The Math of Marketing: George Khachatryan on Reinforcement Learning and AI Personalization

Step into the world of success with Your Brand Amplified, the ultimate podcast designed to unravel the intricacies of thriving businesses. I'm your host, Anika Jackson, and I'm on a mission to uncover the stories, strategies, and secrets that have propelled entrepreneurs and business leaders to success. My guest today didn't just start out in marketing. He actually has a PhD in mathematics from Cornell, George Hotchatrian's first company was Reasoning Mind, an education technology company that went on to serve over a million students. From there, he spent years at McKinsey as an associate partner leading business transformations, and then did that thing that a lot of consultants think about and never do. He left and built something. That something was OfferFit, an AI decisioning engine that replaced the manual grind of A/B testing with reinforcement learning agents that experiment and learn on their own. He co-founded it, served as CEO from day one, and led it all the way through the acquisition by Braze in 2025 in a deal valued at $325 million. Today, he's the head of OfferFit by Braze, and he's arriving with something a little bit uncomfortable in hand. Braze just released its 2026 Global Customer Engagement Review, over 2,000 marketing executives and 4,000 consumers surveyed. 93% of marketing leaders believe AI is helping them understand their customers. Only 53% of consumers agree. So let's talk about the math, the exit, and the trust problem sitting underneath all of it. George, welcome to Your Brand Amplified, and I do have to let you know, I actually use that study in my classes that I teach at USC for the Master's in Digital Media Management. So I'm using all of the things, materials that your company is putting out. in my classes to talk about what we need to really be looking at in digital media. So, I'm doubly appreciating you being here today.

No, that's awesome, Anika. Thank you so much for having me. It's great. It's great to be on the show.

Yeah. So, talk about, first of all, the career, right? And all the careers that you've had that wouldn't necessarily be a through line, mathematics, edtech, management consulting, AI marketing. So, let's look at that. What was that through line? And And what took you from one to the next?

Yeah, well, it was, you know, almost everything that I've done with the exception of the time in consulting was various things that were like math related. So that first company, Reasoning Mind, we were teaching kids math online and we were, it was one of the earliest intelligent tutoring systems where we were, we were using the AI tools that were available at the time in order, which were obviously vastly more limited than what we have these days to personalize instruction to students. And what brought me there is I think I was excited about math. I was excited about teaching math. There's just things that I was like personally passionate about and I really enjoyed building companies. And so it, it kind of came together naturally. And it was actually a family project with my parents. So we, we worked on it together and we were the, we were the co-founders. And then after that consulting was because I felt that I had had a very insular career to that, to that point. I had followed things that I was very passionate about in math, but I hadn't really learned from other people of how do you, you know, properly build companies. So I spent a few years as a, as a consultant, and it was a chance to learn from my clients, from the other consultants at the firm I was with. And it was, it was kind of like, like, uh, like, yeah, like going back to school a little bit alongside obviously a chance to work with clients and, and, and help them. I mostly did, I started as a generalist and then I specialized in turnaround and transformation work because that was very fulfilling because it was very hands-on, you know, in the trenches with our clients, helping them actually, actually make changes in their companies. And then for, for OfferFit, my co-founder Victor Kastouk, he and I met in grad school. We did our PhDs together at Cornell. And then we worked together at the edtech company Reasoning Mind. And we, we had always worked together very well. Uh, we're best friends and we knew at some point we wanted to start a company together. And then Victor had this idea. And it was also a math idea, perhaps not surprisingly. So Victor at the time was with the Boston Consulting Group and he was building custom machine learning models for marketing personalization. And he was using this approach at the time, which was the state of the art as of a few years ago called next best action, where let's say you're a marketer, you want to personalize how you market to different people. You would start by setting up or having your data scientists set up predictive machine learning models. So these were things like propensity to churn that would predict how likely a customer was to churn or affinity for different product categories that would predict how likely the next purchase was going to be in a certain category and so forth. And the scores that these models output, they're not yet decisions on how to market to people. They're predictions on how likely someone is to do something. And so then you turn them into a decision by having some kind of rule-based system. Like a simple way you could do that is like, okay, whatever is the category that they have the highest affinity for, market that category to the customer. And so that whole approach is called next best action. There are all sorts of problems with it that Victor experienced. First of all, that decision seems to be one-to-one because you might be using all your data. It's not really one-to-one. because you're still basically segmenting people by the scores.

Right.

