Conference Transcript - TWLO 2026-09-08

Speaker 1

All right, let's go ahead and get started. Thank you so much, Aiden and Inbal, for making it. And, yeah, let's go ahead. So I'll start with just a broader question, and then we'll kind of dive into some specific questions for each of you. But Twilio has historically helped customers connect communications across a fragmented carrier ecosystem. As you increasingly connect context across different channels, applications, AI agents, where do you feel like Twilio can create the most value? Is that delivering on the communication, managing a conversation, or kind of increasingly helping customers act on those things?

Speaker 2

I think it's all of those things. I think our bread and butter is the channels, right, connecting our customers to, you know, the end consumer. But we're increasingly trying to make those communications more valuable. I'll let Niball talk about it, or I'm sure you'll get to it. But we've launched a suite of conversations products from memory to orchestration into intelligence in the last six months or so, and that's where we kind of see the future going in terms of agentic as well as human communication.

Speaker 3

And I think the uniqueness for Twilio is, if you think about Twilio as three layers, we have all the communication channels where the conversation is happening, and then you have the contextual layer, which is the data, and then you make the AI agents that are operating across all these channels with the context layer so much better, more effective, more productive, more accurate, which is what you're trying to achieve. So it kind of operates, as Aiden said, across all of them.

Speaker 1

Yeah, and then Aiden, as the product has broadened, how has Twilio been able to balance kind of investing ahead of these new product cycles while maintaining operating discipline, particularly given that newer products can have a wide range of different financial profiles?

Speaker 2

Yeah, I mean, I think we've actually done that pretty well. So like as growth has kind of reaccelerated, if you think about 24, 25 into 26, like we've been quite disciplined on cost. I mean, we were flat to down 24 and 25, up a little bit this year as we launched some new products. But I think we've kind of run the company differently, right? Like I think with Kozema, who's our CEO, since both of us kind of took our roles, we really introduced a lot of financial discipline and operating rigor into how we run the company. I'd say how that manifests itself in innovation and investments on the R&D side. Like, I'd say, you know, if you look back over, you know, years ago, a couple years ago, I'd say we just tried to invest in too many different things, right? Like, we tried to place a lot of bets in different areas. We're much more focused now. We're much more focused on platform efficiency to help us drive kind of innovation velocity. Imbal's actually been the person driving that for all of us, and we're very ROI-focused. You know, I'm fortunate in that I have a CEO that has a very strong financial background, right? So he's very ROI-focused as well. I think a great example of that in terms of an area where we've invested recently is our self-service platform. Twilio has always been developer first, PLG. The reality of it is we let that process get too complicated for our customers over time. And so over the last couple of years, we've really undertaken an effort in product and engineering and go-to-market to simplify that experience. And most recently, we launched what we call OneControl, which we can get into, which makes it even easier for customers to come in and adopt multiple products on Twilio. So that's one example where the return was just obvious, right? And we put a lot of investment beyond it. Maybe just on your last point around different financial profiles, like what I would say is from a pricing model perspective, all of our products are usage-based, you know, for the most part. So in that sense, they're very similar. Where I'd say they're actually different is in the gross margin profile. So, you know, aside from messaging, most of our products are quite high margin. I think this new Conversation Flayer that we just launched, I put in more of the high margin bucket.

Speaker 1

And then Imbal, one of Twilio's historical strengths has been giving developers these, like, module building blocks. At the same time, customers increasingly want Twilio to solve more complexity for them. So how do you balance kind of remaining flexible and being developer first with moving higher in the stack and owning more of that layer?

Speaker 3

Yeah, so I think that the first thing that we should anchor on is that the concept of developer is changing. It used to be that developers, we were able to segmentize them as like it's a software developer. They're here to solve a complex problem. You expect them to solve the end-to-end, but the cost of building is going down. And as part of that, what we're seeing from the different enterprises and the different ISVs and businesses is they want to be able to control full customization. So also remember we're coming from kind of the era of SaaS, that businesses got like a black box and they needed to customize it to a specific level. We see that what our customers are asking is to get more flexibility. They want to be able to create these use cases that really solve their problems. They want to be able to build some of these bespoke solutions, and because the cost is going down, then suddenly the developer is getting a different context. And why does that matter? Because Trilio serves both. So we're serving the existing developers, which is the software developer that won all these bits and pieces, but we're also introducing new interfaces that enable the new builders to be able to build the solutions they want on top of Twitter in the fullness of time. Like, conversation memory is a good example for that because when you're looking into how are we making conversation better, conversation, in order to last a lifetime, it needs to have some sort of a memory. So a true customer 60, a true journey that is happening throughout the lifetime. and you can customize it and build it by yourself or you can use a Twilio component of that. You can introduce it to your conversation and you can also build a brand new business like an AI native that is building AI agents that are using our conversation memory on that.

