Five9, Inc. (FIVN) Earnings Call Transcript & Summary

All right. We're going to go ahead and get started. Thanks, everybody, for being here. I'm Will Power. I cover cloud software for Baird. Pleased to have, I guess, starting with Dan Burkland, who's President of Five9. We have Jonathan Rosenberg, who is the Chief Technology Officer and also Head of AI. So that will be a subject I know of interest to all. And there's Barry Zwarenstein, who of course is the Chief Financial Officer. For those that may not be familiar in the room, Five9, of course, is a leader in cloud contact center. Barry has a quick safe harbor, and then we're going to jump into Q&A. And if there are any questions from folks in the audience here, you have instructions in front of you, and I'll get to those as well via the iPad I have.

So thanks, Will. And just before we begin, a reminder that we will be making forward-looking statements about events and trends in effect, the industry, the company, its operations, including things about product development, AI and automation, our growth projections and so on. Actual results may differ materially from what we say and should not be unduly relied upon by the investors. For factors that could cause such differences between the actual results and what we're saying, filed with the Securities and Exchange Commission, the annual 10-K and 10-Q and the captured risk factors and other areas.

Okay. Awesome. I think you've done that before. So maybe, Dan, just to start with you. I think it might be helpful for the group. If we just have a 1-, 2-minute kind of quick overview of Five9 kind of the core solution, what kind of differentiates you in the market?

Sure. Thanks, Will, and good morning, all. If you look at Five9, we help companies of all sizes deliver customer experience through our platform and when we say customer experience think of all the customer interactions that a company may have, okay, from you as a consumer reaching out to a brand and wanting to interact with them. That first interaction, whether it's a voice call into their business, whether it's a chat, whether it's an e-mail, whether you've got on social media, whatever it might be, we're going to receive all of those interactions, we're going to identify who you are, we're going to identify and derive the intent for your inquiry. And then think of us as this platform that then is going to decide how to treat or how to handle that interaction or that inquiry, and we're going to determine if it needs to go to an expert agent, human agent, or whether it can be handled with automation and be self-served and handle your question right there where we can receive input, fetch an answer and speak it back to you or whether it's a combination of a human agent being assisted by AI and automation. Think about it as we're listening into the conversation, we can transcribe it. We can apply NLP to it. We can go, "Oh, we know the question. We can go to the right data source, fetch the answer and bring it back to even an agent and assist them in being more effective, more productive and more effective at providing that answer." So we're this platform that handles all interactions. And the reason I mentioned that right now is a lot of folks are talking about the AI thing, and we'll get into that, about how is that going to offset the human agents. We love that. We've been at this for many years. In the last 3 to 5 years, we've made strategic acquisitions, inference being the leading intelligent virtual agent solution, which allows us to fully automate some of those transactions that are straightforward. We acquired a workflow automation solution from a company called Wendu, which allows us to do more complex workflow. So we look at this and say, we're going to go into companies, help them set up their platform handle all their interactions. And that volume not only doesn't change, in many cases, goes up as we make it easier for consumers to interact with their brand of choice. And we want to give those companies a variety of choices for how those consumers interact with them and make it as simple as possible.

Maybe before we get into the AI discussion, just any perspective on kind of strategic areas of focus, maybe for you, Dan, from a go-to-market standpoint, what you're focused on targeting here over the next several years? And maybe, Barry, too, just on the financial side.

I'll start it, and Barry can chime in. But from a go-to-market perspective, there's 3 key growth drivers to our business right now. One is as we deliver cloud contact center, the 1 thing that keep in mind estimates are somewhere around 20% to 25% of contact centers have moved from their on-prem legacy in many cases, end-of-life platforms where they're really delivering a poor customer experience to make that migration to the cloud. They have to make that migration as those platforms have been discontinued in some cases. Certainly, they're not innovating or enhancing them at all. And so that migration will occur. So that's kind of the first phase. This move up mark. As we've moved upmarket, we've had to prove ourselves at every step of the way. We believe we've got 2 of the largest, if not the largest cloud contact center deployments, large global parcel delivery service and a large health care conglomerate, both Fortune 50 companies. And by serving them, it's opened up that market in the high end of the enterprise. And we've seen that pipeline double in the last year. And so that's one driver is move up market. The second is international expansion. We've now built out our not only infrastructure, but our sales and go-to-market teams throughout Europe and Latin America and Asia-Pac next. And then the third is our channel expansion giving us more reach, not only to get more opportunity and make sure we cover the market, but actually enable many of our partners to now implement our solutions so we can get margin leverage as well, leverage from them implementing our solutions. Those are the 3 main drivers on the go-to-market side.