So for example, everyone who has an affinity for that category, you're going to market the same way to them. It's like, so you're not really being one-to-one. You're still being segment-driven. You're not really personalizing as well as you could. And so that was the most obvious, but also it was like very, very difficult to set up these things and maintain them. The rules were often like made up. Like, how do you know those are the right rules? So there were these challenges and Victor had the thought that, hey, there's a type of machine learning called reinforcement learning. Where instead of having these predictive models and then have rules, you can just set up a reinforcement learning agent, which you can think of as this, you know, ML thing that autonomously experiments and decides. So you're automating the whole thing, soup to nuts, where it makes decisions at the individual customer level. You know, for Anika, I'm going to send this message at this time with this offer. For George, I'm going to send this other message at this time through this other channel. And then it learns through what comes of that. And so as a result, there's no segments, it's truly one-to-one, there's no rules anywhere in sight, it's all truly, truly automated. And he was doing this as a consultant at BCG, and then he had the thought this could be productized. And so he talked to me about it, we were already thinking about starting some kind of company together, we talked to some marketers that we knew. And to your point, I wasn't a marketer, so I had no clue about marketing at the time. I was just a math person who wanted to work with his friend Victor. But then I had friends, fortunately, who were marketers. And I talked to them and I said, hey, is this stupid? Like, or is this potentially useful to you? And then they were like, no, no, no. Like this, if someone could build this, like we would want this. And when we heard that from enough people, we're like, okay, I guess it's worth, it's worth doing. And that's how we started it. And, you know, we used to joke that OfferFit wasn't really a technology company. It was a math, it wasn't like a software company. It was a math company because a lot of the nuance of getting this to work is actually figuring out the math. Like, yes, you also need to program and have all the interfaces and all that stuff. But at the end of the day, like this stuff lives and dies on the quality of the data science, which is essentially math. And so, yeah. And then eventually we can talk separately about joining Braze, but that's been the most recent thing. And it's been great. It's been great being, being part of that team.

Yeah. Fantastic. And you actually answered a lot of my questions with that, which is great because this is something that we are all talking about, whether I'm teaching for the Digital Social Media Masters, whether I'm teaching for Digital Media Management. Whether I'm talking to clients, whether I'm talking to educators about how they're using AI in the classroom, or even for my own MBA program, right? We're all talking about this is the best possible use case of this technology. It's not to create content that may be the same as everybody else's content necessarily. It's really to drill down the analytics and the data so that you can be super hyper-personalized. And so, you're already leap years ahead of where a lot of marketers are and where they aspire to be.

I think it's a very rapidly shifting landscape. And yeah, I think, I think there's the capabilities over the last few years have become so much more powerful than what was the case before. And I think marketers are, are catching up, but obviously every company at its own pace.

Yeah. Can you give us an example of a client that walked through the process or even yourselves deciding to market this to other people as a test case?

Oh, sorry, can you repeat that?

Yes. So, can you walk us through a potential campaign, a brand client situation that somebody was trying to figure out how to change, how to get a better measurement system in place and get to that point of being able to hyper-personalize one-to-one instead of We know that this segment likes the color purple, even though they all might like different things within that color, different brands and products, right?

Yeah, yeah, for sure. Like one of our customers and, and the, the is, um, is Yum Brands, which is, which is the world's largest restaurant company. They have Pizza Hut, Taco Bell, KFC, and their chief data officer actually spoke at our conference Forge last year, Cam Davies, and he kind of He talked about this. So, you know, very sophisticated company, obviously. And also one part of why I like this company is it's so like personal for so many of us, right? Like these are, these are restaurant brands we grew up eating. And you can imagine when you're communicating with folks, let's say you have a loyalty audience and, you know, folks who have signed up for the app, entered the loyalty program for one of these restaurants. It's very important for the brands to be able to communicate with that loyalty audience. And try to get folks to engage there, as opposed to if they have a more impersonal relationship with the brand and they might be making purchases through aggregators, which obviously is much worse for the margin profile. And also just, you know, long-term loyalty. And so if you communicate to people, there's all sorts of questions like, okay, which of the menu items do you talk about? How much humor should you use in your messages? What time of day do you communicate? Do you communicate before lunchtime? Do you communicate late at night? Do you communicate in the evening? And we all have very different consumption patterns, right? We have very different times, not only when we eat at these brands, but also, maybe we eat lunch, but emailing about lunch isn't going to help because that depends on my work schedule that day. So just because I eat it mostly for lunch doesn't necessarily mean that 11:00 AM is the right time to email me. Maybe still 11 AM is just not incremental. The right time to email me is in the evening because in the evening it's discretionary, whereas during lunch, like it's not, it's just going to be determined by my work schedule. There's all these nuances. And so before, before using our Embrace AI decisioning, you know, young brands would still do A/B testing and A/B testing lets you discover what's a good subject line, what's a bad subject line, but it doesn't get you to that level of personalization. And you can get, you know, a level deeper by segmenting. You can say, okay, for the people who are in their first 3 purchases, we're going to have a separate journey. You know, for the people who are, who haven't purchased in a while, we're going to have like a winback journey. So at that level, you can segment, you can get some personalization, obviously, but ideally you want to go a lot deeper, you know, within that winback, you want to be sending really different people to folks, really different things to folks. And so that's where they started using Braze AI decisioning. in conjunction with the Braze engagement platform. And so there you can set up the Braze engagement platform, which is where you make all of your creatives. It's how you can, you know, execute sending messages to individuals. You can put all sorts of, you know, suppression rules and guardrails and so on. And it's very, very scaled real-time infrastructure. So you can imagine for these companies, scale is extremely important, having infrastructural scale, and then implementing AI decisioning to be able to autonomously decide and actually discover Based on the entire population of folks, not just Anika, but folks who have had similar behavioral patterns and look like Anika, what is the incremental action to take? Do you send them an email with emojis or no emojis? And that can drive quite a bit of uplift. So, you know, when brands do this, they typically see whatever their current journey uplift is over holdout, over do nothing, you see 50% to like several X improvement on that.