Speaker 1

And you've delivered 5% plus organic revenue beats over the last two quarters. It's obviously kind of like been very well received by investors recently, but what drove that specific upside and how should we think about that level of performance persisting over the next couple of quarters?

Speaker 2

Yeah, I mean, it was pretty broad-based, which is kind of what we've said. So when you look at it across product, when you look at it across sales channel or industry vertical that our customers play in, it's actually quite broad-based. Now, messaging is 60% of our revenue, right? So that tends to be the biggest driver of variability in any period. And when you look at that business, it grew like 18% in the first half of this year. It's a large part of why we outperformed performed relative to our guide. But with that said, like, we're usage-based, which is why we tend to, you know, guide and plan a bit more prudently. And if you look back maybe a little bit further, the last couple years, we've really beaten in the range of our guide in the range of like 2% to 4% on the top line. I'd say that's more the norm, right? 5% plus hasn't, you know, nothing has changed in our guidance philosophy in that sense. So I'd say we're not expecting five percent plus to be the new normal now when you look at gross profit and you look at our other products like voice or software add-ons those are all very high uh gross margin or for every dollar revenue they carry most more gross profit than than messaging does and so when you think about like the what drives the gross profit uh strength it's really much more balanced uh across the portfolio where i'd say at any given period messaging can can drive the revenue uh one way Voice AI is something that's been top of mind, I'd say, for everyone recently, but it's something that can work really well in a controlled setting, but when you introduce it into production can face challenges with latency,

Speaker 1

interruptions, accents, background noises, network variability. So how do you think about the relative importance of those issues and how to help solve those?

Speaker 3

So I think we all need to remember these are very early days in the voice AI journey. We're just at the beginning. So the models are still evolving, the infrastructure is still evolving, how we build AI agents that are voice first is still evolving. This is a problem that the industry has been trying to solve for years, but with the introduction of LLMs and AI agents, we're now in a point in time that we can solve it, but the infrastructure has not yet kind of built towards that world. It was built for human engagement. Now, where it starts playing a role is where we see adoption and we see enterprises and different businesses taking these AI agents into production. There are not many of them. And if they do that, they do this on like a low-hanging fruit use case that they feel confident delegating to AI agents, or they do that in a specific or a small amount of customers that they are taking through that flow. Now, what are the barriers to take some of these into production? Accuracy is number one. And accuracy, it's a combination of the infrastructure, It's a combination of the ability of the model to really interpret what's happening in the conversation, and all of them together are playing a role. Now, where accuracy is critical is in the infrastructure layer when we're talking about latency, when we're talking about quality, when we're talking about all the ability to do proper turn detection, and this is where Twilio is playing a significant role. The second part, the trust. So trust today is a big blocker from taking these AI workloads into production because suddenly you have an entity that is there to complete a task. It is very task-driven. It doesn't have judgment. It's there to complete the task. So how are you enabling that AI agent to do that work unsupervised? So we're missing a lot of that layer of supervision. If it's the identity verification or it's the government or it's the kill switch, the way that Cole likes to call it, which is really about how are we monitoring what's happening in the conversation. I think the third part is, like, we need to understand that the regulation will evolve and they will change. So there will need to be new policies that are introduced. And as part of that, it goes to data storage, what is happening in the conversation, and how we monitor these conversations, where is the data stored, what data are you allowed to store and not allowed to store. And the way we're thinking about it is really across all these facets. And the other part is thinking about a conversation. And a conversation rarely happens on a single channel. it tends to be, by nature, a multi-channel engagement. You start on a voice, then you work to messaging. As I said, it's early days, so we see mainly the voice use case, but all the messaging and email is coming. And Twilio is uniquely positioned to be able to host, through our orchestration, a multi-channel conversation.

Speaker 1

And then you support multiple speech and model providers today. How do you think about eventually becoming more of a layer that helps customers select amongst these providers in real time based on maybe quality, latency, cost, versus kind of continuing to put that in the court of maybe other platforms?

