Twilio Inc. (TWLO) Earnings Call Transcript & Summary

Well, good morning, everybody. Thanks for once again attending the Goldman Sachs Communacopia and Technology Conference. This, I think, is the fourth year in a row that we've had it at this venue. Tremendous excitement. Thank you once again for your support. And I'm very, very delighted to have Khozema, the CEO of Twilio, almost kick us off. This is at least the first presentation we're going to be moderating. So you're the first. Welcome to the conference. Thanks for your continued support.

Thanks for having me.

Wonderful. And I have my colleague Matt Martino, who is also going to be joining me in this amazing fireside. He said, "Kash, I'm going to ask way better questions than you. I know that you've been doing this longer than me, but I'm going to be better than you." So we'll see.

Khozema, as you take a step back, first of all, congratulations on the significant and impressive turnaround of the company under your leadership as CEO. So having come this far. How do you see the next 4, 5 years ahead for the company?

Yes. I mean, I guess the way that we think about the company kind of writ large is that our job is to ensure that every digital interaction between one of our customers and ultimately their consumers is nothing short of amazing. And so until we deliver on that promise, like we got a lot of work to do. And I think from our perspective, like what's exciting about the way that we're running the company, kind of speaking to turnaround, which again, is very much a work in progress, like we're running the company better, more financial discipline, more operating rigor. The way in which we're innovating, it's a lot more focused, I think, now than perhaps it was before. And so we've got this awesome collection of assets, communications plus data plus AI. And I think, as I look out 5 years, the way that I really see it is, is that we want to be able to deliver on that vision to be sure. But more importantly, I think we want to be indispensable to a company by being this customer experience layer of the Internet, if you will, which has to require communications plus especially contextual data that then gets activated by AI and LLMs and stuff like that. So we're proud of kind of what's transpired. We're looking forward always, and we're kind of nowhere near to where we want to be.

Got it. I think Matt and I had a chance to attend your Analyst Day, and I think Matt subsequently attended your user conference, SIGNAL, both were amazing events. I think at SIGNAL 2025, there was the idea about customer engagement life cycle. Can you tell us more about what goes behind it? And are there some aspects of the customer engagement life cycle manifest through a set of products that have a lot of potential for the company in the years ahead?

Yes. So I think a couple of things. I mean, first of all, I'd sort of start where I ended in the prior question, like the idea is how do we deliver on this notion of being the customer experience layer. And again, if you're a customer in whatever the industry is, I mean, it could be retail, it could be financial services, it could be health care, in every single interaction that you have between one of our customers and one of their consumers, you have to have some sort of communications channel to be able to do that. Even if it's a person, a live agent, which I'm sure we'll talk about later, you have to have some sort of communication that's happening back and forth. We have all of the channels that are required to be able to do that, and so that's an exciting starting point. I think the second thing is, is that as exciting as AI is, as incredible as the LLMs are, I think where things get really interesting is where you're able to apply context to the communication so that you're able to drive both persistence and this notion of customer memory so that over long periods of time, our customers can develop longer-term relationships with their consumers and be able to feed them increasingly more value, more intimate interactions over time. And so that's what we're really, really focused on. Of course, we developed some of our own AI but we're also leaning into the LLMs and the capabilities that they're bringing. In terms of products, I think where you see all of this come together and you saw this demo, like our ConversationRelay product, I'd say, is pretty exciting. The ConversationRelay product, for those that don't know, what it allows for is -- and we used it in the context of actually a pretty complex transaction, a mortgage application, where you can speak to a virtual agent during the context of that interaction. The virtual agent is able to store all of the data that transpires during the course of that conversation. Using our Conversational Intelligence capabilities, it's able to harvest the words and the context that really matters. You're able to complete the transaction through that interaction. You're able to switch channels. And for example, if you need forms delivered to you, you probably want that done over e-mail. If you want to be able to send a confirmation of a transaction, you probably want that done over SMS. And so now you have this complex interaction and something that is very emotionally charged, very complex, like a mortgage and a home purchase in the first place where you're able to handle that transaction through multiple channels, use context. And then as you can imagine, the mortgage companies, the financial services companies they don't envision that being kind of a one and done. I mean what they really want to be able to do is develop this lifetime relationship that actually probably starts with renting if your college graduate and then progresses to various home purchases over time. And their ability to harvest all of that data and be able to drive additional interactions over time allows us to be able to create more value for them and for them to be able to create more value with their consumers.