Wow. Yeah, it's truly amazing. So we've kind of gotten into the next phase, right? So can you talk about what was the decision process to become part of Braze?

Yeah. So the way that we always thought about AI decisioning is that it, it sits between the data layer and the kind of traditional customer engagement layer. So you've got your data systems like warehouses. and customer data platforms. That's where you unify your customer data, you get your 360. Then data from there flows into your decisioning engine, like Braze AI Decisioning Studio, and the decisioning engine makes choices. George is going to get this push message at this time, Anika is going to get this in-app message at that time, et cetera. And then those get executed through your customer engagement platform. And you can think of it as a loop, Because then when it's executed to customers, then they take certain actions and then that ends up back in your data systems. And that was kind of the state of the world as of a couple of years ago. And we actually, we weren't planning on being acquired. We had offers from investors for a Series C at the time. And then when I kind of, you know, we were already a partner with Braze and Braze approached us and said, hey, like, would you be interested in joining forces? And for us, actually, it was interesting to have the conversation.

Yeah.

Because since we were kind of neighbors in the stack with Braze, we knew the system and of all the products that were the kind of enterprise that were widely used by enterprises, that was the one we most respected and admired. And we also knew, you know, we liked the team because we were partners with Braze. And we also saw that there was still like product energy that Braze like shipped. And a lot of the alternatives in the space were more kind of previous generation technologies where we weren't as excited about the technology. And we also felt that if we ended up joining those companies eventually, that we would be less likely to make progress on our product roadmap because just the culture of those companies was just less, less hustle, less shipping, you know, less, less momentum. And so for those reasons, like, yeah, let's have the conversation. Because if we're going to be acquired by anybody, like Brace is our top choice. So like, let's, let's talk to them. And then as we had those conversations, we kind of really liked the Brace team. And the thing that I think ultimately made us say, yes, like this is the right time to do it is if you look at the integrations, the data system and the decisioning engine are a very clean, one-directional integration. So you've got your customer data, it needs to flow into your decisioning engine. And these data systems are like built for that. Like that's the whole reason they, you know, a big part of why they exist as products. The other integration between your decisioning engine and your engagement platform is much more complex. So for example, let's say if you're authoring creative, like you create a push notification or an email, information about that creative needs to be given to your decisioning system so that your decisioning system, it's not just like an abstract creative number 53, it needs to know structure about it. Then when you make the decision to activate, that needs to be activated to people. If a customer engages with it, like clicks, that's important information. That information needs to be passed back to your decisioning system. If there are certain suppressions that are in your engagement platform, like don't send, you know, like frequency caps, if that information doesn't get back to your decisioning system, it could cause problems with your data science. So it's a very, very, it's a much more messy integration. And so what we realized is you can be a pure play decisioning engine, but long term, it's probably going to be very attractive to have decisioning engines that are also very deeply integrated with engagement platforms. And then Braze is the one we'd most, most like to be a part of. So that's why we decided to do it.

Nice. And you're able to continue to innovate within the platform and continue the work that you've started, that you've become passionate about, this thing called marketing.

Yeah, it's been great. It's been all like, like if anything, it's accelerated our roadmap because, you know, Braze obviously has a very very strong, like top in the industry engineering team. And so there are a lot of things, you know, we were always very strong in data science and we were decent at engineering. You know, I'd say we're like, we're, you know, best in class data science team and machine learning team and a, you know, decent engineering team. And Braze is a world-class engineering team. And so being able to combine with Braze and get access to all the infrastructure and expertise has been very, very helpful.

Let's talk about that gap that I mentioned at the beginning, right? The gap between Brands and marketers thinking that AI is helping them understand customers, but customers feeling like brands are not necessarily. I mean, 53% of customers feeling that brands are actually predicting, that's a huge gap. What do you think is causing that difference?

