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

Okay. Let's get started. Thanks for joining us today. As a brief introduction, I'm Ryan MacWilliams, Small and Mid-cap Software Analyst here at Barclays. Excited to have this AI in the context in our fireside with the Five9 team today. I really appreciate their help in setting up this event. Now we have a full agenda, so we're going to dive right in. With me today is CEO and Chairman, Mike Burkland; CFO, Barry Zwarenstein; President and Chief Revenue Officer, Dan Burkland; along with some special guests and Five9 attendees. From an agenda perspective, over the next hour, after starting with the Safe Harbor statement, we'll start with questions to Mike and some brief remarks on AI and the contact center. Following that, there'll be a 15-minute AI deep dive with Five9's product leaders, a 20-minute customer Q&A and then a 20-minute Q&A with the Five9 team. Now we do have a packed agenda with multiple participants. We're going to be passing the ball around a lot here, but please feel free to send questions into [email protected], and we'll try to get those in, if possible. So let's kick things off with the Safe Harbor from Barry.

Thank you, Ryan, and hello, everybody. Before we start, I'd like to remind you that certain statements made during today's discussion are not historical facts, including those regarding future events, [ transactions ], projections and beliefs that may affect our industry or our company, prior developments, AI and automation, and potential growth drivers are all forward-looking statements. Such statements or predictions should not be unduly relied upon by investors. Actual events or results may differ materially, and Five9 undertakes no obligation to update such information in such statements. Please refer to our most recent Forms 10-K and 10-Q under the caption Risk Factors and elsewhere with our filings with the SEC. Thanks, Ryan.

Yes, of course. It's just so great to have some of the members of the Five9 team and multiple different customers, and we're going to be passing the ball a lot around today. So for investors on the line, if we have some like technical difficulties or we're doing some screen sharing, just bear with us, but we got a lot packed into the schedule.

Mike, I've been dying to ask you. We've seen a lot of changes in the contact center over the years with the introduction of IVR to the widespread adoption of Internet and e-mail. How do you view the evolution of the contact center with large language models in AI? And why do you think Five9 is in a great position to help clients utilize this new technology?

Yes. Thanks, Ryan, and thanks, everyone, for joining us today. Today, we have an opportunity to have a meaningful and important conversation about the realities of AI and automation. And how recent developments, such as LLM and generative AI, will likely impact our market and our company. So over the next hour or so, we're going to cover a lot of material, and we expect that you will walk away after that hour with the following 2 conclusions. Number one, generative AI is the next wave of opportunity for Five9. It broadens our TAM. And number two, platforms will win on point solutions and engines alone. The platform is the control point, and Five9 is the platform. So let me take each of those 2 one at a time for a second here. The first, generative AI is the next wave of opportunity for Five9, broadening our TAM. Five9 has been riding the wave of AI and automation for the past several years, and as we are positioned very well to continue to push this industry forward. Not only is the AI revolution a tailwind to our technology and innovation, but it's a tailwind for our business model. We provide software for our enterprise clients to manage all of their customer interactions, from 100% human interactions on one end of the spectrum to full self-service automated on the other end of the spectrum, and a blend in between. So let me remind you that as AI drives efficiency and productivity gains in the form of a mix shift toward more automation over time, that leads directly to an increase in revenue per customer and a TAM expansion for Five9. Let's take the second conclusion here. Point solutions and new technology engines can accelerate and enhance a platform, but they cannot replace it. By now, most of you have heard, we've made this analogy between the airplane and the jet engine. Think of Five9 as the airplane and LLMs as the jet engine. To fly across the country, you need the entire plane and all of the systems. LLMs are like the jet engine. They enable our airplane to fly faster and further, but they are not replacement for the airplane. Because Five9 is an end-to-end platform where all customer interactions run through our platform, we act as the control point, which is quite different from a point solution such as a stand-alone IVA or a chatbot. So Jonathan and Callan are going to discuss this in much more detail. So with that, let me turn it over to Jonathan Rosenberg, our CTO and Head of AI.