So before I turn it over to Matt, I just want to dwell on that for a second because there's this fear that AI is going to do what software does. And with the way you explained it to me, the context that Twilio has with its customer engagement layer, it's important for the LLMs to actually work. Can you tell us more why that is so differentiated and why is it that conventional belief, perhaps misguided as it is, might have you believe that the LLMs are going to figure it all out somehow magically and just obviate the need for not just Twilio, but any other software layer. What is wrong with that? What is mistaken with that?

Well, I don't think it's entirely mistaken. I mean I do think that there's a lot of truth to LLMs being very disruptive to a lot of different SaaS companies, especially single point applications. If you can orchestrate all of that specialty systems of record, I think there is a disruption that's about to happen. I think what's a little bit different for us is that we're not a SaaS company in sort of the classic sense. We operate at the infrastructure layer. I mean just at the bare primitives like what we do is we're allowing for this level of connectivity across 150 countries, thousands of different telecommunications companies. I don't think that, that's well served by LLMs. I'm not going to say never. I think people get into trouble when they make statements like that. But I don't really think we're susceptible in the same way. Like the super network that we have that sort of underpins our capability to deliver on these things, I really think that, that is like sort of our own version of AI before AI was cool. That allows you to be able to deliver these seamless interactions all over the world in microseconds across whatever the channel is. And I think that, that is very unlikely to be disrupted by AI. That's one thing. The second thing is that I actually think AI and the LLMs in particular, like they very much augment what we do. And the reason I say it that way is that if you think about the brands we serve, again, it doesn't matter what the industry is, health care, financial services, insurance, retail, I find it incredibly, incredibly difficult to believe that there would come a day in which they turn over their proprietary data, their knowledge about their consumers to the LLMs because if they did do that, they would completely allow for their own hard won customer relationships to be disintermediated. And instead, we're hearing quite the opposite, where their CISOs are basically saying, "Hey, listen, this data needs to be safe and secure. It's got to be in my data warehouse. You've got to be interoperable with my data warehouse. I want no copies of this data. I want all of the transactions to take place there. The LLM will augment and power." The context, though, is what I'm really interested in, in terms of being able to drive the interaction, and having that sit on top of our CPaaS layer, I think that is really defensible in terms of moat and stuff like that.

Yes. Khozema, you've talked a lot about contextual data at the start of the conversation. Segment stand-alone growth has been relatively sluggish for some time. But as we understand about [indiscernible] pretty [ intertwined with ] the core Communications portfolio. So can you talk about some of the innovation you're bringing to market in this capacity and the early traction you're observing around this?

Yes. I mean I think we certainly want to maintain the growth characteristics of the Segment business. But what's way more important going forward for us is how does Segment actually complement the whole. I mean the reality is, just to put it in perspective, like we're talking about less than 5% of our revenue, right, whereas Communications tends to be about 95% of our revenue. And so, so long as that product and feature capability can augment what we're doing in Communications, like that's where it gets exciting. To answer your question more directly, the way we wanted to operate is that when a customer comes to us to be able to avail themselves of voice or SMS or e-mail, they immediately start utilizing data characteristics so that they can drive context from the very get-go. And we've always been the company that abstracts this communications complexity away from our customers so that we can work on that, and they can do what they do best. I think now that comes with data and through a single API, they can immediately get going. They can start incorporating all the different data elements that go with all of these different transactions and consumer history and then use the channels to be able to activate against that. And ConversationRelay, I would say is perhaps the most fully borne product in that way. But we've also launched APIs with respect to Flex. With respect to voice, our Conversational Intelligence product, like that on a stand-alone basis, if you overlay it on top of voice, that has that capabilities. And so there's a number of different places that you can go with this. Ultimately, what we're really shooting for on behalf of our customers is persistence in that data and then customer memory.

