Braze, Inc. (BRZE) September 2026 Conference Transcript

Let's kick it off. There's a lot we want to get through here. For those in the room, I haven't had the chance to meet. My name is Billy Fitzsimmons. Started at Piper at the start of 2026. I cover the hyperscalers, application software, and vertical software. We are incredibly excited to have Bill Magnuson.

I assume we're both Williams as well. We share that, yeah. Didn't clear that up front. B2C brands, but brands of all kinds who are looking to better understand their first-party customer relationship in order to build stronger customer relationships over time. And so what that literally means is that we collect tens of trillions of first-party data points every between four and five trillion messages last year on behalf of brands, all of which are being orchestrated and personalized using first-party context and first-party data and advanced AI. And we're going to talk about a lot of that today. I think from the print, definitely an exciting quarter for us. We, you know, we beat, raised, growing 26% year over year. We had a record-free cash flow quarter in the quarter as well. Operating income guidance, we moved up our operating income margin percent on the year, up another 30 basis points, and, you know, pacing to a great full-year result there as well. You know, we had consistent bookings actually around the world for the first time in recent memory, which is great to see. We've been investing a lot in our go-to-market leadership organization since bringing on a new CRO last summer. And it's been great to see, you know, we're a little bit over a year from him joining and just got a strong sales leadership bench around the world, and we're seeing consistent results from that as well. So, you know, a great quarter all around, and we're really excited about the back half of the year. We've got our customer conference coming up in just about two weeks. It's called Forge out in Las Vegas. It's going to be bigger and better than ever. We're going to talk about some of the AI roadmap as well, and so I'll be able to give you a bit of a preview of what's coming at Forge, but I'll also probably drop more to come a few times in this chat.

And, yeah, I mean, it's been an exciting year, so excited to dig into it. There's been a lot of changes to the products. Can you also set the stage on some of the product innovation and how you've kind of enhanced the platform in this AI era?

Yeah, so high level, we're in a golden era, I think, right now. in Brace product. Our product delivery velocity is moving faster than ever. We're in an incredible position as we've accelerated our own AI roadmap in that Brace already has proven scale. We've got the reliability, security, stability, performance at massive scale. You know, I just mentioned some of those numbers before, and what that has meant for us over the last couple years is that a lot of the really difficult distributed systems and scaling problems and all that, you know, to be able to have smart pipes, to be able to, you know, drive a lot of this agentic orchestration acceleration through was already there for us. And so we've been able to actually all the way back in 2011, about 15 years ago. I took over the CEO seat about 10 years ago. And my technical co-founder, John Hyman, is still Braze's CTO today. And our SAP of engineering and our head of AI products have both been with the company more than 14 years. And so, you know, when I look at, I always chuckle a little bit when I read the founder mode articles and what have you because we still have both of our technical co-founders and our most kind of tenured product and engineering leadership that have been around since the beginning. And we went through and we were born in this last phase of disruption around the rise of mobile. And now as we look at this wave of disruption from the rise of modern authentic AI, we're doing so with a founder mode team. And so it's just been really exciting and awesome to be right at the frontier of change with us being able to leverage underlying science that's moving so rapidly, both in the LLM space as well as on the reinforcement learning side, and to be able to combine that with a system that we have laboriously and lovingly scaled over the course of the last 15 years. And so now, you know, when you look at Braze's ability to just deliver customer engagement through, like, smart self-optimizing pipes to our, you know, to our enterprise customer base, and then be able to kind of build on top of that for much more advanced orchestration and personalization than was even possible before, we're able to deliver that combined solution, and it's moving faster than ever.

Let's talk about some of the individual products, because there's a few different ones here at different stages of maturity. So Decisioning Studio, Agent Console.