Yeah, I mean, I think there's probably many things causing it. One is customers now are working with a lot of, they interact with AI in their daily lives. You know, everyone is talking to ChatGPT Claude and so forth. And they see how magical it can be. And brands, you know, depending on the brand, a lot of them are being very cautious. And it's also, you know, it's, it's tough to have change happen in any large company. And so I think there's a mismatch where customers know that they can go and have this magical experience talking to ChatGPT. And then if they don't have similar experiences with the brands, then, you know, they're sitting here being like, this could be better. Meanwhile, the brand is sitting saying, oh yeah, but we've, we've done these, you know, 57 AI initiatives this year. You see, we did this and we did this and we did this. And that, you know, both are true, but the bar for consumers now is very high. And if brands aren't, aren't meeting that bar because they're slow moving because they're large companies or because they're being too, too cautious or, or just because it takes time, then that creates that mismatch. So I think that's a factor. I think AI is extremely controversial in general right now, right? Like people have very mixed feelings about AI. It's obviously an incredibly powerful technology. It's also kind of very personally threatening to almost everybody, right? Like everybody is wondering, what is the future like? Will I have a job in 5 years? Like, I don't know anybody.

Will I have a job when I graduate with a master's even?

Exactly. Yeah. Yeah.

Yeah.

Every single student is asking, right? I graduate with my degree, like what will happen? And the closer, you know, it's not like getting close. It's not like it's some kind of unsubstantiated panic in the Bay Area and in the tech community. There's perhaps the highest level of concern with this where I think I read an article this morning that the general, the best description of the state of like everyone right now in the Bay Area is hysteria because, and like warranted, right? Like we're all kind of in this existential questioning period of like, what's, what's our future as reasoning entities when now we've created these other reasoning that also do this thing that we thought was like just ours to do. So I think that's causing, you know, at the same time, it's a technology that's incredibly powerful and that's obviously creating so much value already for, you know, everyone. And that creates a lot of tensions where folks are in many cases both enthralled by AI And also quite uncomfortable with it.

Absolutely. And I think we're seeing so many push-pulls when it comes to marketing and when it comes to brands. Brands like Apple having to prove that they didn't use AI to create a new logo or to create a commercial. As just one small example, we see a lot of, you know, the AI influencers that now, if it's a short-form video, you can't necessarily tell that it's not a human promoting a brand. So I would say the content play is one of the worst cases. It's the one that's, it's the easy one for anybody to use, but it's not the best. Getting into the data and using predictive analytics, recommendation engines, that part is the part that actually makes sense and puts the right message in front of the right customer at the right time.

Yeah, well, and I think, you know, our view is that the future is going to see hybrid systems that use, because there's value in these different All the different types of AI. So for example, one of the things that, that a lot of brands are starting to do is use LLMs in order to personalize copy that they send to individuals. And that can be, you know, we've seen a lot of evidence of Braze customers doing this because we have a product called Agent Console that lets you do this. And so we see our customers that use Agent Console get, you know, huge lift Like it can be over 100% performance improvement. And you can sort of understand why, because you can, you can take the information about your interactions with the brand and then write something that's much more relevant in the information. And of course, that's, that's much more compelling, right? You don't feel like you're being spammed. You feel like the brand actually went to the effort to send you something that like, you know, is for you. And then meanwhile, decisioning using non-LLM machine learning models is also quite important because that's how you can figure out statistically, Right? What's the best time to communicate? Which, you know, product types are the best to talk about? And so I think that the most compelling systems that marketers set up are going to be ones that take the best of all these different flavors and stitch them together in the best way. So for example, you could imagine using an AI decisioning engine to decide what time to communicate and what message to send, maybe even what tone to have in that message, but then you could use an LLM agent to personalize the copy within that message. And then you end up, you know, or you could even use your AI decisioning to decide when to use an LLM agent versus not, because that way you can learn statistically who are the customers that maybe don't want the personalized message and they better resonate with a more, you know, neutral message versus the ones for whom it's going to be much more valuable. They'll far prefer get an LLM personalized message. And then you can, you can make those decisions accordingly. So that's where I see it going in the next 2 or 3 years. We're still seeing just the very earliest examples of people starting to do that. But I think, I think long-term, that's, that's what's going to perform the best. And it's just a matter of time before, before folks implement it.