All right. And you talked about ConversationRelay a couple of times. I think it's one of the more exciting part of the Twilio story. You've stated several times that voice is poised for a renaissance. So Twilio has been doing voice for years now, right? So what excites you most about the opportunity today?

I think two things. One is every interaction basically that's taking place, and you see this in venture funding, there are thousands, I mean, literally, thousands of venture-backed voice AI companies that are emerging, and we are delighted to count most of them as our customers, right? So that's an exciting starting point. The way that you see that show up in the numbers is that as recently as the last quarter that we reported, like we've seen this return to double-digit growth in voice. And I think what's especially exciting about that is, is that if you're talking about real life agents, voice had become a little bit clunky, and especially as you're talking about fighting like things like robocalling and stuff like that, like outbound voice to consumers became super clunky, right? And in most cases, I suspect that you weren't even picking up the phone just as I wasn't, right, because you didn't know who was on the other side. Now for all of this incoming stuff, it's perfectly suited. And the really interesting thing about that is, is that you're able to drive lower cost, you're able to drive more revenue because you have limitless time effectively because the consumer, they don't really care about how long it takes you, right? But it matters in a real life agent dynamic because they're measured on cycle time. But with a virtual agent, you have unlimited time, but most importantly, the consumers' problems are actually getting solved. And so I think you're starting to see that really develop. From a Twilio standpoint, too, our top 10 voice AI companies, they're growing at -- their revenue run rates are 6 figures, in many cases, 7 figures. That's exciting. What's equally exciting about that is it's a tiny percentage of our revenue, right? So on the one hand, we're not super exposed, on the other hand, I think there's a lot of upside here.

You talked a lot about the AI customers, and you guys disclosed in the Analyst Day that you have $260 million in revenue from the AI natives, right? Twilio was a direct participant in the hyper growth from companies born out of the cloud cycle. You've seen many cycles yourself. So any parallels you draw on here as to why next-gen companies continue to build with Twilio?

I think the big things from my perspective, I mean, I think all cycles have their ups and downs, right? And it wouldn't shock me to have us experience a down as part of the cycle. But I think what's more important about AI is that I think it's very much a secular trend. And I know a lot is sort of being written about, are we in sort of a hype cycle. Maybe, I don't know, one way or the other. But I think long term. There's just absolutely no question in my mind and probably on the minds of all of your speakers over the next few days that this is very much a -- not a trend, but it's going to become the fabric of everything that we do. I think for Twilio, just to knit a few concepts together, this idea of context, I think, is really powerful, and I think it really presents an unlock for our customers. I think that the AI's ability to do three things at once: Drive out cost, increase revenue and solve customer problems. I think that's really exciting. And then I think inside of Twilio, our own ability to be able to drive productivity, especially as it relates to customer service and SDRs, which is just the starting point. I think that's really exciting, too. And I think importantly, we're not overexposed as I said a moment ago, like it still remains to be -- it still remains a fraction of our revenue, and we're excited about the upside potential there. But being back to kind of double-digit growth in that context feels pretty good. Yes.

So just one segue and then I'll turn it back to Matt. $260 million is a lot of money. If an AI voice startup did that -- of course, it's your revenue from the AI voice startups. That is pretty significant. What is happening with these AI voice start-ups? I'm asking you this because over the weekend, I was watching Andrew Ng, the Professor of Computer Science at Stanford, who's also a CEO, talk about voice as a frontier that AI models have not really quite taken on, and you saw it as a fertile opportunity. And then you're telling me about these voice start-ups. What's going on -- what do these voice start-ups do exactly? What is kind of a very interesting thing that these companies do that can get us -- arms around what exactly is...