Yes. So I think Dan summarized it pretty well, a move-up market, continue to grow internationally. We've been having a lot of success there. We've made a lot of investments across the globe. In part, we dragged there by these major orders that Dan talked about -- growing there internationally at over 40% in 9 out of the last 11 quarters. If you look at it in terms of the income statement, on operating expenses, we are pretty much where we want to be in terms of our long-term model, our long-term model at the midpoint is 47% in 20 of operating expenses combined, we had 44 so we actually got some room there. We do want to increase somewhat is in R&D. We -- in terms of gross margin, which is the other area that we're going to be expanding, currently, at 60%. We've got a road map to get it to 70% plus. And we can go into that separately at this time.

Okay. Great. Okay. Well, I think everybody here knows the bear-bull debate just been kind of centered on AI. So we'll jump into that, especially with Jonathan here. So thanks for being here. Given your background, and we're just discussing outside. Well, maybe it would be helpful, actually, when I get into question, just provide your background at both the company and prior. But to help level set, maybe you or Dan can talk about where you are with AI today. And then what's the next step, right? What is Generative AI mean? Because obviously, the question has been whether that's opportunity for us. So what are the key AI products today, solutions? And where are you going with Generative AI?

All right. I'll take it away. So I'll just do a quick introduction myself because I'm not normally in the financial community, I'm Chief Technology Officer and Head of AI. I run our overall AI engineering teams that build the products that -- like our IBA or agent assist and other applications, also direct the overall technology for the business. I've been here about 5 years. Previously, I was the Chief Technology Officer for Cisco's Collaboration business unit. And prior to that, the CTO for Skype. I have been a long time in the telecommunications industry, known in industry as one of the vendors of this thing called SIP, which is the foundational technology for voice over IP. We also use it here at Five9 as it does every telecommunication provider. So I've been doing both telecommunications and real-time communications as well as AI at all those places. I ran the AI teams at Skype, at Cisco and now here at Five9. So to directly answer your question, I guess, like is this good or bad for us? It is without a doubt good for us. And the way to understand that is that Five9, we provide a platform that's focused on delivering an end product of supporting customer interactions, whether they're phone calls and e-mails and chats. There's a ton of things you have to do to make that work. From routing calls, queuing calls, connecting to agents, deciding -- integrating data to answer questions. And so our platform has never been one that actually does what we call the AI engine. The thing that like transcribes the voice to text or understands the intent of a spoken word. Our platform is built to allow us to plug into third-party vendors who build those technology engines. And so we today make heavy use of Google, for example, which is what we use for speech recognition frequently and some amount of what's called intent detection. We use Nuance, which is one of the leaders in sort of the prior realm. And the strategy that we have of integrating the engines that are provided by third parties and turning them into this finished product, right, has proven extremely, extremely fortuitous and a great decision for us because we've now seen another change of engines. We went from the prior generation that was intent-based and now to this new large language models from vendors like OpenAI, and our platform has allowed us to quickly switch to those engines to allow us and then ship finished products that take that raw nugget of capability and turn it into something. And again, the way to understand it is like, if you're a contact center, you don't want to buy like you're not in the market for an LLM, you're in the market for a customer experience solution that can handle customer inquiries over e-mail, chat, voice, digital and then like answer them. And you have -- and so the ability to recognize and understand the spoken word is like a piece of that. And so we provide that whole solution.

Okay. So I guess one of the big questions has been -- I think we're going to try questions from the audience here, too. It's been how the new LLMs OpenAI, Google, et cetera, kind of impact the road map that you've been on? And what kind of new competition could that inject, right? I mean does that can enable start-ups that are funded by VCs, et cetera, to kind of jump into this market with a lot of data that's publicly available.

So I'll start with answering the road map question, and then I'll hand it to Dan probably for the competition question. But from a road map perspective, it's been a massive accelerant to our capabilities. And I'll give you a concrete example. One of the products we built and shipped on the prior gen technology was agent assist summarization. What is that? So if you're a consumer and you call the contact center, you chat with them, you're speaking to a live agent. At the end of the call, the agent has an obligation to make these notes on the call and put them into the case system so that next time you call back, the next agent knows what happened. Historically, in the contact center, this is a timely process. It takes a while for the agent to write notes. They're not very good at it. They produce -- and some of them just don't even bother. And that's why you've all probably experienced this, you call the contact center, you tell them your story, you call back the next time, why don't they remember who I was? Because they're not really, the agents are not doing a great job at this. So we shipped a product about 2 years ago that listened on the call and summarized it automatically. The agent approved it and then it put the summary into the case history. And it worked -- okay, good, but the technology could only go so far. When these LLMs came out, we did a quick test, and this thing is brilliant at summarizing conversations, and it didn't require any configuration or setup which the prior generation did. So because our technology was built to allow us to swap the engines out, we chuck the other one aside, we pulled in the large language model, and the product got like 10x better with 0 configuration. And now what we're doing is we're hitting the gas pedal on distributing this product into our installed base because it doesn't require any configuration or setup. So now everyone can try it. So we've seen this technology take a product we already had and make it way better and allow us to accelerate it. And that's true for other areas of our product portfolio in addition to opening up new capabilities we can never do before.

