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

And -- yes -- I mean it's one thing after another from popping the screen back in the '90s, which will save 10 seconds to routing more effectively and efficiently. That's been the goal all along is to continue to reduce agents, but there's the technology, they have to pay for the technology in order to do so. The benefit to the customer is they save probably a $4,000 per human labor rate and replace it with a $400 self-service solution. So we think it's a great ROI and customers are very eager to take advantage.

I'm glad you put that last one because I was going to push you to go that. Why the heck would a customers say, I'm really jazzed to pay $400 instead of $200 on the last 2. It's the ROI and savings that's there. The other thing that I've started to talk to investors about is I think some -- they've lost sight of, you're not a contact agent software company, you're contact center infrastructure software, meaning that, yes, could be human, could be automated. It needs to be the full breadth of solutions.

Interaction management. We're managing every interaction. Human, self-help, all sort -- it's all sorts. And some of it's chatbots and e-mail auto response or human response. Again, regardless of whether there's humans or automation involved, it's a combination. It's always going to be a combination. And the more we can infuse the solutions with automation, the more efficient and cost effective it's going to be for companies to run their, call it, a contact center or call it just handling their customers' interactions, handling their communication to their customers. The more we can help them automate that, the more it's going to save them because the technology will be less than the human manual labor. And that's the -- that's where we look at it and say the scales are tipping towards automation. Halleluiah, it's about time. We've been looking for this moment.

The other element that -- and I think this is a legitimate pushback is, okay, we go that route, what's going to maintain or expand that $400 on the automation, won't it commoditize at some point as we shift more to the automation side?

Yes. Yes. We're finding -- actually, our ARPU goes up because you add automation to do self-service, like with an IVA at $400, guess what? We can apply summaries to -- and I would argue that it's perhaps even more important to have a summary of the conversation that took place between the consumer and the automated system because there are no human witness there. I want to be able to have a summary of that call and make sure that, that's put in. So you start adding in other automation solutions and the numbers just get great if not less.

I think the thing to stress here too is this idea of bringing engines in to understand human language and automate some sort of interaction. Five9 has been doing that for years and the company that I worked before that many more years again. The price of the virtual agent has continued to rise over time. Because it is literally looked at as an alternative to what the human agent would cost. So even if the engine prices are becoming commoditized, we have not seen over the last 10 years a reduction in the effective automation solution price in the market. And so that's why we're able to offer very high-margin IVA solutions today because it's the overall package, and it's the overall platform that people are actually investing in. They're expecting the underlying engines to change and improve and we switched in and out over time. I mean just to give you an example, in a simple IVA interaction today, it would not be uncommon to be consuming 3, 4, 5 different engines from different vendors, all within one conversation. You're switching in and out as you need them. So yes, this is just another example of a need class of engine.

So I think a key part of this is the rate and pace in which we see the deflection from a traditional human agent to some sort of automation. Where are we in terms of that state generally? And then where are the outliers in terms of what are the high end...

I'll start and let Callan take over. But if you think about that, I've been in the contact center industry for over 30 years and it's quite interesting. We are at an inflection point where this technology is being introduced for the very first time. But we're just getting started with it when you think -- and as Callan mentioned, we've got years and years of introducing it and really creating it. But I'd ask all of you, when was the last time you called into a company and fully self-serviced speaking to an automated virtual agent? Perhaps never. Not yet anyway. So introducing the technology to companies, the big question out there is, what percentage of calls are going to get deflected? I talk to customers every day, and they ask me, well, how many of my calls are going to be able to deflect and it completely depends. And it depends on a lot of variables. What's the current service you're offering? Is it very simple to fetch an answer and feed it back to the customer? Does it require any human consultation at all? What's the demographic of the people calling in for this service? And do you normally make them wait 20 minutes for an agent for a human agent? Because if you think about it, I'm going to want to opt for self-service, if some of those things are true. I may be apt to want to use self-service when I'm in one environment versus another, right? I may go to the website for self-service. If you think about self-service offerings and automation, we've been at this for years. We all go to a website or an app now to make our airline reservations to make our hotel reservations. That's all self-service now. But guess what? The airlines and hotels have massive contact centers. Because when I get stuck and I'm in Houston, and I need to get to San Francisco and my fight got canceled, I'm going to -- hey, how do I get there? What are my options? Is there any first class seats left here? There's lots of back-and-forth consultation. That's very difficult to do with automation. So there's a place for it, but there's also -- the human, I don't think will ever go away. But I think there's a lot more human interaction that we can certainly enhance, we have what's called Agent assist, where we can be eavesdropping on the conversation, looking up suggestions and coaching the agent to be more effective, but still have the human on. And that is where I see the most impact is to helping the interaction and helping the human. Most of the self-service stuff, a lot of it we've already moved over to self-service. People say, "Oh, password reset." I haven't called a contact center and said, I need to reset my password, I do it online. So a lot of stuff has already been done. Smartphones and websites have taken care of a lot of that. We're looking for the next lowest hanging fruit to automate in the contact center and that will happen. But a lot of customers find that, "Gosh, if I can deflect 3% to 5% over to automation, that's huge." Depending on the number of agents, again, that labor savings can be significant and oftentimes pay for the whole spend with Five9 just by having a small percentage move over. Now people say, oh, 30%, 50% someday. I'd argue twofold. One is perhaps if it's very simple questions coming back. But think about all the time when you contact and I use this example, we have a customer select quote, as an example, you see their ads on TV. We Shop, You Save they're going to find gene insurance policy. When you call that number, they have 4,000 agents every day, all day, selling these policies. When you talk to them, guess what, you're getting a salesperson. Well, they're not going to automate that. And also a lot of servicing companies. You call in, they take care of your service problem. Hey, you're eligible for an upgrade. They want to upsell and cross-sell you more services, really hard to do that with automation.