Well, I mean I think in many cases, if not most, I think a lot of them are very much customer service oriented, which I think is just the beginning. I mean, I think there's a lot of different directions that this stuff can go in. I think for now, it seems to be customer service. But I think what's exciting about it is, is that -- I'll just speak about it from a Twilio perspective, I can talk about it more generically if you'd like. But what we find in our own capabilities as well as in the partners that we work with is that latency, which historically had been sort of the roadblock here, like that's more or less becoming a solved problem, right? So you don't have these awkward pauses when you have these interactions take place between a human on one side and then a voice AI agent on the other side. I think that's important. The second thing is, is that, again, your ability to incorporate context, right? No customer care is really what they're interacting with on the other side. In fact, like we have a certain amount of research that says, especially in health care, most customers would prefer to interact with a voice AI because if it's a person, there's this feeling that there's this asymmetry in knowledge between the two sides, whereas voice AI is just that. I mean it's just virtual, right?

Could you spell first name again? Your last name again? I'm sorry...

But that stuff is gone, right? Like that's the interesting thing, right? So all of these like horrific IVR menus that we've become used to, if you can identify a voice signature upfront and then do a light verification on the backside of that because you've got to take out spoofing because that is a real thing, if you can get past that, you're right into the conversation, able to drive the interaction and outcome. What's also interesting, I mean, I've said this a couple of times, maybe just to put a fine point on it, there is a revenue upsell opportunity for the businesses as well, right? So if you think about this like in a retail context, food delivery, and this is an actual customer that we have, 35% their orders come in still, I mean, to this day, like over the phone, okay? And they flipped to doing voice AI interactions instead of human interactions, and it has the cost -- obvious cost impacts as you would think. But the ability to upsell inside of that interaction -- because again, human agent has to worry about cycle time. They're trying to rifle through an order and move on to the next one because otherwise, another customer is waiting on hold, right? The voice AI's capacity is unlimited, right? And so think about like GameDay, when you're ordering pizza or wings or something, I mean, we just said the opening of the NFL season this last weekend, like those are heightened periods in which...

Yesterday was GameDay in New York.

Yes. Well, for me, it was more about tennis. But anyway, the reality is that they lose a ton of business on those days because you cannot get through. So now all of that's gone. Not only is it gone, but the voice AI agent during the course of the interaction using context and memory about the customer can actually refer to prior purchase history and suggest upselling other products that are super high margin and allow them to have a better customer relationship. And the customer doesn't mind because this is stuff that they want. Now they don't want this with every brand, okay? But the brands that they really interact with frequently and really care about, they definitely want it. And the research all shows that.

That's great.

Maybe let's touch quickly on some of the more recent business momentum that Twilio has observed. You guys have delivered 4 consecutive quarters of accelerating double-digit growth. It's an impressive feat at a $5 billion run rate. What are the primary drivers fueling kind of the improving growth profile here?

