HubSpot, Inc. (HUBS) Earnings Call Transcript & Summary

I was tempted to leave the doors open because we always talk about having a fireside chats, but this is like having an oceanside chat. So like that's not too bad, especially on Day 2 of a conference. But for those who don't know me, I'm Ryan MacWilliams, mid-cap software analyst here at Fargo here for the 9th Annual Wells Fargo TMT Conference. With me today from HubSpot is CEO, Yamini Rangan.

Thank you, Ryan. Thanks for having us. It is a great location.

Like people are complaining on Day 1, and they're like, "Oh, it rained today." I was like, this is -- like I'm used to New York City in the winter right now. This is amazing.

Right. We'll take it in California. .

I'm jealous. Well, it's been an interesting time in software coming back after my break in launching coverage. And it does seem like it's kind of a rainy day in software right now, but that's a lot of opportunity. And I'm excited to hear about what HubSpot and what you're doing with AI. So for investors in the room, we're going to be speaking mostly around the product to start and then maybe some of the earnings at the end. So if you have questions, let me know, but we won't be taking questions directly from the room. So e-mail me at ryan.macwilliams@wellsfargo to get those in. But Yamini, just to start, with your Investor Day, there's been a lot of new products from HubSpot at this point. What would you say like differentiates HubSpot's AI offerings?

Yes. We just had our annual conference a couple of months ago in September and we did launch a lot in terms of AI. It's an exciting time to be in the industry. Even though it feels rainy, it is actually an exciting time to be in the industry right now. And just to give context, our AI strategy is to take something that is super powerful and apply it to the segment that we serve, which is we serve small, medium businesses. And we want to take AI and apply it in a way that can help them grow. That is really the strategy. And in order to do that, we are embedding AI into all of our products. So you saw a number of feature releases there. We're building agents that can help do work. We have featured agents that we launched at our conference, which is customer agent prospecting agent and data agent. And we have a world-class Breeze Assistant that forms like the Copilot that every go-to-market employee can use. So that has been strategy and we kind of launched a lot of products to drive that. And the reaction has been very, very positive. I mean, talk about AI adoption broadly and where we are seeing green shoots in terms of consistent usage. If we step back, you asked a question what differentiates our AI strategy. The first thing is we know SMBs and we've taken an approach of taking super sophisticated technology and making it very accessible to SMBs. That's been the business that HubSpot has been in, and we're doing exactly the same thing with AI right now. And that is the #1 differentiator because we know SMBs, we know what they need and their day-to-day job and their day-to-day challenges, and that's like #1. The second thing I would say is context, and I'm sure we'll talk a lot more about it. The way to think about HubSpot is we bring the context of every sales conversation, every campaign that was sent out, every e-mail that was generated, every deal that was closed and turns out while AI is really good at generating insight it needs context. Ryan, if you go to an LLM and say write me an e-mail, it's going to write some e-mail that might not actually convert the right outreach for you. But if you have the context of all of these conversations over a period of time, then it generates much better responses. And so our differentiator is that we have 280,000 customers using us for marketing, for sales, for service across the customer journey. We've become a platform that small, medium businesses rely on to drive growth, and that context helps make AI much, much better. And the combination of the domain expertise plus the context that we bring in helps us drive AI adoption and give value back to our customers.

And a lot to touch on there. And on the domain expertise side, it's always so interesting to me when I hear like the bare cases on what AI can do versus application software and people are like, oh, what if you can recreate this exactly, but it's like, okay, what happens the next day, right? Like who is going to think about like what's the best ways for customers to make more money off their own customers, right, or new use cases from a customer service standpoint? And I was like, that's what HubSpot does. You have hundreds of people and like decades of experience doing this. So like it's not static in terms of like, okay, this is what a lot of can do with, okay, how do we continue to advance on the platform for what customers need.

Exactly. I mean, look, coding has become easier, but expertise is still important. And I do think that the domain expertise and being able to apply something to a certain segment and make it easier for them is still where there is value that is getting created. And we think that HubSpot today is adding way more value for our customers than we did 3 years ago because of AI. And so it doesn't just go away with AI.

And I think that's a great point in terms of like there's a difference between like the coding advances that we've seen from large language models because that's a publicly available data set, right, that has stack overflow and has a more deterministic outcome, right? Like code is either right or wrong. You can debate like a better way to do it, right? That's like chess. It's a much easier type of problem to solve. We're like, how does the B2B organization correctly address this one use case for their customer. That seems a little more challenging.