And if I may, on the question of newly funded startups, yes, you can run out and build a chatbot in a weekend. However, when you call into a contact center, you're typically calling for -- you have a question about you. What's my balance? When is my next payment due, very specific questions about you or your product or you're standing with that company. The large language models understand now how to know what you're asking, but they don't have the answer. It's all those data systems. In some cases, in some of our larger customers, it's dozens of back-end systems we want to integrate to the billing system to know when your payment is or was due. We want to integrate to the CRM to know what products and services you purchased. We want to know the value that you create to us so we can prioritize you a certain way. But that information is all within the company's data sources. And we have all those integrations done with a single platform regardless of whether you came in over voice, chat, e-mail, it's one set of integrations so we can deliver a consistent customer experience across all those different means, those different channels. If you put in a point solution on the edge, guess what, I got to take that chatbot and integrate it to all 40 of those systems. Cost prohibitive. There's no way you'd never do it because I've got to know how and when I hear a question, I got to know where to go to get the data and it might be in many, many different sources. The other thing I need to do is I may not answer the question effectively in that stand-alone start-up bot, and I need to go speak to an agent. So I hit 0 to go talk to an agent or I say representative or agent. Guess what? All that data and information that was collected in that session that may not have been completed or contained within the automation needs to travel to who? The right agent within that environment. Who knows where the right agent is? We have status on every agent, what their skill set is, what their availability is, and are they the expert that we want to put you in touch with? Okay. Then if it's Barry that's expert agent, we need to take all that data and all the other data that might be pertinent to a conversation that's about to happen and deliver that to Barry. That can't happen from a stand-alone solution. It has to be part of the platform.

So an agent system is a great example. So I guess maybe 2 parts kind of following up on that, a, our customers are already starting to recognize the benefit of that and what kind of reception are you getting? I mean because you've talked about in the past, I think Barry like AI being 10% of bookings or...

10% of bookings, yes -- in revenue.

I mean, should we start to expect that number to rise based on the conversations you're having? And then the secondly, like IVA is LLMs or some of the -- does that new engine capability impacting some of these other key products, right? What's that road map?

Yes, a few questions to unpack there. If you take the question about moving towards this automation and the adoption of agent assistance as an example. Think of AgentAssist as a system is listening and it's transcribing to English. It now understands what's being asked. It then needs to be instructed and taught where to go to get answers and then bring back those answers to the agent. So my brand-new agent can be an expert and not have to put you on hold and go search everywhere for knowledge and information. So your question about adoption rate, the criteria by which people are making platform decisions to go with Five9 is 2 things. One, they know in order to leverage this technology, they need to be in the cloud. So that's kind of step 1. And so we're seeing more and more interest. We believe this can be an accelerant to these organizations moving to the cloud because they got to go through a whole process of evaluating, making a decision and then transitioning on to the platform before they can even take advantage of the AI and automation. So AgentAssist is a great, great example with lots of criteria. It is being adopted faster than ever. But we also have companies like that large health care conglomerate that said, "Look, it's going to take us close to 2 years to get all of our 30,000 agents globally onto your platform. We'll implement the AI and automation then because we know it's going to change so rapidly and advance so much between now and 2 years from now. So we have a lot of decisions being made, probably 80% to 90% of the decisions that are made to go Five9 onto our cloud are made with the criteria of AI and automation and where we're headed with it, but they don't necessarily implement it immediately because they know it's rapidly changing and they get on to the platform. So it's a great criteria, but not a needle mover on the revenue front yet.

Okay. Thank you for adding that, Dan. It takes a while for this to permeate installed base.

You're not reading guidance yet. Yes.

Okay.