I'm a believer that I think the future of application software requires that you have bespoke intelligence like Braze Operator, that you can deliver to customers that is the best user of your software possible, right? And our litmus test on that is that we need Braze Operator to be the – I want Braze Operator, and it already is in this case, and it's actually accelerating, but, like, we'll continue to measure it by, you know, what is the most state-of-the-art model from OpenAI or Cloud when they read our documentation and try to use our, like, dashboard product. I need Operator to be head and shoulders above that in capability because Braze Operator benefits from proprietary use case libraries, from advanced skills that we built from the vertical integration into the dashboard's, like, metadata environment, both from the metadata of the Braze, like, objects and the development environment itself, as well as the first-party context that the customer has access to and the way that they've integrated Braze. And what we've really benefited from, in particular this summer, is that we built Braze Canvas to be a proper programming environment. You've never heard me talk about it as, like, a workflow tool. It is a visual programming language, and Braze canvases, which are the primary unit that Braze customers kind of build and configure their Braze environment in, it has full features with a runtime environment behind it. It's got a full application context concept that is, like, mutable and visible to the It's got debugging and observability capability. It has, you know, multi-user, like, versioning and collaboration features. Like, it really has been built as a full development environment, very similar to, like, Cursor for marketers to be able to, you know, to be able to build out advanced marketing and customer engagement programs. And as a result, like, when we built the bridge from the underlying foundational models, which, of course, have all been enhanced and optimized for coding, into really understanding the Braze development environment, the acceleration and capability has been tremendous. And so what Operator has really done is unlocked all of the capability that was already there in Braze, but was being underutilized because a lot of our customer base just didn't have the time or the resources or the full skill sets to be able to really use all the advanced and comprehensive capability that's in Braze. And so, you know, I think that is the first major unlock, and that gets into the headless conversation and such, and we can talk about that more as well. But the critical starting point is that Operator is unlocking all of Braze for all of our customer community, and it's doing it rapidly. And that is, you know, the stories that we hear back from customers who have adopted Operator are incredibly glowing. I shared a stat actually during the earnings call that over the last 90 days, we've had almost 80% of our customer base has used operator more than 10 times. And more than half of those have then gone on to use it more than 100 times because it is just like it's addicting like that, right? Because of the value that it brings to you. And we're seeing customers who are completely changing the way that they run their own day to day. And it mirrors what we saw happen in the software engineering world, which is that, you know, as tools like Cloud Code and Codex and such, like, got better, and especially as they leap forward over the course of the last nine months, developers just don't write code by hand anymore, right? They still need a development environment in order to be able to inspect the entire system and be able to test and review and have the right observability and all those other things, which are, by the way, all these things I just mentioned that Canvas has, right? And you need a place to deploy all that, and the distributed runtime that automatically scales for you, which is effectively what lives behind Canvas, is the critical way that you take the programs that you build and you get them running in the cloud, being driven by first-party data, and doing so in a way that's secure and scalable and high-performance. And so we're delivering all of that effectively as a agentic harness for marketers to be able to build customer engagement programs and have operator build it for them. So that's operator. Then within each of these Canvas programs that I've been talking about, there are more and more advanced features that we're building. So before, your Canvas programs would have, like, you know, let's say you have a forgotten cart or abandoned cart campaign, and it's like, okay, the Canvas trigger is based off of someone viewing a product page, and then it starts a timer, and if the purchase is not made within the timer, then we, you know, we create, we send a message to try to motivate the sale, right? Simple forgotten cart concept, this exists across all verticals. Agent Console provides the ability to inject an LLM into that same flow. And so an example that we've seen customers have driven really great performance uplift on is when you are about to send an abandoned cart message, you have the LLM rewrite the entire content of it to be a custom compare and contrast of the last few. Let's say for a hotel brand, for instance, that we work with in APAC, they had the forgotten cart message send a custom compare and contrast of the properties, the last few properties that the person had looked at, along with great vacation ideas and other sorts of inspiration that was tied to the context that they have about the user as to business traveler, luxury, for pleasure, romantic, et cetera. And so by bringing in that first-party context about the user and about the recent user journey, they then rewrite the content so that every single email or SMS is completely one-to-one, and that is just injected into the flow of the program that they were running before. So that's an example of Agent Console getting used, and some of the really great use cases we've seen there are being able to drive personalization off of even sparse data points. And so feeding them with first-party context about the user is obviously where a lot of the differentiation comes from. And Braze manages all