That's exactly right. .

So when it comes to HubSpot, you have a large customer set you've been working with for a long time, like they're going to be asking you for certain AI solutions that you can help with. But how does your data advantage also help build like a more holistic workflow?

Yes. So I think before we go into the workflow, maybe fundamentally, what is different in an agentic architecture that enables all of these workflows, right? So if you take it down to the foundation of this, you can do much more with unstructured data, like CRM and customer platforms have always had structured data in roles, in records, that's what we are good at. So an example is like a customer record. You have the name of the customer and the address of the customer, the revenue, that's a customer record. That is what we've always been good at. Now with agentic -- becoming an agentic solution, you need to be able to handle unstructured data. the conversation that we just had, the transcript of a Zoom call, something that you said that is out on social that we can now grab, that is all the unstructured data. And what AI made it possible is to process all of that unstructured data and add it to the same context that you have. So that's the one big change at the data layer that we need to. And so our solution, it was easier for us to go from having all that structured data to now capturing e-mails, like Zoom, transcripts, video calls, audio calls, all of that and add that unstructured data layer. The second thing that you need within an agentic platform is really orchestration. It's not just about dumping raw data but having the context across all of the data, where do you get feedback, was that answer good or not, which is evaluation and where do you have the memory of the questions that you've asked in the past. So the orchestration layer becomes really, really important with evaluation, feedback and memory, and that is what we have built. And then what is changing in terms of how applications work is it used to be that you would go and point, click and navigate to something. Now you can have conversational ways of asking software to do something for you and you have agents that actually do it. So just to be like really clear, what has changed is the amount of data that you can process, the level of orchestration that you bring and how you can have a conversational way of interacting with software, that is literally what we have built over the last couple of years in terms of the foundation. And so then it becomes like how do you enable workflows? Well, because we have much better unstructured data, now the workflows become much better. So instead of using a CRM where someone had to go and contact and create a contact and say, I met Ryan for the first time today, and this is a contact and this is the conversation. Instead, you can process all of that through the unstructured data that you're able to capture. So workflows become much more dynamic and workflows become with much more context of unstructured conversations that you're having, and that enables much better output for our customers.

Yes. We were just talking about that in regards to what we do every day. And how like before you log a call, and you're like, I'll put a couple of notes and it doesn't matter. But now that you know it's going to help you down the line. okay, maybe I'll add more detail there.

Sure. Yes. Exactly.

Your data second is even stronger. One of the funny things I keep running into is like we talk about like Oh, I'm going to do this, and it's going to write all my research for me. But sometimes it's just like summarizing an earnings call is like the most helpful thing right? So it's like some of the unstructured use cases are actually getting adopted faster or customers are more interested in. So far when it comes to your customer base, is there any use cases that might have surprised you or are things that your customers are more interested at this point?

Well, I mean, a ton. So I'll tell you, I got started in sales like multiple decades ago. And back then, the hardest part of the job was you'd get a set of accounts. let's say, at the beginning of the year, you got like 500 accounts. You just never had the time to look at 500 companies look at 5 people per company of who was the right contact? And are they going to buy? Are they ready to talk about your product? Like that was all manual work, Ryan. What has been fascinating with AI is that the prospecting use case, specifically one where you can now use AI to go and grab the right information about each of your 500 prospects that you have, did they get funding? Are they adding marketing reps or sales reps or service reps? Are they mentioning initiatives that you can help? This is all kind of unstructured information and structured information that you can now get. And based on those intent signals, you can say, next Monday morning, these are the 10 that you need to talk to based on the intent. That is so much more value than we have ever been able to deliver. And so I think the prospecting use case. The other ones are pretty obvious, like support is done, right? Like everybody knows that AI can be used to resolve support inquiries, and it's getting better and better across the different channels that we're supporting, marketing in terms of content creation. But to me, like sales as a function is fundamentally changing of where you spend time and where you get leverage out of AI, and it's just been fascinating to see our customers adopt that and get value.

Have you heard any pushback from salespeople who are saying, no, I'd rather dig around through all this information?