And on the IVA question, so yes, we're integrating LLMs across many components of our product suite. We start with the low effort -- the easiest high-value first, which was swapping out the prior engine for AgentAssist. We're wrapping up development on the integration of LLMs into our IVA platform, which was designed to allow for flexible integration with these different engines, literally get drop-down menus of pick your Google, your [indiscernible] there's many different vendors in this space for different pieces. So we're adding the LLM support. And what that's going to do is that's going to give sort of customers a way to quickly take advantage of this technology to improve the accuracy of these bots and to reduce the time to deploy them while still giving them like the safety of control that they know it's not going to like, say, crazy things, which is one of the risks you have if you just sort of give it the raw ChatGPT. So that's coming in with a long road map of technology, assuming we get our R&D budget increases there, Barry, we can...

Well, it's actually -- with co-pilot you can do twice as much with...

Indeed. Well, that might be a separate question internally, what the opportunity looks like, I guess, from an efficiency standpoint. But another big question I think we get from investors, just trying to understand the use of the customer data. Do you have access to the customer data. You have all these voice records, and is that more important than ever? Is it less important with LLMs. And how do you kind of -- I guess what the question is, how do you marry the power of the LLMs with the customer data to ensure that you're delivering the right answers, right? I mean, your customers want to make sure they're end customers get the...

Yes. And this is important, too, is I think this is something that's changed in the technology landscape that's not well understood. I think -- if anyone have been sort of working on AI, you probably got it in your brain, even if you're not technical, you need data to build AI and I got to train models with lots of data, right? You all probably knew that already. With large language models, that's not really true anymore. What's happened is that we've got these new super powerful models like GPT-3, ChatGPT and it's competitive. They required enormous amounts of data to be trained. Once trained, they are capable of understanding vast array of human language without being further trained on specific use cases. So now -- but okay, but it still needs to be sort of customized, like Dan gave this example, like when I talk to a chatbot, like, I want to ask my account balance, like if I log in to ChatGPT and ask what's my account balance, It's like I don't know what you're talking about. So how do you tune it to those business use cases is through a new discipline that's called prompt engineering. And this is going to be the thing that you're all going to be hearing about endlessly probably over the coming years. And what happens in prompt engineering is you take all of the data, but it's not historical recordings or calls or chats. It's those APIs, it's those knowledge bases. It's those articles. It's that information that the chat bot wouldn't use to answer the question. You take all of that information and you like feed it into the large language model. And so to do that, it's all about integrations. It's about getting access to these data sources, plugging into the APIs, having the ability to perform those automations take all that data you feed into the large language model, along with the customer question and you ask it. The way to understand it is, let's say, I'll give you an analogy, like, let's say you wanted to ask a question about like please summarize a book like a Jane Austen Sense and Sensibility novel in 2 senses. And let's say the large language model didn't know about that book already, which it does. But let's say it didn't. How would you do it? Well, a human being would read the book and answer the question. So that's what we do with prompt engineering. The prompt that -- the LLM reads the book, we feed it in and then we feeded the question and then it answers question. That's what we do in the Contact Center. We feed it in all the knowledge of the contact center that's unique. We feed it in the APIs it has access to, and then we ask the question, and we use it that way. So the discipline changes and the winner is the vendors that have the depth of integrations into all the systems, APIs and data that define the Contact Center. That goes to the platform plays, which we are.

And that, by the way, has not changed because that data changes dynamically every day. All the information about you and your balance those integrations are key to the platform. What's become enormously efficient as we've taken the ability to understand spoken English, and we don't have to train that anymore. It now understands how to understand what's being asked. So we've just sped through that part of the process, and we can now apply our expertise, which is how to go find the answers and how to get them back to the consumer.

You should talk about how easy to use to switch on the product.

So for -- especially for summarization, yes. So again, when we applied it for agents to summarization because in the prior generation, we had to predefine every intent, like every question or comment that an agent or a customer would make that we want to have a note in the summary. Like so if the customer called up and said, yes, I'm unhappy with my product. We'd have to have an intent called customer unhappy. And like we have to spell out the 50 ways you could ask that and train it to recognize that, right? That was what you had to do before. Now we don't have to do that anymore. So we just turn it on and say please summarize this. So the configuration went from like usually a month and a 2-month sort of implementation process where our services team would go in and understand what might be spoken in the conversations and what are the 75 things you want to put in the summary and train the model and took that time. Now it's like, click, there you go. One click, turn on summarization and it's there. So time to revenue for these products that are based on these new large language models dramatically less.

So I guess what I'm hearing in part is it's the actual customer data at least the voice records in some respects are maybe less and as long as you have the API integrations into the CRM platforms, your customers and whatever the records you need access to. Does that -- if that's the key integration, does that create more competition with like the CRMs? Like Salesforce which has some functionality or others? I mean, how do you think about that competitive landscape and how you stay ahead of the competition.