that for you and then invokes the agent call. But marketers are still experimenting with exactly what the great strengths and weaknesses of taking this agent or this LLM-based approach is. Now, one of the places to bridge into Decisioning Studio that agentic personalization is not particularly good at is in discipline around discounting. And if you think about your own interactions with these chat-based tools, like, they've been aligned to please us, right, and to stay in an engagement loop. They don't want to upset us. And so when something is the equivalent of, like, a price negotiation, they don't, like, hold the line and hold the tension in order to, you know, they just give up the discount right away. And they want to give you as much of a discount as possible because they're trying to please you. And so if you want to be able to do – but, like, that, you know, that is the opposite of a lot of brands' goals right now, right? You don't want to train your customers to expect discounting. and discounting hits the bottom line and even when you use it as a tactic you want to use it in a highly controlled way and a lot of times especially around subscription you also want to sometimes make the call to be raising prices on people when you get to you know an annual renewal or what have you and so those are all examples where we need a different science approach and the answer for it is constrained optimization using reinforcement learning where the model is actually able to experiment with different discount levels and different price changes across the customer base, across the thousands or millions of interactions that you're having with other customers over time, and then the model learns over time where the points of resonance are and where it can kind of take risks in order to optimize for something like decreasing discounting or increasing prices. And so that's where Decisioning Studio comes into play, is when you've got a reinforcement learning need where the model needs to achieve a more advanced goal than merely engagement and where there will be benefits from it learning from all the other interactions that are happening across all the different messages that get sent. When you think about the context window and there's a lot of conversation around making sure that agents have the right focus and first-party context and they can be non-deterministic if you don't have the right guardrails around them and what have you. And all of those considerations are at play when we're engineering Agent Console. Part of the answer to that is keeping the first-party context that you deliver to the agent tight, right? And so part of the job of Agent Console in Canvas is we need to deliver the user context and other first-party data into that window in an efficient way, so that when the agent runs, performance, like there's less non-determinism, etc., you're getting value out of that first-party data, but without the context window potentially distracting or slowing down or ballooning the costs of the agentic call that you're about to make, right? And so that need also means that it's harder for that agentic approach to learn from the millions of other interactions that might be going on in your customer base, right? But we already have an approach for that. It's also how, like, your TikTok feed personalizes itself over time, right, by looking at the mapping between different user interests and different content properties and being able to find the points of resonance between them, right? So Decisioning Studio runs in effectively the same way but adapted for these marketing use cases where you're finding these points of resonance along with the content personalization options and using that to drive higher performance, just like your TikTok feed would personalize over time to drive higher levels of engagement. And so those are the two major product families is the LLM-based approach and then the kind of heavier weight math-based approach. And they can actually be combined together. So you can, for instance, use decisioning to drive your discount optimization strategy if you're trying to constrain discounting to go back to that example before but then still have the LLM rewrite the content in a fully personalized way before it gets sent out and we have customers that are experimenting with combining those together as well and one of the great things about the Braze platform is that you can now string all this together where you go and you come into the Braze dashboard experience you chat with operator and you say hey operator and we we just shared an example of this on the earnings call actually from our city by city Sydney hackathon it was a coffee cafe chain of cafes in Australia and they are already using our survey tool in order to collect post visit surveys from people and what they wanted to do was actually update the survey response to automatically adapt itself based on the content that someone put into the survey feedback. And so in Braze, that is extremely straightforward now. They had a quick chat with the operator. They told it what they wanted to do. It went into their Canvas environment. It built a multi-step response Canvas. So it's like send the survey, process the response, and then depending on whether the customer sentiment is negative or positive, If it's negative, we want to send an SMS follow-up, and then the content of that SMS follow-up should be based off of what they complained about, basically, in the survey, right? And so the operator then went and built the agent that processed the survey response and built the personalized reply and then pulled that back in, configured it into the Canvas flow, and set it up to automatically send those SMS. And that was all done in under 15 minutes. And so you start from just this goal and operator navigates around, it programs it, it does it in front of your eyes so you can watch it and you can have certainty that it's building the right thing and you can edit it and change it and, you know, observe it and everything afterward. And then launches and has this much higher level of personalization across these channels using these strategies that were being run before but they're now being delivered at much higher quality.