Oh god, no. Actually, salespeople, the things that they don't want to do is dig around for information put in notes of the calls that they have had show the work and activities that they have done to the managers so that they can contribute, like those are the kinds of things that salespeople do not want to do, which AI is actually really good at doing. What salespeople enjoy doing is being in front of customers and having deeper conversations. And when you take a lot of the extra work that you used to do and you make AI really good at that, then the time in front of customer and the relevance of the conversation that you're having with customers goes up, which means your close rates need to go up. And I think that's the exciting part of what AI can do for sales.

Yes. You can spend more time on actually selling with all the back-end work.

Exactly.

And as a former customer support agent, I can tell you a lot of those use cases, I'm okay with.

Oh yes. If you start in support, that's awesome.

Yes. It was, for student loans. Buy siders are tough, but they're not as tough as people calling in about their student loans. But when it comes to like your new consumption and credit usage model, For me, like I'm still really interested in apps here because I've used Cursor and Claude Code. And when I'm using those services, I'm clicking, AI do it for me constantly, right? So something like Claude Code, developers use like $6 a day worth of token usage. So that's something around like 40 or 50 times, you're clicking the AI do it for me button, right? So when I think about that, it's like, oh, well, what people live in all day and the platforms they use all day, right, they'll start to like do the AI do it for me button or whatever it is Atlassian or HubSpot or what they're used to. So in terms of like that credit motion that's newer for HubSpot today, can you just discuss what you're seeing like the early customer trends and use cases where those comes?

Yes. And just to maybe step back. We talked about the AI strategy, which is embedded AI agents and Copilot or Breeze Assistant, and our monetization strategy is hybrid. We monetize AI both through seats as well as credits. And so for the example that you just gave, within HubSpot, if you are a Sales Hub user and you're clicking multiple times today to summarize the e-mail and get me the next follow-up e-mail, which you can do today. You can say, summarize that call and give me the follow-up e-mail, that is part of seats and that does not consume credits. So I just want to make sure that people understand that part of the way we monetize is through seats. Now the credit specifically is for agent work that we do. our customer agents resolve support tickets and consumes credits. Our processing agent does the account research that we just talked about, and it consumes credits. Our data agent brings in data and cleans up data that consumes credits. And so there are a handful of agents as well as Data Hub that consumes credit. And that we launched in June for all of our customer agent customers, and it moved into installed base in August. So it's early days but in terms of credit consumption, Customer Agent is #1, leading, because it's been in general available mode. And our customers, we have over 6,000 customers resolving over 60% of their tickets using the Customer Agent and they're consuming credits. And the second area is Prospecting Agent. This is where I do see a lot of promise because this is a known age-old problem for sales that we are now able to solve, and that's the second area. And the third is intense signals within data. So those are the areas that we are beginning to see. But look, it's early days. We believe that AI monetization for us is hybrid monetization, both through seats as well as credits, and we're seeing kind of all the right signals.

I actually appreciate that distinction as a part of your pricing strategy. So it's like if you're doing work in tandem with like HubSpot, then that's a part of the platform, right? But when it's doing work for you, that's where you can start to monetize those credits?

Exactly. And for example, our within Marketing Hub, you can be at content, you can remix content. That's all just part of Marketing Hub. But if you take agents that do actions for you, that's where it consumes credits. And so I'll probably keep repeating this, that our AI strategy is hybrid across seats and credits.

That makes sense to me. When it comes to software today, like if I was a power user of HubSpot, I use it all day every day and there's some one who use HubSpot like once or twice a day, you might be paying this amount on a per seat basis, right? And I don't know if that is going to be the same way in the next couple of years given some of the usage dynamics here. But in terms of who is adopting AI first that you've seen so far in your platform, is it power users? Is it SMBs versus like larger businesses? Or does it run the gamut?