So that takes me to a question on the consumer side, and then a different question on the small business side. Maybe somebody doesn't have This size of infrastructure. On the consumer side, when we think about consumers, you know, that your report saying 43% of consumers would stop engaging with a brand if their data were misused. Yet we see that in the United States, it's very patchwork. Of course, we have Europe, the GDPR, and their AI regulations are much stronger. There are other countries that have different regulations. And so, It can be really difficult to get a sense of, is our information being used correctly? Right. And I think brands often harvest a lot of data on customers and from my perspective, aren't using it in the right way necessarily. That doesn't always mean a leak. It also just means that they're sitting on a lot of data and they don't know how to interpret it. So, what do you think about that? What would you—

Yeah. Yeah. It's a great question. I think it, it, It actually, it ties back to some other points you were, you were making on, you know, the provenance of certain information or images or logos, right? People wanting to know, was it AI created or not? I think that, so there's a really important trust component, like trust and transparency. And maybe the best way I can illustrate this is with a, with a personal anecdote. A few years ago, I was living in the Boston area and we were shopping for an apartment and we found an apartment building that had, you know, some rentals we liked and we emailed them to see if we could see, you know, if they could show us the apartment and got an instant response from Sydney. They're, you know, an agent from this apartment building. And Sydney said, yes, absolutely. We'd love to have you come in. You know, here's some floor plans that we would recommend and, you know, let's schedule, you a visit and here's some times. And I wrote back saying, that's awesome, Sydney, you know, let's do, you know, this time. And then Sydney, great, you know, confirmed. See you then. And so to me at this point in the interaction, I was thinking, this is great. Like I love this company. They were so quick to respond. It was personalized because, you know, I said we need, you know, 2-bedroom floor plans and they gave us 2-bedroom floor plans and it was all, it was great. And so then, uh, you know, right after the workday is over, My wife and I pick up our son from daycare. He was 1 and a half at the time. And we go, you know, we take the train and we go to this apartment complex. And, you know, it's, it's, it's a bit of a complicated procedure because he's pretty tired by the end of the day, but that was when we could do it because we had, you know, work the whole day. So we're like, you know, it's fine. So, you know, we, we show up at the apartment building and the kind of the, there's only one person there, kind of the person at the front desk. And he says, oh, sorry, I don't have anything for you. And I'm like, no, but we, we had an appointment that we had made, you know, Sydney confirmed it. And he said, oh, Sydney, Sydney's just a bot. Sydney, Sydney's not an actual person. And so tough. I'm like, well, wait a second, but you know, she's your bot. Uh, and he's like, yeah, I guess it just, you know, when Sydney replied to you, we were already closed for the day. And so, and so tough. And I think it was so frustrating, right? Because here, you know, we are, we, we, we, did this big trip and, you know, with, with, by now my son's like very upset and tired. And now we have to go back and do it all over again. And then if you think about it, like, what was it about it? It wasn't just that this is inconvenient. We felt betrayed, right?

Yeah.

Like we felt that Sydney had held herself out to us as a person and we had trusted in this person. And then it turned out that Sydney was not a person. This was all a masquerade. Sydney was just a bot. But who hadn't like disclosed this to us. And now the company was not taking responsibility for the actions of their bot. And that was just very, like, really undermined our trust in the company to where we didn't even want to go back and see an apartment. Like, this might be the best apartment for us. We're just like so upset about this. So I think the same thing is happening just at scale right now, where if you think about it initially, I loved the fact that they were using AI to personalize. It was fast. It was personalized. It was super convenient. But I was betrayed by the fact that I didn't realize that it was AI. And I didn't realize, and then they weren't taking ownership. They weren't taking responsibility for the actions of, you know, this agent of theirs. And so I think that's like, you know, consumers kind of have, I feel like maybe 2 or 3 expectations. To your point, there's an expectation that companies will use the data that they get to interact in personalized and convenient ways. Like, I don't want an, uh, you know, an interaction where I have to like call or wait a long time. I do expect that instant, you know, personalized response because it's possible technologically and like you have the data. So like you should use it. But then also they expect that it's done with a level of transparency that builds trust to where you don't feel like you're being misled, right? Because no one wants to feel like you've been made a fool of, right? As we ended up feeling in that situation.

That's a really great example. It just makes me think of all the AI that's on all the social media. And that's not being disclosed and how companies are trying to mitigate it, right? TikTok has a policy, Meta has a supposed policy, YouTube has a policy, but we still are inundated with messages and things to respond to that aren't disclosing. So we do lose that brand trust just like you did with that building. Did you ever, did you end up using that company at all?

No.

See, exactly. They lost a customer and they lost any recommendations that you would give to potential other renters? Yeah.

And if only they had disclosed that Sydney was a bot, I'd be perfectly happy to, you know, book with a bot. I don't care. I just, you know, just don't want to be lied to.

Yeah. So is, would you say that's one of the things that you see is one of the mistakes that you see most often that costs these companies credibility is the lack of disclosure, or are there other things that you see as other, as issues?

Yeah, I actually think that's a relatively rare mistake. I think. That mistake obviously does, it does come up, but usually the mistake is actually on the other side where companies are too slow and too cautious and they don't actually use their data to provide a great customer experience where, you know, they have a bunch of data and they're just sitting on it. And, you know, it's, it's, it's a lot of work, right? To like do these things. And it can be complicated in large companies to get, to get the necessary alignment. And the more people you have involved in a decision, the more conservative you're going to be about how you make those decisions. But meanwhile, the world is changing and And consumers do have, you know, startups that they interact with and including the big labs. And so their expectations have been set in a, in a different place. Eventually brands, brands are all going to have to catch up if they want to survive. And then it's just going to be a matter of who, who does it quickly enough to make it and who doesn't.