Speaker 3

So we already are doing some of that with our Conversation Relay products. Our Conversation Relay products enable customers to pick and choose and bring the models, if it's the STT or TTS model or their LLMs, to be able to customize and create their own AI agents. And the premise of Twilio and why we're very adamant is about being a neutral platform. We do not think there's going to be only one. We see customers choosing different models, different AI agents, different use cases of implementation of AI agents based on their specific teams. They don't want one to rule them all. They believe that there is going to be a multi-agent type of a world. And we're giving customers the flexibility to bring these agents into production by building on top of the conversation memory that basically keeps the contextual on a conversation, even if there are multiple agents that are involved, or being able to have a conversation across multiple channels, even if it starts on voice and then goes to messaging or throughout a lifetime of a customer, which is maybe you start with a marketing engagement and then it goes to a sales and then there is a support issue and then it goes back to kind of a marketing opportunity. So really connecting all these together and giving the customers the flexibility to pick and choose the workload they want to bring to Twilio, we're giving them the rails to run on top of.

Speaker 1

In voice, AI can benefit Twilio. It feels like in a lot of different ways. Maybe it's more minutes at existing customers, greater software attachment, or just overall growth in the number of customers that are leveraging Twilio, which one of those are you think having the biggest impact on Twilio's gross profit growth right now?

Speaker 2

Yeah, I'd say the two primary drivers of the gross profit growth or the revenue growth in voice is increased minutes, so just more volume on the channel, and then the software add-ons. And when we look at Q2, the voice product grew 20% plus. And when you look at the year-over-year growth in dollars, 50% of it came from the channels and 50% of it came from the software add-ons. So pretty evenly distributed between those two things. You know, we used an example. The AI natives, you kind of referenced as well. Some of that is driven by the AI natives coming in to our self-serve channel, kind of building on our platform, still a relatively smaller part of the business. But what we see, as an example, we have a horizontal agentic builder customer. You know, when we look back about a year and a half ago, they came onto Twilio's platform. They were spending kind of low hundreds of thousands of dollars per quarter. All of it was voice connectivity. Last quarter, that's a $1.5 million revenue per quarter customer. A third of it is actually the software add-ons. Two-thirds of it is the connectivity. So as they grow, we scale with them. and they're starting to add these software add-ons, whether that's conferencing or media streams or answering machine detection, things like that, that make their product and their agents smarter and more valuable.

Speaker 1

It feels like a lot of people are increasingly using voice as kind of a channel to communicate with these LLMs. How do you think about that as a growth driver going forward?

Speaker 2

Yeah, I think that today I'd say not a big driver. We have had a little bit of it here and there. But to the extent that they're using a channel, whether that's voice or in some cases it might be like an over-the-top channel where they're using WhatsApp voice or whatever it might be, they can definitely leverage Twilio's Rails. Where we see some of the model companies today tends to be a bit more of a traditional use case, which is verification, authentication, coming to our platform. As customers come to their platform, they need to verify who they are. and so they'll leverage Twilio to do that, whether that's through 2FA or something like that.

Speaker 1

Yeah, and then we touched on this a little bit when we talked about the outperformance on revenue over the past couple of quarters, but when we look at it from a gross profit perspective, how do you think about the different drivers of gross profit being kind of above your expectations over the last few quarters, and how do you think about that momentum continuing to the back half of the year?

Speaker 2

Yeah, I'd say the biggest driver is obviously revenue, right? So, yeah, revenue outperformed, gross profit followed, But when I think about the fact that in Q2, gross profits grew faster than revenue, I'd say there was two primary drivers. One is product mix. So voice, the software add-ons, our self-serve channel, those are all high margin, right? So as they grow faster, and in particular as they grow faster than messaging, we have what we call favorable mix, right? So that's helpful. They just carry more, you know, for every dollar revenue, they have more gross profit. The other is cost actions that we're taking. So there's a lot that we're doing to actually reduce our COG, our cost of goods sold. So that's things like establishing more direct connections with carriers around the world so there's no middleman or aggregator in between. We've taken a lot of initiatives between my team and Imbal's team on our hosting costs. We've also shifted some of our products that were still on-prem to the cloud, and we're seeing some benefits from that. So we're trying to come at it from a couple different angles. Growing gross profit dollars, you know, sustainably has been a big focus for us as a company, and I think you're kind of seeing that play out in the numbers.

Speaker 1

Yeah, and then another thing you brought up when we were talking about the revenue performance, but messaging, I think, has been a big contributor to the first half of the year, and I think that's where people have a little bit less understanding of exactly what's been driving the relative strength there. So we'd love to get a little bit deeper into what the most important drivers of that inflection have been and how to think of even just the mix between maybe new customers adopting messaging versus increasing volumes and those different things?