Yes. I would just say you really just think about it as any customer contact care workload, you can divide between conversations and transactions. Anything that's transactional in nature is going to be a good candidate for automation. And any part of a conversation that is transactional in nature is a good candidate for automation. So if I am paying something automated, if I'm tracking something automated. But if I'm engaging in some sort of back end fund, refund or customer complaint, you absolutely want those interactions being handled by people. And so it's really about identifying which customer care workloads that you have should be sent here and which should be sent there. And in fact, one of our -- one of the Five9 platform products called AI insights does exactly that. It looks at all of the interactions coming into your contact center. It works out the reason for why those people were interacting with you and then suggest to you, this one is a good candidate for an automation. That one is a good candidate for automation. So.

I would actually wonder, I thought in a number of industries, this isn't new. IBM Watson has been used in a couple of industries extensively. And it certainly hasn't seemed to impact your growth.

No. Well, I think the other thing is that cumulative customer care workloads, so the number of simultaneous interactions for the industry continues to grow year-on-year. So despite the fact that Watson has been around for well over 5 years, despite the fact that Amazon Lex has been out for years, Dialogflow from Google has been out for years and we use all of these engines. The overall interaction volume just continues to grow year-on-year, not necessarily voice traffic, but when you look at all of the channels interaction, they just grow.

Which kind of brings up the question, all right, is there a risk that the AI portion could somehow subsume some portion of the traffic away from Five9 and you're left with, let's say, a lower value set of traffic.

Yes. I think it comes back to why do people buy something like Five9 in the first place. So you're buying -- when you buy Five9, you're buying into a platform, I think it was mentioned earlier, the reason why people are having to get off of, say, premise-based legacy systems is that they're looking at systems that don't get upgraded every year or every 2 years or every 3 years, we push features and functionality into the cloud several times per day. right? So when you buy into Five9, you're investing in a platform that essentially does all of this end to end. So one of the questions I would have for anybody considering, let's say, a point solution like just a chatbot out there that you could just use and maybe that takes some interactions away. You could do it there or you could do it on the Five9 platform. Why would you do it on the Five9 platform? Well, because all of the reasons you come to a platform in the first place, the integrated reporting, the service availability, the language support, the professional services support, the tech support, if you ever need anybody. All of those things -- security compliance and so on, yes.

So the other element that is starting to bubble up in more of my conversations is the compressing of innovation cycles that -- if we think about the use of AI in code development, for example, does this actually shrink the moat and allow companies to come in and compete in your space that either weren't there before? Or does it somehow level the playing field across the competition?