And that is something we saw, particularly for not enterprise solutions, but for the SMBs, COVID, right? Brands didn't even have websites, small town companies, they didn't, even have a way for people to purchase if they weren't in the storefront and people couldn't go to these storefronts. Then once that was resolved, we have this thing called generative AI and we're seeing, oh, okay, now you small businesses really need to understand and they need to have infrastructure capability to utilize these tools. They need to have this digital literacy. And like you said, they're getting left behind. The next thing, of course, I think is this level of personalization. It is the level of, engagement, whether it's person to bot or, and making sure that the bot has all of the correct information for the brand or even bot to bot. I mean, one of the things I've been saying about agentic AI is I'm really excited to have an agent that's working for me. And I mean, I, that's how I use Claude, right? It puts things on my calendar and it does a whole bunch of great things for me. But I'm excited to be able to just have somebody call my dentist and talk to the dentist bot and schedule those appointments for me and my daughter, right? Or whatever the other case may be, because it knows enough about me. To be able to have those conversations on my behalf.

Have your people talk to their people. In this case, have your bots talk to their bots.

Exactly. And I mean, we've seen the funny clips of bots talking to bots, right? But there's not a heavy lift right now in terms of consumers who understand how to use that and have their own agents dealing with brands bot to bot. So can you talk a little bit about that shift? How do you think we're going to increase? Because I believe that your own company says they expect that to shift from about 19% use right now to about 46% by the end of this calendar year.

Yeah, I mean, agenda commerce in general, right, is a big, is a big, big area. It has many, many facets. Like there's, there's even something as simple as folks go into ChatGPT and ask it which products are better or worse, right? And so companies are needing to respond to this. And if before they, you know, you would, you would worry about SEO. Now you have to worry about AEO or GEO. You know, there's all this acronym soup for this stuff. And people have specialized tools for that. Then there's, you know, conversational AI obviously is another important thing where people or bots want to be able to come to a site and have a conversation. That's, for example, that's, for example, a capability that we're, that we're including in our product and it's, it's kind of in flight. And then There's also just making things accessible to different bots. Like in our case, it's true on the consumer side. It's also true even for, for our product at Braze. A lot of marketers want to be able to use, you know, bots for working with Braze. And so that's a large part of our roadmap is making sure that we can support that and make it convenient for our customers when they want to, through Claude, you know, do work in Braze. So I think, I think on all those things, basically Companies like ours that are software companies and brands are no longer just building for consumers. They're also building for bots. And I do think for a consumer, to your point, for a consumer to dispatch a bot to go make a phone call is still pretty rare. That's like very cutting edge, but it will become very normal. I'm sure in a matter of a few years for marketers to want to be able to use MCP in order to operate in a system like Braze, that's already here. today. For consumers asking ChatGPT which product is better, that's already like dominant today. So the different facets of this agentic commerce are all at different stages, but they're all, but it's moving very fast.

Okay. This is going to be a little out of, hopefully not completely out of left field, but I was just reading yesterday about Cloudflare and what they're doing with web traffic because they control about 20% of traffic and how they've decided that There's gonna have to be a new payment mechanism, kind of a paywall, so that bots won't be able to do double crawling. And I don't know if you know any, you know, if you've thought about that, what happens if people do start not allowing bots to crawl and not get information from a search engine and then also take it into an LLM search or another kind of agentic search? Is that going to affect the way marketers can do business? Do you see it having any implications? On what you do? Are there any workarounds that we can start thinking about?

It's a great question. Unfortunately, I don't know the answer. I think I want to learn more about it. I'm not sufficiently familiar with it, but it's an awesome question. I'm definitely going to look into it now.

I just had to put it out there because it is a very recent article, and I was talking about it in class last night, and I said, this isn't happening till September, and I can't give you any other information. But I thought since you know a little more than I do, I would just throw that out there. So, But thank you for entertaining it. Yeah.

I mean, for what it's worth, you know, this, this may be, this is just a hunch. This might be an uninformed hunch, but I think it's going to be tricky to completely shut it down because as bots get better and better, they probably will find ways to navigate where you don't know if it's a person or a bot. Right. A good example is these reCAPTCHAs, which have been getting progressively harder and harder. And now I can't, I don't know what impact they have on bots, but I can't solve them anymore. So I think, you know, it's, it's going to be probably some kind of arms race. And I don't know which way it's going to shake out, but I don't think it's going to be clean cut for what it's worth. That's my hunch.