Speaker 2

Yeah, so messaging is about 60% of our revenue. So it's a big business, right? It grew 18%. So in any given period, just given the size of the install base or the existing customer base, volume with our existing customers is the biggest driver. So new customers are growing quite well, actually very well, but it's just off a smaller base. So now why is our existing customer volume growing? It's actually not one thing, which I know is like a very unsatisfying answer. It's a lot of things. When we look at it across industry verticals, like our top five industries where our customers play are growing really well. When we look at it across sales channel, whether it's self-serve or enterprises or ISDs, again, all growing quite well. So it's not one thing. So some of it's macro. It could be, like, our competitive positioning. It's a number of different things that are driving it.

Speaker 1

And then when you think about the carrier fee side of things, obviously that's just kind of a pass through at the end of the day, but has it changed how you're seeing customers evaluate the ROI of messaging as a channel versus other channels, or does their behavior kind of continue to be very similar despite that?

Speaker 2

Yeah, not yet. I mean, I think you see that in the growth rate. 18% is ex-carrier fees, right? It's 28% with the new carrier fees in place. But we shouldn't look at it that way. The operational number is 18%. so no like have we seen a shift in demand because of the increased u.s carrier fees we haven't yet but you know customers talk a lot about it and they certainly don't like it right it's higher pricing to them we've been very transparent with them around the fact that these are coming like we try to message it as early as possible we also are very clear about the fact that if you know for a small or mid-sized business that you know may be under more financial pressure or something like that, like there are other channels available to them. Like we support WhatsApp, you can go to email if it's just, you know, need another form of written communication or other OTT channels. So to date, like I said, no real change in demand as a result of it, but certainly not something our customers love.

Speaker 1

Makes sense. You recently made conversation orchestrator, conversation memory, and conversational intelligence generally available. What did you learn from the beta customers that kind of most influenced how those products were designed and then any kind of other early comments you have on?

Speaker 3

Yeah, so I think beta customers is such an underrated asset that a company has because like what we've seen when we started going after the new conversations, we have a thesis that conversational AI, especially the voice one, is what customers want to solve right now. And some of our beta customers really help us figure out what exactly are these set of capabilities that we need to build into the product and prioritize. You know, every R&D leader has above the line and below the line. So for us, really figuring out what are some of the features that we have deprioritized with the assumption we have time to build them, our beta customers help us understand that we need to solve it right now. Like a warm handoff is a good example for that. How are you handing off a conversation between an AI agent and a human? Like how to detect an escalation happening was a top priority feature that our customers indicated. That's very critical for them, so we prioritize that. The second thing that we learn from our customers, and that's true for every infrastructure or platform company, is you're building something with a specific set of use cases in mind. And then we're starting to see customers adopting. Sometimes they will come up with new news cases that you never thought about. And for us, we prioritize conversational AI through the lens of support, but some of our beta customers started using that for sales use case. So identifying leads, especially when the store is closed and how to do a warm handoff to a sales agent once the store is opening and kind of understanding insights from that conversation, What's the likelihood of buying? So that enabled us to do kind of another set of thinking through what are the next set of features we need to build into our platform to unlock these new set of use cases that we not necessarily prioritize when we went into the beta. And I think that the last thing is that it helped us refine our go-to market from a pricing perspective. For example, we had the chance to survey some of our customers and kind of test different pricing models with them. Like what are you likely to want to be able to see? What will you be willing to pay? and how to think about it, what are the areas that will keep you up at night if suddenly you see them on the bill, but also help us refine our go-to-market motion because we had a chance to test a lot of these use cases and a lot of these sales pitch on our existing customers in a safe environment where they can give us an unsolicited feedback and help us refine how are we thinking about taking this to the market.

Speaker 1

I do want to give the audience a chance to ask questions, so I will circle back on that in a second, but start to think through any questions you might have. I want to double-click on the conversation memory piece of this. How do you determine what information should be remembered, where that information should live, and how it should be governed kind of across Twilio versus maybe some of the other tools in the customer stack like the CRM?