Yes, I think in many ways, I think it's less because it puts the focus back on the true moat and the differentiator is this customer interaction platform. The differentiator isn't the LLMs. We all have access to the LLMs. Every system, every company's platform is going to be able to understand what's being asked. The key is going and getting it and fetching it, bringing it back in the most precise manner. And then it's giving all the information back to the business about what's happened with all your customer interactions across all these different channels. So I need to have the reporting and analytics that tell me when customers take these journey and ask these questions, here's the outcomes versus when they take this other path, which channel is more effective than the other. So we can get back to looking at contact center efficiency and effectiveness as opposed to focusing on using massive amounts of recordings, spending months and months training the system, how to understand spoken English. Now we can go back to focusing on how do we deliver a better experience. So you need to have a full end-to-end platform to do that. We saw companies try to enter this space back in 2017, 2018 and failed and have exited, and it's because of that moat of the platform. And it's not just information in, fetch data, bring it back to the customer. It's how do I do that on a global basis for this parcel delivery service company that's in, I don't know, 152 countries, I believe, and deliver a virtual solution globally for all of their agents that are highly distributed all over the world and be able to make on the fly those decisions. Oh, do they need an agent? Can we self-serve and give them information with the system? Do they prefer to talk to an agent. Let the customer have the choice. So there's lots of complexity that goes into the platform that is not only still there, but I think will be a reemphasis on that.

And the channels is the key one, I would say the channels form our competitive moat against point solutions, against the CRMs and other people. And when I mean channels in our industry, that means the phone call, the SMS, the e-mail, the web chat, the web page, the WhatsApp, the Facebook Messenger, all these things are channels that are ways you connect to the consumer. And when brands or our customers want to deploy these products, they don't want a solution for each channel independently. They build a platform that can receive all of those things, and then they build those integrations once, they build the routing once is what our platform does, and then you can plug it into all these things. And the barrier to entry to support all of those, especially voice and SMS, oh, man, this is really, really hard stuff. I mean we have a global telecommunications presence, phone numbers in all over the world. You have to deal with regulatory, quality of work, call routing. I mean, I can go on and on for an hour on all the things the platform has to do to deal with the complexity of voice and SMS global interconnectivity, you can imagine it's hard. And so we have that platform. All the channels come into us, and that forms the moat that has been always and will continue to provide competitive strength...

And you do want to see a quick demonstration and see Jonathan highlight the value of that true platform and walk through it simplistically. If you go to our website under the IR section, you can see a webinar that Jonathan helped describe that.

Okay. I'm going to try to get 2 separate questions, and this has all been super helpful on the AI front. Maybe Dan, parcel company, health care, maybe just update us, how are those going? And what's that driving in terms of large deal pipeline?

Great question. They're going extremely well, right, on track from a rollout perspective, the large parcel delivery service should be completed by the end of this year. And the healthcare conglomerate is 12 different companies, big brands that we know that are owned and operated by the larger business. Those are -- think of them as almost 12 separate projects, and that's rolling out throughout the rest of this year and into 2024 as well and those are going very well. We believe that puts us in the pole position for the large enterprises to get confidence that we can deliver. They've been very, very supportive and referenceable for us in that front. And it's caused our pipeline to increase to more than double year-over-year from this point last year to now in the large enterprise space. So companies are recognizing, CaaS is ready for us. It can scale and it can deliver the innovation platform that they're looking for to revamp their customer experience.

Okay. And maybe Barry, moving into our last minute. And there is going to be a breakout session for anyone who has to follow up with anybody here, one of the rooms outside here. Financially guidance, second half of the year, how are you thinking about puts and takes? Where could there be sources of upside? Where are the potential risks?

Thanks, Will. So when you think about Five9, Think about it in terms of revenue growth in 2 buckets, approximately equal. The 1 bucket is new logos, they're coming on to our platform. They're going live, generating revenue. The other one is from the installed base, the other I have, which we sell more seats and more revenue per seat. On the first one, we have really good visibility, not just what Dan just talked about, what the parcel delivery service and the health care company is going to be doing but more -- actually bigger than those 2 combined are the orders that were born in the second half of last year and the first few months of this year, which will roll out the rest of this year and what you're bringing in now, we're going for 2024. Where we have less visibility is on the installed base side, where we have had the effects of the crosscurrents in the macro economy in our top -- we track 17 verticals, the top 3 in order of health care, financial services and consumer. And at different points in time, I don't have time to go into all of it. They've been buffered one way or another. What we've assumed is that the macro economy will stay roughly what it was last year. So nothing -- no major leg up, no major leg down, but we do have less visibility on that side of the business.

Okay. Yes, we're actually out of time. So we're going to need to wrap it there. Please join me in thanking the team here for their comments.