But yeah, that's very fair. And that's why I think it's going to be really interesting to see somebody try to govern what we're doing, right? It's not being legislated. And can one company own that? So lots of questions out there. But going back to other things, thinking about Bot to bot and thinking about 35% of consumers thinking that a reason to use agents is to negotiate better deals. So basically it's a bot trying to optimize against another. So any thoughts on that?

Yeah, I think that's, that's another, it's a great example of another surface area that marketers are going to need to be very sophisticated with. Because, you know, there's the pre-bought version of this, right? Where customers could call into a call center. And some of them were serious about, if you don't reduce my rates, then I'll churn. And some were not, they were just calling because they knew that you could do that and get a rate reduction. And so in those cases already, the companies that were very sophisticated about this were setting up, you know, AI decisioning or similar systems. Like, for example, one of, one of our customers is a leading home security brand called Brinks Home Security. So they've been using. AI decisioning in order to, you know, make decisions on when someone's renewing a contract. What is the right rate to offer? And when can you make some adjustments? When can you not make adjustments? When do folks need free equipment in order to renew? When do they need a rate reduction? When is a rate lock, you know, the thing that'll get the job done? And so I think that what'll probably happen is for the most sophisticated companies, what they'll do is you need to have an LLM agent that's powering the conversation. But you're also going to want to have an AI decisioning engine that's making decisions on guardrails. And it may be saying, hey, you know, there may be data that this person's called in 5 times in the last year and like threatened, threatened to churn over the rate. And maybe that's a data point that suggests, you know, here's a cap and we're actually not going to do it for this person. Actually, we know that they had 2 customer service issues where they were like outages of our product. In the last 3 weeks and they've never called in before and therefore actually they're serious. And kind of passing that through a statistical algorithm that can correctly determine which of those signals are meaningful, output, you know, a decision on what to negotiate and how to negotiate it, maybe even what style to use in having the conversation, and then have an LLM agent that can execute that conversation with either the person or their bot. That's probably going to become the state of the art in a couple of years if you want to take seriously, you know, defending against these negotiation bots. Yeah.

Is there a difference generationally in what your customers, your customers who are the brands that are going to customers, you know, that they're seeing, whether it's the ages, the psychographics, in the way that they're dealing with the trust issue of Companies that are using a lot more AI?

Like generationally among brands' customers or among our customers, like the people at brands?

Well, actually, that's a good question. So I was thinking about the brands' customers, but your brands as customers is also a really interesting question. Yeah. Yeah. Yeah.

Yeah. Well, I think among the brands' customers, you know, I do think you have, well, as an entrepreneur, you can see, The Gen Z is the fastest adopter of AI. So it's definitely this kind of, maybe what you'd expect in terms of generations, right? That, that the, that the folks who are the newest to the world are also the ones who are the fastest.

The ones who've grown up with a phone in their hands.

That's right. That's right. That said, I think there's also, and I don't know, you know, this is maybe more anecdotal, but I'm sure there's data on this too. There's also a very deep concern with AI in Gen Z. And that's probably, it relates to what, you know, I brought up before.

Yeah.

Yeah. In many cases, those are the folks who are also the earliest in their careers. And they've, in many cases, they've worked very hard through school and college in order to prepare themselves to be very, very effective knowledge workers. And now knowledge work is in the state of flux with a lot of unpredictability. And that's probably especially frustrating for someone who's still at the start of their career. So I think we're seeing that that's both, you know, Gen Z is both on a percentage basis, the fastest adopter of AI, but also, I don't know for a fact if it's more than other generations, but anecdotally it seems so, that, you know, those are a lot of the folks who are the most anxious and the most concerned about what it means for their futures. And then, you know, at brands, I don't know. Yeah, I don't know that I've detected as many patterns. I think it's a different situation because it's a little bit less discretionary, right? Like if you're a consumer, it's somewhat of a choice if you use AI more, if you use less. But I think at this point, pretty much every marketer knows that it's just the job. So I haven't observed any generational differences, again, just anecdotally, but—

Yeah. I would say the one thing that I've seen shift, and as I get more students who are coming early in their careers or just out of undergrad versus a couple years ago, I think I had more students who were also older generations. They're coming to me with more knowledge on AI, but they're still very cautious and they're a little surprised that we have to underpin anything that we're talking about in digital social media or digital media management, no matter what play it is, right? Whether it's buying, you know, advertising and the analytics behind it and doing the A/B testing, or whether it's, are you, what kind of content are you creating and everything in between that they still have a little bit of that skepticism. They know that it's a necessary evil, but I'm trying to get them really excited about it and show them all the different use cases and share. When I have interviews like this, I share them with my classes as well to get them excited from somebody who's in industry right now. So thank you for that.

Yep. Yeah, of course.