Yes. It's a good question. I'll tell you, it is not based on the number of employees. In fact, the really distinguishing factor is, is there a C-suite leader that's pushing AI priority within the company. Like I've looked at it by industry, and I've looked at it by segments, and we've done a lot of analysis. It's really not based on the size of the industry. It is like, is there someone there that is top-down pushing AI because I would say that I know there's a very different narrative with investors, but when we talk to customers, there's still a level of fear and uncertainty and lack of trust with where is the data going and how do I make sure that my company's data is not used for some LLM training somewhere. So there's just a lot of like mistrust associated with that. So the #1 factor is top-down kind of initiatives on AI. And once you get past like there's someone within the company that is looking for an AI road map, then I do see that there is a strong ops role. And in order for AI to really drive value, you need kind of what we used to call RevOps back then, and now we call them AIOps. And there's someone within the company, the power user is like an AIOps user that trains that gets the right quality of the data, trains the data, uses the AI features and then makes it available for everybody else. That role is kind of like the power role. And when we see customers that have these AIOps roles and someone like a go-to-market engineer, that is leveraging AI, then the AI adoption actually accelerates within the company. And for a lot of our customers, they're still at the stage of, we want clean data that we can trust that we know is not being used outside of our company, and then we want to have a road map of reasonable set of use cases that we can experiment, scale and then grow with.

And it's amazing how like just one champion can really move the needle and you can activate a lot of others. Like we did a Wells Fargo off-site, and I had ChatGPT answer a question then I had everyone build the agent on their ChatGPT. And I think actually got commissioned about the number of ChatGPTs. They paid licenses we sold after. But once you get started to like, oh, I can do this, I can do this, then that makes sense to me. And that's actually a good segue into, I think the Data Hub strategy is one that kind of needs more airtime with investors here where your customers are already trained to like put everything they can in HubSpot and then activate off that. So you've rebranded that at the most recent Investor Day. Can you just talk about how Data Hub helps your broader AI road map?

Yes. And look, already today, we have talked about the criticality of data and context for AI to work and that is known. And if you step back, Data Hub does a few things. One is it pulls data outside of HubSpot into HubSpot. So we have things called data things that are kind of like integrations that pull data from across. And typically, we'll see HubSpot customers have anywhere between 8 to 14 integrations, but sometimes they're bringing data from more sources to bring into HubSpot. So that's like Data Hub and helps you do that. The second thing that we are finding is that data quality is exceptionally important for AI to be accurate. And Data Hub actually helps you improve the quality. It can run a set of prompts to LLM. And so for example, if there's like a column of like funding data that you need, Data Hub will pull the right problems and get the funding data for every contact that you have within your database. So it improves the data quality. And then the third thing is it provides a workspace, we call it the Data Studio to now manipulate the data. You've brought in the data. The data has higher quality. Now can you build workflows and sequences and automation with that data so that your AI works better. And so it's almost like a foundational workspace for AIOps and for RevOps to do much more with higher quality data. And that is kind of one of the reasons why we rebranded it from Ops Hub into Data Hub because you're really working on the foundational in a need for AI. And as we look into the future, we haven't talked about marketing. But in order for marketing playbook loop to work, you need higher quality data and Data Hub provides that. So that's the vision. And it's now one of those areas where if you're a Marketing Hub customer and you want to do AI, then you need to get like Data Hub as a foundation for it. And it's the same thing with Sales Hub. It's part of the multi-hub play that we have.

And we've kind of touched a lot about the data advantages and the differentiators of HubSpot here. But it's interesting when people was like oh, well, I use these AI cases. I'm like of what or how, right? So I mean, I think like the larger question that's been on top of mind for investors is like, does HubSpot add AI features first or do I end up doing a lot more of what HubSpot does. So we touched on a lot of these things today, but we'll have kind of just like you probably had this question over and over again over the last month, but we just kind of love to hear like all the things we talked about today like why you guys are better positioned in that world?

Yes. I like the framing, is it easier for a SaaS company to add AI? Or is it easier for an AI-native company to add like CRM. It's a very -- I like the question but foundationally, I'd go back to the conversation that we just had, Ryan. We've already become an agentic platform by adding unstructured data, by adding a context layer that ties together all of the data by building agents that can then take the context across that. We've kind of done a lot of the actual plumbing and the architectural changes that are needed to support becoming an agent platform. If I were starting out as an AI native from scratch, I need to build still a CRM record and I need to build the structured data. Many of them are starting as point solutions. So let's say, you start as a support agent and dealing with support tickets. The minute you get a question on sales pricing, then you don't have that context. So you now have to extend and start building what are products that you sell, what is the pricing associated with it and what are the common questions and objections. So you have to go from structure to unstructured. You have to go from point solution to full platform, then you have to build the full context associated on top of it. And then I would say a couple more things. We're still finding that AI features get adoption with feedback. So we have the advantage of building an AI feature and making it available for thousands of customers to use that we can get the feedback and improve AI development cycles are very iterative in nature based on customer feedback. But if you're an AI and native company and you're starting with 10 customers, where you're going to get that feedback. So there's an inherent advantage in agentic world that benefits from the scale of the number of customers. And then the final thing I would say is you still need a partner ecosystem. AI is still one where you require someone to look at your road map to help you with what are the use cases to start and drive, and we have partners within the ecosystem. They're all driving AI readiness. So if you are a new company that is getting started as an AI native, you also need -- that's why you see a lot of forward deployed engineers, right? The model is like you not only build a product, but you also invest in forward-deployed engineers that go and sit in customer sites. And we have that, which is an ecosystem that we have. And so I think like there is platform advantages, there is scale advantages because of the distribution and then there is an ecosystem advantage that we have.