Speaker 3

So everything, the way we are kind of dividing it right now is that the conversation memory is everything that needs to make the current conversation better. So we're refining the transcript as we go. We're giving insights and indication on the conversation as it goes. so the agent will be able to complete their task in a better way. And then we're kind of taking that transcript and we're storing it in our longer-term memory with the idea that eventually customers will be able to query that memory. They will be able to train their agents to be better in the fullness of time. But we're not asking customers to kind of copy any of the data that exists in their CRM or exists in their data warehouse. We're kind of building on top of that where we have connectors to all these data storage, and we're trying to keep the most relevant information to be able to keep the human or the AI agent more successful in handling that specific conversation and that kind of the delineation that we have between the long-term data storage that are not real-time and what is needed for a real-time conversation.

Speaker 1

Any questions from the audience quickly? I can keep going. I want to kind of touch on what you mentioned earlier, which is the self-serve motion and the momentum that you've seen as a result of the new console. Can you just walk through maybe some of the things that you've seen over the past quarter and some early signs of success with that.

Speaker 2

Yeah, I mean, maybe better for Imbol since she did all the work. I can start, though. Not all the work. Well, most of the work. I certainly didn't do any of it. So really, it's what we call one console. We launched it in May at our Signal conference and basically allows customers to come in, developers to come in to one place, access all of our products. I know it sounds novel, but they couldn't do that before. Get things like, you know, one bill. or, you know, that makes it much simpler for them. And we've leveraged AI to make the experience much better, to help them along the journey. So when a customer comes in and wants to, say, adopt messaging, there's actually hurdles they have to go through to do that. So they have to register and all that, but they actually have to provide what's their campaign, what is their use case. And carriers want to make sure that it's a legitimate use case, not spam, essentially. And so we help them through that process, and AI is there to kind of tee up, like, hey, as they're filling out the forms, as they're doing all this, hey, you know, be careful, that looks like you've got it wrong, or maybe think about this instead. And it also tees up new products based on the use case. So we've tried to make this as easy for our customers as possible and AI is a big part of that. And as a result, what we've seen is conversion rates for our customers in that platform is up like 90% relative to our old platform. So great success. It's only been like three or four months or whatever it is, so it's still really early days.

Speaker 3

Yeah, I think basically what we have done is we kind of gave the cognitive load of figuring out how to work with Twilio to our customers. But the first thing we did is take it back. It's like our customers don't need to be an expert in our product. They don't need to get their way into using Twilio. So by bringing all the products into a single console so it's available for all our customers to see in one place, they don't need to log into several different consoles just to see what they're doing with Twilio. The second thing is removing some of that friction of experience experiencing, experimenting with our product, so we created kind of a playground. You can think about every developer coming to the console. They now get a credit. They can try it out. They can upgrade very fast. They see everything that is happening, and the third bit of that is really giving them insights, if it's on our billing or how to use the product or how to start a new campaign or how to onboard into a specific new product, so removing a lot of that friction on how to use Twidio, and then in addition to that, guiding them through the process with the AI agent was the biggest differentiator that we've seen from customer satisfaction starting with the new console.

Speaker 1

How do you think about maybe other parts of the product portfolio where you can maybe achieve similar results, like identifying the next self-server new console unlock?

Speaker 3

So for us, it's a lot of working backwards from the customer. Everyone knows I came originally from AWS at some point in my life. So there is a working backwards. How do you identify what are these customers' problems that we're trying to solve? How do we look into these signals of where the market is heading? What is the next set of problems that customers need to solve? And then looking into our portfolio and assessing, do we have a solution like that, but it is not used to solve that specific problem, or are we missing some components? Are we missing the ability for customers to connect the dots? I talked previously about some of the adoption blockers in terms of taking AI agents into production. Some of that is trust. What does that mean, trust? Trust is compliant, so how are you filing for all this information that needs to be shared? The second thing is how do you validate that whoever is engaging in that conversation is a legit player or a consumer or AI agent? The third part is that something is happening in that conversation the way it should be. So we're combining all of that together. It's like what makes a conversation trusted is a big opportunity to unlock, and that's an area we're investing in.

Speaker 1

So you mentioned this before. You allow customers to choose their models, what cloud platform they use, data warehouses, other business applications. How do you think about the strategy of remaining open and neutral versus still owning enough of the architecture to ensure that you have a durable differentiation going forward?