I'd point to a few that are really big. I mean I think ISVs have definitely been a big part of the story. The interesting thing, and what's sometimes misunderstood about ISVs, I think people tend to think about the mega centers, like the really big guys, and I'm sure those guys are at your conference. But it's a super, super long tail. And so actually, the irony of this category is, is that ISVs sort of writ large, like pull up the overall average margin rate just given the absolute breadth of ISVs that we have in our portfolio. So I'd say that's been a big one. The second one I'd really point to is self-serve. Like self-serve has been awesome for us. And everything that we've talked about in the context of AI, like you can bet that we're applying it to our console experience to make it super easy, super simple for a customer to be able to navigate and get to that next use case or even the initial use case. I mean you guys have followed us for a long time, so you probably know that a couple of years ago, because of like a lot of regulatory stuff that was kind of shoved into the mix, like we threw a lot of, albeit necessary, but all the same complexity and clunkiness into the onboarding and sign-up process, like that stuff is predominantly gone and we've made it really, really easy to navigate those workflows. And we're trying to make it more and more intuitive so that if you're an actual human developer that you can just kind of sail through the process and get up and running to that magic moment as we've kind of talked about it in the past. And if you're a virtual agent, really the only thing that we need to make sure of is that your intentions are business-oriented and positive. Otherwise, if it's someone that's entered a prompt and just trying to reach an API endpoint, if it's an MCP and again, you're just trying to reach an end point, even if it's a virtual agent, like we're ready for that, too. And so I think that's been a really exciting part of the story. I think we talked about AI, that's been great. And then finally, international. International has been wildly underpenetrated from our perspective. It's a little tougher in the sense that we want to maintain our pricing discipline there. We don't want to just chase business, which I do think some of our competitors have done in some cases. And so I think that's why we remain underpenetrated, but there have been a number of instances in which we've been able to do competitive takeouts so much so that customers of ours who were previously using multiple providers have kind of gone all in on Twilio, and that's been exciting.

Yes. Maybe double-click on the international piece, right? It is a strategic growth pillar for Twilio. Is there any difference in the go-to-market playbook there? Like why can you guys win, especially given kind of the pricing dynamics in that market?

Yes. I mean again, the irony of international, lower gross margin rate, but better unit economics and those unit economics have been really stable over long periods of time. I think the reason anyone wins to start with is that these phone numbers and SMS, in particular, it tends to be like the ubiquitous channel. I mean I'd love to see WhatsApp take off for a variety of reasons. But SMS remains a really ubiquitous way to communicate with customers. All that said, the way that we approach it is we're very deliberate about where we have boots on ground. And so we only have boots on -- in the U.K., in France -- sorry, in U.K., in Germany, in Western Europe. In South America, it's Colombia and Brazil. And then in Asia, it's Singapore, Australia and Japan. And so very deliberately, those are the only international markets in which we actually have people. Now we do serve customers from other geographies. And in Singapore, for example, we tend to serve Southeast Asia, we tend to serve outbound China. We have difficulty participating in China itself. But when they're sending outbound to other countries, we'll serve China in that way. So that's kind of the way that we do it. And again, I mean, for us, it's really about maintaining price discipline. And I think we tend to serve a lot of enterprise customers because they want to be part of the broader Twilio story in which it's less price sensitive. They're putting together an entire package to deliver value to their consumers. That's a better way for us to win out there, I'd say.

And maybe the pricing dynamics is a good segue into some of the gross margin dynamics you guys have been observing more recently. Like some of that's been tied to the messaging mix. but you also saw an unanticipated headwind from A2P fees. So what steps is Twilio taking to stabilize and improve the gross margin in the near and long term?