During the break between the jobs, I was sitting around thinking like I could probably try to build one of these. You know what I mean? And then as I did some research on it, it's like...

Did you?

I tried to build a project management tool and it looked a lot better in my head than it ended up being on paper. And it turns out like having the initial draft is a lot harder than having actual working SaaS, right? But as I talked to other developers, it was like would probably be one of the most difficult things to build given like the high data density, mission-critical workloads and unstructured data but it's such a big market opportunity that people are going to try.

Yes, absolutely.

And so when it comes to like you have the data already with your customers, and you're already training your models on that data that also extends like the time that a challenger would have to take in order to create the same, like use case based off your new training model.

Yes. I mean, look, I think we've always been in a very, very competitive or has never been a winner-take-all market or a noncompetitive market. It's always been the case. And I'd go back to like why do platforms win over point solutions. One of the most common use cases for HubSpot is that we'll go to customers and they'll say, our data is fragmented across 15 different solutions that are point solutions, and we've lost visibility of our growth. So that's like -- that continues to be the case in the agentic world. When I talk to customers who have even attempted like different agents. They're like, well, we now cannot manage this across all of these different agents. And we want to look at your road map, if you have it, then we're just going to continue adopting with you. And so I think like it comes back to why platforms win over point solutions. It's the same reason why an agentic platform wins over point agents that are kind of sprawling across go-to-market.

I would love to touch on that point. Just like that complexity that you're speaking to. For me, like it's a very interesting dynamic of like, oh, I can custom build my and use case now. But like that's investors kind of thinking that's the case? Where like are you seeing customers being more willing to like know whether buy AI to start and build on my own?

Yes. I mean, custom building has gotten easier. But think about our segment true, right, which is a 500-person company and let's say, a manufacturing company in the middle of the country that's trying to grow their business. Now they have to figure out Sonnet 3.7 and ChatGPT 5.1, the latest Gemini that dropped today for 10 different use cases, and they're building a custom application on profit, that is not what an SMB wants to do. And while there is a narrative that yes, custom applications and custom agents are easier, average mid-market company still focuses on growing their business versus let me run and adopt AI just for the sake of it and build a customer application. And you said this earlier, there's a lot in terms of building an agent, which is there's a front-end piece to it, there's a back-end piece to it, there's getting the right data into it and there's this constant iterate process of getting feedback and improving it. And in order for a company to do that, they really have to change their strategy and invest pretty significantly in either AI engineering talent or AI pods that just do that. And that is not the conversation that we hear with our customers, they, in fact, want us to make it like easier for them to adopt and they just trust us to do that because of the level of innovation that we have brought to them.

Absolutely. And the thing that worked yesterday is not going to work today. And you guys have definitely been on the forefront of speaking about how SEO is changing as a part of AI and you guys had the acquisition XFunnel in regards to SEO, and I love all this about it. We've been following pretty closely for what we cover. But can you just talk about your own efforts to solve some of those top of funnel changes? And then like how you're helping your customers with the new SEO?.

