is AI changing B2B marketing for you? But I think you gave like a great, like that explanation that you just walked us through that. It's like, miles, we gotta go build what Jamie's building. That's great. Let's go do that. Um, okay, Jamie, we are about to wrap up any final kind of notes for the listeners on like, what's coming at Klaviyo. Any final, like big tips or tricks that you have for marketers around? Like, hey, it's, you know, for thinking about AI stuff like think about this first. Any parting thoughts for the listeners?

Speaker B: I think we hit on a lot of it. But I will just say I think the future is bright. Uh, and I think agents. Uh, our agent is a composer, which is. Makes your job easier, turns you from a building a campaign to building a prompt to build a campaign for you, which is great. Um, and things like customer agent that can act on behalf of your brand, whether it's. It's marketing or support. It should be one voice, uh, which is our customer agents. I think, like, AI should be an accelerator. And there's a lot of cool tooling already built for you, where you work, how you work. So, um, you know, that's. That's where we're going. That's where we're investing. And, uh, you know, I'm excited about it.

Speaker A: Awesome. Thanks, Jamie. Thanks for joining us. And actually, on that note, uh, we have an AI accelerator coming up in September for marketers. Uh, it's our flagship bootcamp. We run it every few months. We've just opened enrollment for the September version of it. It's being sponsored by Bliss, which is, uh, dsp, owned by T Mobile Ad Solutions. So, really excited about that. Enroll if you haven't yet. Thanks again, Jamie. This was an awesome chat. I learned a ton. Thanks, Miles, as always. Thanks to our listeners, and we'll see you guys again next week.

Speaker C: Thanks, Jimmy.