Speaker 3

I think the biggest differentiation is threefold. First one is being the largest telco in the world. It is what it is. We are the largest telco in the world. We connect so many customers, so many carriers into a single network that can serve customers globally worldwide. we can terminate a message in almost every country that exists today we have direct connect with some of the carriers in addition to that we have voice connectivity in addition to that we have an email solution we have an over-the-top channel so if you think about like the largest network of communication is a strong mode that Twilio has and it's not just the connectivity it's the ability to operate in a highly regulated industry that the compliance policies are changing over time again and again and being able to catch up and meet that compliance. So that's the basic Twilio premise. On top of that, it's the contextual data. The contextual data is not just a data storage. It's not something that happens after the conversation is done. It's what's happening in that moment. How are you making the conversation better in that moment? But also, how are you making that conversation better in the fullness of time? So some of that is really focusing on creating that contextual layer that solves the problem in the moment, But the other part is refining through transcripts and training data and connecting to a knowledge base that the company has to maybe improve their documentation because something the customer has been reading is not really clear. And maybe some of that is how you train your human agents to operate in a better way because we've seen that engagement not operating. So when you think about, like, that glue layer, the one that sticks the conversation together from the communication to the ability to have a contextual information and then make AI agent better and human agent better or any engagement better, it doesn't matter if it's schedule an appointment, that's kind of a unique mode that Twilio has, and it is the right choice for us to be AI agent agnostic and a model agnostic because we don't know if it's going to be only one. We don't think there's going to be only one. There's going to be multiple agents. There's going to be multiple models. We see some of these transform, like the frontier model to open source model. And we want to cater for the customers wherever they are, what is the use case they are trying to solve, which model they are trying to use. So that is becoming kind of, I would say, the commodity layer of the AI world, while the infrastructure itself is the sticky part.

Speaker 1

And given the breadth of your customer base, the breadth of products you have, the number of new products that you've come out to market with, Like, how do you think about directing incremental investment into each of those areas and kind of balancing different priorities, whether it's enterprise versus SMB or kind of other ways of splitting the business?

Speaker 2

Yeah, do you want to talk about how you have to kind of run a team on air, food, water, horizon one, horizon two?

Speaker 3

So one kind of mechanism we've introduced in the past three years is annual planning. This is something that historically we have not been doing as part of the R&D organization. And the idea is to look into everything that we have on the backlog, if it's keeping our stack alive or making sure that we're improving the core business to kind of focusing on core excellence. So how are we making channels and data better? And then innovation. So what are some of these innovations that are horizon one, horizon two? And then what are these experimentation we run around on horizon three? So we see some signals. It's not yet kind of fully mature. We don't necessarily have like a clear customer demand. So we've introduced the practice of taking all the backlog once a year, looking into what is our budget kind of foundations, what are we working with, what is the headcount allocated, what is the cost of our infrastructure and making sure that we're kind of prioritizing across each one of these buckets our headcount allocation and sometimes it's more towards getting rid of tech debt in specific channels because we see for example voice AI taking off, there is some tech debt we need to pay there to make voice quality better or investing in a new set of products because our customers are signaling to us that they really need new solutions So really being very much focused on solving into these big buckets and kind of prioritizing our investment based on what will get the biggest ROI for the company.

Speaker 1

There's a lot of debate right now in the software ecosystem of what the impact AI is going to have on application software. So curious, from your perspective, it feels like you're in a position to kind of benefit no matter who ultimately wins at that layer. But how does that kind of dictate how you think about where to invest in the product and how to think about what are the key priorities for Tulio going forward?

Speaker 2

Yeah, I think we're, like, when we think about AI and the impact on Twilio, so Imbal's talked a lot about what our competitive differentiators are, like, where we play in the infrastructure layer, but we're seeing voice as the place where it's all starting, right? As companies build agentic solutions, voice is the most natural place for it to start. It tends to be oriented toward more customer support. Again, voice makes sense there. So I think voice is definitely seeing a benefit. it. But ultimately, as we talked about earlier, we think it goes multi-channel. That's two-way messaging, email, et cetera. Communicationals have become both synchronous and asynchronous. In terms of AI in the company and how we think about it, we're certainly leveraging all the AI tools. SelfServe's a great example where we've gotten a lot of ROI by leaning into AI. I think our global operations or global support, our customer support is another area where we've leveraged AI heavily. And then with an Imbal team, like with our very technical folks, all the coding tools and everything that we have. So definitely leveraging it a lot. Very still, I would say, focused on getting operating leverage. I would say we were not one of those companies that was ever token maxing. It just goes against our financial discipline and operating discipline culture. But we're leveraging it broadly. We're just trying to do it in the right way.

Speaker 1

Thanks. Well, thank you so much. Everyone joined me in thanking Imbal and Aiden for their time.

Speaker 2

Thank you. Thank you. Thanks for having us.