There's always going to be these point solutions, as Callan mentioned, around the periphery, and there have been for years and years. If you take the example of chat and e-mail for a while they're -- 10 years ago, folks were out coming up with stand-alone chat solutions, stand-alone e-mail solutions and companies would take them and then try to bolt them on to the contact center platform. The difficulty there and now we have chat and e-mail that gets sold with just about every system that we provide that we sell to our customers. And the beauty of having it be part of the same system is I want to create a very consistent experience regardless of how a consumer has reached out to interact with me. I'm going to want to integrate and retrieve the same data whether it's the agent retrieving the data, whether it's the chatbot retrieving the data, whether it's the e-mail system retrieving the data or the IVA as we now have. We want to be able to retrieve that data and give it back to the customer in a consistent fashion. So we don't have to build multiple integrations to those back-office data systems. It's just one from the platform itself. In the last earnings call, we used the analogy of the airplane versus the jet engine. These jet engines, as Callan referred to, are going to be continuously upgraded and changed very rapidly. And we can insert new ones and enhance what we offer. But to go build the airplane, the -- going off of these legacy systems onto the cloud, that's a huge moat. It's really all -- for all the reasons we just said. But to try to build that airplane and get it off the ground and make it fly, you don't see a whole lot of new entrants. You don't see virtually no new entrants into our space in the last 10 years. Because the companies that have tried that. It's been a long multiyear effort to be able to build all the capabilities and not only the thousands of features and functions that are -- and the compliance and the security and everything that goes along with that. But when you build a feature for a multi-tenant platform that's going to be hosted -- hosting it for hundreds or thousands of companies, every company wants that one feature the way that they want to implement it. So you got to build each feature 6 different ways because they all wanted to work a little different. So it's more complex than would meet the eye.

Yes. I mean it's a big, big decision like Korea limiting decision to get customer care wrong in a big organization, right? Like you've got to make that decision very, very carefully. You may be enamored by a point solution. And in fact, I built a company which was a point solution. And what happened, it grew and grew and finally bought it, right? So because it was seen as, oh, that's a fundamental part of the modern airline, if you like, or jet airline. So that tends to be what you see happening. You have a lot of entrants that are trying different things. But for a big company or even a midsized company to take a risk on something like that is done very, very early. So they inevitably go to a trusted platform provider to do it.

Yes. And inevitably, when those point solutions do make traction and they are best of breed and unique, they get scooped up by companies like us because they're great technology tuck-ins.

Starting to get the impression on how we feel about this.

Yes. Just a little passion there. One of the key elements that I'm a big believer in that's going to separate the winners and losers within AI is access to data sets. I think couple of weeks ago, CEO of Reddit said, "Listen, if you want to scrape Reddit to train your LLMs, we want to get paid." And I think this is just the very beginning of companies realizing -- further realizing the value of their data sets, and I think they're going to close their arms around it and make it even more difficult to get access to talk to investors about what is that data set opportunity at Five9?

Yes. So actually -- there's actually a fair bit of misunderstanding about how this sort of data is used in an LLM world or even a generation 2 world. If you go back to something like -- sorry, just the microphone. If you go back to, let's say, just let's go 12 months ago to something like a Google Dialogflow, Amazon Lex based, NLP-based solution, let's call that Generation 2. It was all about having access to as much data as possible because every single intention from every single interaction was actually programmed in manually and then trained on many examples of someone saying, "I'd like to track a parcel, I'd like to change my address" or whatever it might be. And so data was absolutely king. With an LLM basically, you're saying, well, it's pre-trained. So essentially understands kind of everything about language and can -- or attempt to answer pretty much any question. So in that sense, LLM sort of resets some aspects of that data collection playing field. However, the most important data is that temporal data. The one that's -- that piece of information you have right now because you're on the phone with somebody or you're in a chat with somebody or whatever. That's the hardest data to get access to. Five9 is the aggregation point for that data. And so not only the aggregation point for that data right now, but all the historical examples of that over time could be many millions and millions of interactions for any given company, for example. So LLM allows us to make really good decisions on the very small amounts of data, but very important amounts of data is like stuff happening right now, but it also allows us to draw insight from all the data that we've collected over time as well. But that's -- I would encourage people to think of that as separate to building, let's say, generation 2 language models, which is sort of what we're going to see fade away gradually.

I guess my thought was the training aspect of it in a vertical application, the optimization for intent within a company, for example, and then being able to take your expertise because of the millions and billions of minutes of interactions that you have and being able to apply that to a customer data set?

Yes. So I mean, where you do see a lot of that happening is you see it in the way that dialogues are actually designed in general. So if you think about what is the secret sauce of having done this lots of times, the next time you do it, you do it very efficiently and it works really well for a customer. So if you look, for example, at Five9, we have been developing out verticalized tasks for automation for various industries. Health care is an example. So if you want to come in and do appointment booking, you don't have to invent the wheel, but if you come to the Five9 platform, there's literally a template there that says, this is how we suggest you do something like that. And it's built based on lots and lots and lots of experience across lots and lots of companies doing this. And the types of phrasing that we use and the way we respond to you and the interfaces we present to you are all kind of well thought out, and that's another reason why you would come to a platform provider as opposed to building something from scratch. Yes.