And people don't always stay, right? Founders sometimes have to stay for a couple of years, but a lot of founders also exit. You're talking though about you integrated into something bigger. You're able to move faster to some of the goals that you had. for your company when you were integrated into Blaze. Is there anything in the companies that you've created and exited and created and not exited that you wish that people would know? Like an advice that you would give to your previous self or something that a founder needs to think about as they're going into these situations, maybe questions they need to be asking themselves or the company.

In terms of like potential, potential merger situations? Yeah, I mean, I think one thing that is really important is to make sure that there's a shared vision for what you're going to be building together with the company that acquires you. I think that's especially important when the product is different in certain ways, right? Like in our case, for example, it was very much this data science machine learning product. And at the time, Braze certainly had Braze, you know, had machine learning elements to the product, But the primary focus was, you know, SaaS. And so the business model was different in various ways. And in our case, I think one thing that was very helpful is before we even agreed to do the merger, we spent a lot of time getting to know the Braze leadership team and really having those conversations that are not conversations of, oh, you know, more that were very frank and open about the risks that we saw. Because you want to make sure you're on the same page philosophically. And you don't want to be in this kind of honeymoon period where like, oh yeah, these people are great and they think we're great. And like, let's assume everything's great. You actually want to say, okay, so merging these businesses is a very consequential decision for both companies. And I remember we had a meeting in Boston with a lot of the key people on the Braze leadership team, including, including, you know, Bill, Braze's CEO, and with me and my co-founder. And we did a pre-mortem. This is before we, on both sides, agreed to the deal. Hey, if we end up merging and it doesn't go well, like what could be the reasons? And we actually wrote all those things down and we went through them together and we said, you know, do we understand this risk the same way? Are we aligned? Do we think about the same or do we not? How would we mitigate this risk? And that's the sort of thing I'm, I'm, you know, I would very much recommend anybody who's, who's in that situation where they're, you know, their company might get acquired to the extent that they care about, you know, the future success of that company, which I think most founders, I believe, care very deeply to do that exercise. Because in our case, I'm very happy we did because we really were able to vet that like, okay, there is this philosophical alignment because every merger is going to have, you know, ups and downs, right? That's a normal part of merging 2 businesses. But at the end of the day, if you're truly on the same page, if there's deep understanding of the business on both sides, and if there's a shared vision that's not superficial, but that's actually based on like knowledge. Then it becomes a collaborative exercise and you work through it and you're, you know, you're going to have a great, a great long-term success of, of the product you've been building.

Fantastic. This has been such an engaging conversation and I know I've taken you all over the map. So thank you for allowing that. I'm just wondering if, is there one thing that you would want people to take away from our conversation today?

I think maybe it's, it's the, We're living in such a fascinating time right now. It's like, it's, it's anxiety inducing, but honestly, so have been so many other moments in history. Right. And it's because there's this, there's this like historic change that's happening with AI. And I think marketers are in a particularly, you know, interacting with brands is one of the most important parts of daily life for like everyone in the world. And so marketers are at one of the most important nexuses. for this historic shift that's happening. And it's, you know, it's scary, it's fascinating, and, and, and it carries huge opportunity all at the same time. And so I think part of what we should, we should all try to do is also like enjoy the moment, right? Like it's, it's such an exciting moment for marketers where marketers do have more and more extraordinarily powerful tools at their disposal to implement AI. and to make consumer journeys quite magical. Yes, there's also very high expectations that consumers have. They have expectations for doing that. They have expectations of doing that with a high level of trust and transparency. But it's like, it's kind of, it's an exciting challenge and it's a privilege. So to me, it's been like so cool to be a part of that process alongside all the marketers that I've been able to work with.

Amazing. Well, thank you so much. And for everybody who's listening, if you wanna go deeper, Braze's 2026 Global Customer Engagement Review is free to download at braze.com.

I'll link it in the show notes.

And I'll also link your LinkedIn, George, if that's okay.

For sure. Absolutely.

Yeah.

And thanks so much, Anika, for, for having me on the show.

Absolutely. And to everybody who's listening, this is my ask for you this week. Think about somewhere that you're trying to personalize something for your customers. Ask questions that George raised in our conversations, right? Think about, is this something Do I have the information? Do I have the data I need to understand what I can best personalize or what is missing and how can I get that information? And think about the ways that you as small business owners can also start implementing more agents, more AI, and more strategy into your businesses because as George said, you need to do that yesterday. I mean, you actually said today, But I'm saying you also need to do that yesterday because yesterday's even better because, uh, tomorrow it's going to be too late. So with that, I'll be back again with another amazing expert to share their story, their journey, and some advice for your business very soon. Thanks for listening or watching to this episode of Your Brand Amplified. Don't forget to leave us a rating or review on your favorite podcast listening platform. And if you want to learn more check us out at yourbrandamplified.com.