Yes. I mean the most important thing I'd say about the whole topic, and then I'll answer your question is in spite of the degradation in gross margins in recent periods, it has not impeded whatsoever our ability to continue to drive cash flow, right? And we are very, very focused on driving cash flow and ensuring that as part of the framework that we laid out that we continue to drive improved cash flows over time. And so yes, we've seen some short-term dynamics. I'd call them largely mix oriented as a result of things that we've seen in our messaging business. On balance, that -- I'd take that all day long. We want to continue winning in that product. The A2P fee thing, like it has no bearing whatsoever on the pricing characteristics, and it doesn't even change the gross profit dollars. The other thing about it is, it hasn't actually impacted demand. I mean, again, these channels are very resilient, and SMS is ubiquitous and it's super cheap, especially inside the U.S. Now back to your question, I do think gross margin stability to a degree is like basically a self-help story for us. And so we have undertaken a number of actions. Being the market leader, we've passed through some price increases. Starting on SMS, like that in the U.S., in particular, we've raised prices there. That has immediately hit our self-serve customers. It will ripple through all of our other contracts over time that will take 2 to 3 years just based on the way that those negotiations work. We've also passed through what amounts to a price increase in our e-mail product. Effectively, what we did there was to lift the floor on free trials. So instead -- now, once you hit a certain threshold, you'll start paying for that versus just getting unlimited free. Inside of the company, there are also things that we're doing to be able to drive cost. So we use AWS, for example, there are more efficiencies that we can get out of that. We've been working for a while on transitioning some of our legacy on-prem stuff entirely to the cloud. So that double bubble will start to collapse over the next few periods, that will start to benefit. And then I'd say, finally, we're continuing to just drive the kind of blocking and tackling that you would do. FX hedging, I mean, we obviously have done some over the past. I think the tricky thing for us is that we're predominantly -- not entirely, but predominantly U.S. dollar-denominated in terms of our top line but our bottom line can fluctuate as a result of like where we go and source COGS. So we're doing some things in terms of the way that we build that will allow for more of a natural hedge between those two lines in the P&L. So all of those things in totality, I think allow for us to just kind of stabilize gross margins. And then over time, it's certainly my expectation that many of these other products that we've been talking about, they start contributing in a way in which they allow gross margins to come up over time.

Very helpful. I want to shift to the expanding partner ecosystem. You've launched partnerships with several marquee players, OpenAI, Databricks, Snowflake, Microsoft. Can you talk a little bit about the opportunities you see here from a technical and commerciality perspective?

Yes. I mean they're all different just based on the nature of what those folks do. I think to maybe state the obvious, like we want to work and be associated with the best. And those names that you just rattled off, I mean, I think that they are the best when it comes to the various kinds of capabilities that they offer. Now we're not necessarily going to be exclusive with anybody. I mean, I think especially with the LLMs, you want to maintain optionality. A lot of our customers have made choices about various providers that they want to interact with. Whether it's by the way, at the LLM layer or the cloud layer or the data warehouse layer, everybody has made a different choice, and there tends to be four or more players depending on who we're talking about. And so flexibility really counts. But I think just to kind of double-click on a few of these, I think with Microsoft, what's especially exciting is that their Azure AI workloads have proven to be pretty awesome. I think the functionality that they're delivering to customers has been great. We've been codeveloping products with them. We launched one at SIGNAL. We launched one the very following week at their conference. We're building out more product capability. What's been exciting to me to watch is that without any real leadership involved, the product folks and the engineering folks from both sides have been able to get into rooms and create a lot of innovation. And I think you'll start to see that bloom over the next several quarters. I think on the AI front, I mean, the OpenAI thing has been awesome, and we're really, really proud of the work that we've done with those guys. I think one of the cooler things that we did, obviously, was last Christmas, when they were doing their 12 days of Christmas or whatever they called it, 12 days of OpenAI, I forgot now. But they called us like 2 days before they wanted a thing done, and we were -- the 1-800-CHAT-GPT thing. And we were able to get that launched within like literally less than 48 hours, make sure it was ruggedized for the volumes that it would inevitably experience, and that thing really flew. And then with the data warehouse guys, I think interoperability is really important. Our customers are asking for that 0 copy, the CISOs in particular. Databricks and Snowflake are obviously leading companies. But so is Redshift, for example. And so you want to have broad relationships with all these folks, and I think we've been able to do that.

Maybe take a few questions. If you guys have any just raise your hand. We'll bring a mic over to you. Or just speak up. Yes. [ Sean, ] I can hear you. I can paraphrase the question.

Curious like for like a Sierra, who I think is a big customer of yours, maybe they're not using ConversationRelay but they're still using voice, how do you think about the economics on that and kind of like maybe some framing on just how big you think voice can get in context of -- in the total size of your business.

Economics of voice and how it works in the context of your business, how big can it be?