Yes. I mean it's a topic that we can spend quite a bit of time on. I mean if you step back we, as an industry used to put content and get people to click on the blue links and have people come to our website. Then we captured their e-mails and then we nurtured and then that became part of the marketing funnel. That is completely disrupted, right? Because AI overviews are providing the answers, and that is if you search half of them are not even searching, they're just going to an LLM to ask questions. But because of that, there is fairly massive drop in terms of a very specific type of lead call content leads, right? Content leads are the ones that have really gone down. Now in terms -- you asked a question about how have we navigated it, I mean, I'd say that in 2022, even before ChatGPT came on and AI overviews were part of our language, we saw that customers were spending a lot of time on other channels, social channels, podcast, we're all listening to podcasts and we saw that everybody is listening to like hundreds of podcasts now. And so e-mail newsletters. And so starting 2022, we went through a process of diversifying our marketing channels. And that strategy has worked. We have 10 YouTube channels, and the leads from YouTube channels are growing between 80% and 90% year-over-year. We acquired a podcast network and we now have like over 100 podcasts within that network, and that generates leads. We also acquired e-mail newsletters, which we did, people were scratching their heads saying, "Why are they doing this? But that has also increased the leads. And so for us, the story has been how to diversify out of content leads into all of these different sources. Now one of the channels happens to be AEO, and that is Answer Engine Optimization where you show up within an LLM. But that is still low -- it's nascent, right? It's pretty early days in terms of AEO and how you show up in answers. And the lead volume is still very low. It's single digits, but the conversion rates of the leads from LLMs is much higher. It is 3x because you're doing a much thorough, deeper questioning and you're ready to convert if you get the right answer. And so that's kind of how we have navigated it. And the thing that we are really good at is once we figure something out, creating a playbook and getting our customers to be able to adopt and get the benefit of the playbook. So at our conference this year, we launched Loop, which is kind of the playbook of how to diversify your lead sources and drive a level of personalization. And so we launched that, we launched a set of products across Marketing Hub, Data Hub and content hub features that support that playbook, and we're early days, but kind of getting the new playbook out. And I've now spoken to a lot of our customers Polestar inbound and they've started doing some of this, and it clicked. It's like, oh, I now see that here's how I need to diversify. Here's how I need to use AI for personalization and this is what is happening within the AEO channel. And so early days, but we feel like this opens up a big opportunity for us.

I mean there's always something, right? But it's like there's more complexity now is to reach our customers where they are, more personalization, like these are all things that they need some of the help with.

I think we're actually excited about it because if you look at the last 2 or 3 years in marketing, channels were saturated ROI was really, really hard to get, and conversion was slowing. And the playbook was like super difficult. And if you're in marketing within a small and medium business, you couldn't get even 1% to 2% improvement in any of these key metrics of lead conversion. But now there's a completely different way to do it. And the level of return that you get from figuring 1 or 2 of channels is just much better. And so there's just a lot more excitement in terms of what you can do with AI within marketing. So while everybody talks about the disruption of SEO, what maybe is less understood is that A is actually creating a much bigger opportunity to optimize your marketing channels and strategies, and that's an exciting opportunity for us.

I have a few minutes left and this might be a bigger question. But for me, investors often ask like, oh, why don't they have all these like amazing products now this for every sulfur company. I'd like look, reasoning large language models came out at the end of last year. It takes time when you have hundreds of thousands of customers to put like really intricate products in. So besides like there'd be a little more seasoning across like all of AI within solutions, like what do you think really picks up the adoption curve for your customers?

Yes, I think that I would still say the technology is ahead of customers' ability to adopt the technology. And so I think you're absolutely right. LLMs came on board and a lot of us, we understand the transformative power of AI, and we've been building, but it's an iterative process and you got to get customer feedback to be able to make an AI feature better, and that's one part of it. The other part of it is helping customers through the do I have good quality data. Can I trust where my data is used can I make sure that my problems and all of the interactions are within my company's walls are not being used to train something else. And there's a level of just comfort around that, which is the adoption cycle. And again, it's not dissimilar to any other technology cycle that we have seen in the past. And I've been in this industry when -- I joined the industry in '96 when that was like before we went into cloud, and it was very, very similar. Everybody saw the value of it, but it took a little while for people to begin adopting and it's the same thing that is happening. There is just transformative value in AI, but adoption comes from building trust, better quality data and then making it frictionless for customers to try something and then scale it. And that's the process that we're in.

I'm really sad to see what comes next, and I'm certainly not going to miss searching through our CRM for all the clients I talked to over the last 2 weeks.

That's exactly right.

And to those that comes through next. But guys, thank you for your time, and thank you, Yamini for coming.

Thank you. I really appreciate it.

Thank you.