I think it's not every day you get to sit down with a technical founder, CEO. And so I want to get your perspective on a couple different topics as it relates to AI. But I think the foremost one is agentic commerce and where there's agentic commerce. I think a lot of people in this room probably still have the announcements from Meta with Metamuse. And how do you think about the world evolving in that context?

Well, it's funny you asked this, on the eve of our, we just had our 20th earnings call ever, and we're coming up on our five-year anniversary of our IPO, and so I had the occasion of looking at my letter from them. And if you go read our S1 letter, in the middle I talk about three different generational changes that we see that are going to continue to provide tailwinds for our business. And the second one is that building a first-party relationship will be an increasing imperative for brands as they continue to be disintermediated by, you know, whether it's a agentic retail commerce experience or a bookings aggregator, you know, or an online travel agent in the travel and hospitality space or it's a delivery provider within the quick service restaurant space, right? These are all examples where someone comes in between you and your customer and their play is one where they get to keep all the first-party data and they get to build an ad platform that extracts your profits out of the customer relationship. And it is the case that, well, those are often important sources of new customers for brands. And, you know, certainly we are already seeing this with, like, ChatGPT referral traffic into e-com, the ChatGPT apps, and now with Metamuse and what have you. And also, like, include in that things like TikTok shop and other sorts of Instagram shopping. and there's like all these other places where our consumer behavior and our attention gets like aggregated into these big platforms, and then that drives like shopping decisions, right? And when that happens, the brands that can run the highest like profit margin businesses and be able to, you know, operate with like higher levels of strength as a brand are the ones that double down on building those first-party relationships with their customers, right? Like the new kind of way of shopping becomes a source of new customers, but you need to get that first-party data on them. You need to get the ability to communicate with them. You need to train that customer to not start their customer journey through someone else's platform, but to actually come directly to you, right? More often than not, if you achieve those goals, that customer is way more valuable to you. And so I think the most critical kind of takeaway for brands from the Metamuse moment is that in the future, and again, I think this is already the case, but it will continue to become increasingly the case, the bifurcation between brands of those that you know the name of the brand, you care about them, you trust them for something that is precious to you in your life or to bring you great experiences around things that you're passionate about. about, et cetera, right? These are like luxury brands. They're brands for your kids. They're brands for your favorite hobbies or what have you, right? The special moments in your life, the things where you know and care about that brand and you have a first party relationship with them. And then there's going to be the companies, and I wouldn't even call them brands, right? But the companies that deliver goods to you through your agent and commerce experience. And like, you know, if we go look at the Amazon marketplace as an example today, hey, there's already untold numbers of these that will sell you phone cases and chargers and all sorts of random stuff in the Amazon marketplace. And what has happened as Amazon has built the giant ads business in the marketplace is that luxury brands and a bunch of these other types of brands that I'm talking about, they've actually doubled down on their own first-party ecosystem more to make sure that they're able to keep the first-party data from the interactions they have with customers and to make sure that they can build a customer relationship where they get to keep their margin, right? And they don't have to pay it all to whether it's Amazon Marketplace ads or it's MetaMuse's payment processing or Apple's in-app purchase rake that they have or what have you. And we actually already saw this in the QSR space with Braze specifically because the same thing effectively happened with all the delivery platforms. And when you go look at Braze's customer base today, okay, we dominate the major quick service restaurant brands around the world. And the reason for that is not because they all woke up one day and decided that they were going to be super sophisticated customer engagement like marketers. It's because they're responding to the disintermediation that happened to them because of these delivery apps, right? They started out originally like, okay, great. This gets me marginal demand. And then they wake up six months later and they're like, well, shit. Like, my most loyal customers have retrained themselves to buy through, like, DoorDash now, and that means that my loyal customers, I'm getting no data on them, all my margin is going to the delivery platform, and I've effectively, like, lost control of demand, And so then you look at, like, how do they respond? Well, McDonald's, you know, as an example, responds in the exact same way that, like, Marriott and Delta did 10 years ago, dealing with the same problem with the online travel agents, where they go in, they redesign all their stores, in-store kiosk experience, They're reading license plates in the drive-thru. They're personalizing the menus. They revamp the mobile app in order to provide enhanced takeout options and a stronger mobile loyalty program and mobile wallet. And they essentially rebuild, like, their entire digital ecosystem around getting primacy of that first-party connection with the customer so they can get the first-party data back, so they can get the ability to communicate with the customer back, and they can get their profit margin back, right? And the airline and hotel industries had to do the same thing when online travel agents came into the picture. And luxury brands have had, you know, luxury brands and other brands, as they've navigated this modern, like, retail and commerce environment, have had to do the same thing, too. And I think that, you know, this kind of, like, meta-muse-agentic shopping moment is actually just, like, it's retail's turn to go through the same thing that, you know, travel and hospitality and restaurants, like, already did. And when you look at the impact on Braze in those categories, It's great for us because when brands have a stronger incentive to invest in their first party relationships with their customers, that's when they become great brave customers as well. And so I do think you're going to see some, you know, brands kind of die as they are unable to compete as these like faceless commodities downstream of, you know, agentic shopping experiences, much like I'm sure is already happening in the Amazon marketplace, right? But the brands that thrive and the ones that are great Braze customers are the ones that are going to double down on knowing and, you know, continuing to build that direct relationship with their customers. And so, you know, also probably a great time to remind everyone how diversified the Braze customer base is. Retail and consumer goods is actually only 22% of our business because this problem of building first-party relationships with your customers is one that is broad-based across every vertical. And I think that, like, that's why, like, we have the experience of having seen this same effect play out over time, and we've got a strong track record of it really benefiting us. And I can, you know, point to all the other categories that have had this, like, aggregator disintermediation in the past. And so, yeah, it's obviously a big moment, and I think that more and more customer behavior, much like has happened with e-com over the past 25 years, will shift to buying in these ways. But it actually, I think, just doubles down the imperative for brands that want to survive to make sure that they're investing heavily in their first-party ecosystem.

That's a great answer. We'll wrap there. Bill, thanks for joining us and really excited for Forge in just a couple weeks. Big announcements there. So thank you.

Yeah, thanks for having me.