Switching gears a little bit, but still under the broader AI umbrella topic. Are there other areas within the business that you're using AI yourselves? So thinking about marketing and sales, thinking about content generation, thinking about, well, I leave it open ended. Is it starting to sprinkle into the business?

Yes. I'll speak on the sales front, absolutely. We have several tools that we use today that allow us to understand what's happening among -- I probably shouldn't say too much among prospects and companies out there and being able to intelligently know what they're searching for and what they're looking at and who they're speaking with. It's remarkable how much insight you can gain that is out there in the market.

I mean on the product development side, we obviously use tools that allow us to look at sort of real-time interaction data. We don't necessarily build those tools. Those tools are becoming embedded with AI, of course. In fact, every industry is -- and just as we look as to where we spend our dollars on the various pieces of tooling, we're looking for somebody that has a strategy that incorporates AI just like the same customers that come to us and look for that as well.

And then, Barry, last question on the topic. As you think just conceptually, how does all the AI items that we just talked about influenced the financial model longer term? So I'm thinking about -- does it have a gross margin implication? Does it have -- well, we kind of got to the revenue implications if ARPUs are going up. But what are your thoughts there?

Yes. So just to emphasize on the revenue side, the way we look at it, we currently -- it's $200 to get on the platform just to get that economy ticket. You want first class, you're going to pay a bit more. In terms of gross margins, it varies, and we're still evolving in some of the products. Some of these are still recent. But on the IVAs, which is the more established one, when we bought Inference, the margins were sort of in the 80s, they continue in the 80s. And we have every reason to expect the way Callan talked about it earlier, the fact that these things -- the cost is becoming commoditized. Five9 has always had the policy unlike some other companies of not -- we're not smart enough to take on the hyperscalers for things like automatic speech recognition, text to speech, NLP in the past, now LLM. That's just one small component. And the drive to get that cost down is driven by the fact that a lot of it is used for search where there's no income coming in. So it's going to go down in price. We are charging for -- so we're pretty optimistic on the gross margin side.

That's great. And then we've got a couple of minutes left. I'd actually like to dive into Dan, you guys have executed well in terms of delivering on the quarterly expectations. From a sales execution viewpoint, what elements, what levers have you pulled as the economy has changed over this last year. So I'm thinking about pipeline coverage, I'm thinking about territory allocation, sales quotas, et cetera. What are the levers that have allowed you to be consistent in what you've delivered?

Yes. We look extremely closely at all the different segments of the business to make sure that we have a full coverage on the areas that we want to cover and then the areas we see that have the most growth coming. So as an example, we have sales teams that are segmented from all the way down in the small- and medium-sized businesses or commercial business, as we call it, to mid-market teams, enterprise teams and then strategic. The large enterprise and strategic teams have been growing significantly. Those are the ones that are accelerating. We're seeing pipeline growth. We've got to make sure not only that we have the right people -- the coverage, but the right people, the skilled folks that are used to selling into large Fortune 500 companies with very complex sales that have experience in contact center. And one of the beauties we've had is we've -- I've brought in a leadership team years ago that we're used to selling into the largest enterprises, some of these companies that we've referred to already. And so just be patient. We've got to wait for the market to get here and wait for the technologies to get here. It will happen. I remember sitting in a chair like this 5, 6 years ago, and we've had certain folks that have come into the business and said, "I don't have the patience. I don't want to wait 5 years for large enterprise." And they didn't want to do the transactions at the smaller and medium size. And lo and behold, here we are. And a lot of those folks are thankful that they waited, but they've got the skill set. So aligning -- the biggest thing I look for is making sure that we have enough coverage, but then making sure that we align the right skill set with the right opportunities. And that not only is salespeople, but it's all the supporting teams around them. So it's the solution consultants. It's -- we have strategic managers that handle some of these large, large accounts that really herd the cats and build the team around how to go execute against some of these large, large health care and financial institutions. Some of those, we have a dozen people focused on a single opportunity because it's tens of millions of dollars to us.

Makes sense. With that, Barry, Dan, Callan, really appreciate it. This was a fun conversation. Thanks again.

Thanks.

Thank you.