Yes. So just as a starting point, I mean, like any additional voice business that we do is really powerful in terms of like the way that it contributes to the P&L because it starts with higher structural gross margin. So that's a pretty good starting point. And so any additional dollars that we get in the voice business are good for us. I think the relationships that we have with Sierra is an important one. And it's one which -- what's especially interesting about it, I would say is that to be sure, there's like the voice infrastructure component that they're using, but they're also exploring a number of areas with like our intelligence products, which add additional margin on top of just being their infrastructure layer. And a number of these collaborations that I spoke of like a moment ago with respect to Microsoft, same thing with those guys, where our product folks are getting into a room, engineering folks are getting into a room and they're co-developing what we can both take out to market together. And so I think that's really exciting. And I mean, obviously, they're a great name. They're a great company to be associated with. But we want there to be 100 of those fundamentally. And so as it relates to voice writ large, I think we are optimistic about where that goes. I'm not going to guide to it necessarily or talk about it in terms of the proportion of the business. But I think starting with a return to double digit, that feels pretty good, and I think that's got tailwind.

So we have only a minute and 52 seconds left. Anybody who wants to jump on in with a really quick question? All right. There's one. Yes, please.

Kash, my name is [ VK ], private wealth client of Goldman since 2016. Just a question about voice agents with therapeutic health care. I know Twilio actually powers 1-800-Kokua (sic) [ (833) Hi Kokua ] for Tripp VR, which is in the digital therapeutic space. What are you like -- when you're talking about something as sensitive as mental health and digital therapeutics, and we are going to see so many AI start-ups get into that space, how is Twilio kind of addressing the sensitive aspects of people's mental health, talking to an voice agent and really it's all about giving more accessibility and inclusivity for something that's not used that often. But there's always these downsides. So what is your view on that?

It's a great question, but going very slowly to be honest. And so I'll give you an example in sort of the SMS world and then I'll bridge it to voice. So in an SMS context, one thing that we do today, for example, is that we do send notifications to elderly patients about when to take drugs, okay? So they've already been prescribed something. And the #1 reason that it doesn't work is because they don't take it when they're supposed to do or on time, okay? And so using SMS, we just send simple reminders that say, take this drug at this time. We don't provide any other context other than that and very, very specific instructions that are dictated by the hospital or the pharmacy, whoever our customer is. In the mental health context, we are not today in the business -- and it's not our business to like start providing advice, and we would caution our customers to go very slow. By the way, this is not limited to health care, insurance, same thing, financial services, same thing. Retail and e-commerce is pretty low stakes, right? I mean if you get a sausage versus a pepperoni, like maybe you're upset about that, but like it's not the end of the world, right? But in a financial transaction or a health care transaction, like it is a big deal. And so the most part, what we do in health care is serve use cases that are more customer care oriented, where there's a specific solution to a problem that can be identified using the context of the data that exists inside of a customer's environment and whatever context arises in that conversation. It doesn't tend to be, oh, by the way, with this set of characteristics, here's some advice. We're not in that yet. I would imagine that over time, what starts to happen is that you will have a lot of that traffic pass through our voice infrastructure that does not allow us into the intelligence game, if you will. Over time, we will layer that with intelligence so that the providers of those services are able to use the different bits that are specifically relevant to therapeutic outcomes to be able to drive interactions in the future. But again, our caution to clients is go slow, make sure you guys know what you're getting into because hallucinations in some of these industries can be -- result in the bad day.

So on that note, there's plenty of stuff ahead. It's a 4-day conference with a ton of presentations. So thank you for your time and attention. And I want to just point out to you, Khozema, that it's very rare, I really want to applaud you that a CFO turns into a CEO of a software company and is able to achieve a significant turnaround. So kudos. I don't know how you did it, one day we'll have to chat separately, but how did you make that happen? How did you turn on the CEO hat, the technical leadership hat and make it happen. So I hope you'll be long remembered for that.

Thanks a lot.

Thank you.

Thank you.