Hey, everybody. Thank you for spending your time with me today. I'm super excited to share with you my absolute most favorite topic, which is to talk about AI. And whait I really want to do is cover that in 3 parts. First, talk about our AI strategy; second, talk about Five9 as a complete CX platform. For that, I'll get help from Callan, our Head of Product Management; and lastly, I'll come back to me and I'll talk about how that platform forms a competitive moat for us. So let's dive right in and talk about Five9's AI strategy. And to do that, you need to have some context here. Over the last 12 years, it's not a very long time, we have seen a huge change in the landscape for AI in the enterprise. It started with what we call a pre-deep learning era. And in this era, which ended just about in 2011, it was pretty challenging to build AI applications for the enterprise. You needed to go and build custom AI models and have tons of training data and build and tune those models, and it was expensive and time consuming. In this era, there was a certain set of vendors. Nuance, for example, is one of them, that really were amongst the leaders in this space. And then we saw a change. Deep learning arrived around 2011, and this introduced the era of like Alexa and Siri, and tools that dramatically arrived in the consumer landscape and also made their way into enterprise. And this era became much easier and cheaper to do speech recognition and language speech generation, but still required customization for the use cases for each business one at a time. And that did involve model training and data collection. It was easier. It was a little less expensive, but it was still required customization. Now in this era, we had a different set of vendors who did really well. So Google and Amazon and IBM, for example, were some of the leaders in this era. Now once again, we've entered another new era. This time, the era of the large language model, which just started last year. And in this era, things have completely changed. Now instead of about collecting lots and lots of training data and labeling data in order to build unique solutions, there's a general purpose solution which can be made fit to purpose through what's called prompt engineering, and I'll talk more about that in a moment. And in this era, we have new vendors. So OpenAI, of course, is on everyone's mind. Lots of people are rushing to compete like Google and others, and we'll see who the leaders are in this space. But the most important thing to observe is that there's one thing that's been constant here, which is change. Every few years, we've seen new technologies, new vendors, new leaders, each of which have unique capabilities. And that led to our AI strategy, which was instead of building these engines in-house, we take them from whomever is best in the market at that point in time and integrate them to build a complete product, right? And in fact, we will use multiple vendors often even for the same customer. We have customers who deploy our IVA and they'll use one vendor for text-to-speech and a different vendor for speech recognition. And this agility has allowed us to ride the cost curves down and the quality and technical innovations as these mega scale vendors deliver amazing technologies. We then built our entire architecture to be prepared to swap them out as tech evolves. And this was a vision and an architectural design principle for both Five9 as well as Inference, which is the IVA vendor that we bought some years ago, and that has allowed us to be agile and quickly move and adapt to these different technologies. And that allowed us to execute the third pillar of our strategy, was to focus on building business outcomes, right? At the end of the day, no enterprise wants speech recognition or natural language processing. What they want is to have a contact center that can receive customer inquiries and handle them quickly and efficiently, leaving satisfied customers at the lowest cost possible. That's quite a different thing, and it requires building the whole product of which the language models and other technologies are just a component. In fact, building that whole product solution, a great way to think of it is an analogy that Mike just shared with you, which is that of an airplane. If you want to go across country, you're not going to just strap yourself to the bottom -- on top of a jet engine and take off. I don't recommend that. You have to get in an airplane. And so the airplane provides all the systems that are necessary, the navigation, the wings, in addition to the human pilots and systems for loading the passengers on the plane and off the plane. And all of this is necessary to ultimately deliver what the customer wants, which is to get from point A to point C. The jet engine is the enabler of that. Now this becomes even more crucial in this new generation of technology around large language models. In this generation of technology, the previous things you did, which was building lots of data and labeling it and building train models, that's gone now. And it's replaced with a new technique for making these technologies fit for purpose known as prompt engineering. This is probably something you've heard of by now. And what happens in prompt engineering is you customize it by taking different sources of information and putting them into the prompt. So for example, if you wanted to build a system that was capable of answering customer inquiries about insurance on a website, you have to build an overall product, a system that took the customer data, that took curated knowledge, that took the customer inquiry and integrate it into the website and use that to build a prompt, which was then fed into the large language model that, in the end, produces the result that goes back on the website to the customer. Again, with this new generation of technology, vendors who in the past have staked their futures on lots of labeled data are going to find themselves really without an edge in this next generation because now, it's not about that anymore. It's about integrations and connectivity with lots of systems that are required to build the prompts that are the core discipline in the usage of large language models. So what I'd like to do now, with that background on our strategy, is talk about Five9's complete CX platform. And to do that, I'm going to hand it off to my colleague Callan Schebella, EVP of Product Management at Five9, also the former CEO of IVA Pioneer Inference. Callan, why don't you take this away?

Thank you, Jonathan, and I will just share my screen on this end. And just let me know if that is coming up. Excellent. Okay. So let's start by talking about what we mean by a complete CX platform. Basically, the platform has 3 major functions, and I'm just going to route here. First, we need to provide an aggregation point for customer contact. That is a means to basically take voice, chat, SMS, e-mail and more and basically land them on our platform. Secondly, we then route those interactions to agents or to automations and bring in core engine technologies that Jonathan mentioned as required to allow us to attach meaning to the interactions that are taking place. And third, we combine our understanding of those interactions with our integrations into data sources and systems to create a real-world outcome for the customer and the business, that are often referred to as fulfillment. Now this might sound pretty simple on the surface, but if we dive one level lower and bring up all the layers that comprise the Five9 CX platform, you'll see that there are a lot. And the way that I'm going to help illustrate this is to start with the simplest possible use case, and that is just a customer calling into a business and having their call handled by a live agent powered by Five9 software. So if we look at this diagram here, the customer starts on the left-hand side by making a phone call. You can actually just follow along with the highlighted elements on the screen there. And the phone call is then handled by our global voice network in both a secure and a compliant manner. And then lands on our core ACD, which is the switch or essentially the routing brain behind Five9. Now in this case, we then pass that to a live agent who has logged into the Five9 agent desktop. And that agent desktop is integrated into the business systems that Jonathan mentioned, in this case, a CRM system and a marketing platform as well as some sort of custom back-end integrations that are required for this particular company. Now once the call completes, the call can be automatically summarized for the agent. And to do this, this is an example of reaching out and consuming an engine. And here, we consume a large language model engine to do that summarization. And finally, we post that information along with everything else we know about the call into the Five9 data lake, and that then allows the business owner to use that information after the call to do reporting, to do analytics or to visualize it in an online dashboard. Okay. So this is a really simple example, and in fact, for Five9, this is about as simple as it gets. So now let's look at a more typical customer scenario using the same type of flow. So in this case, we're going to imagine a customer calling into a contact center to ask a question about an existing reservation that they have. You could imagine it could be a hotel or an airline, a restaurant, or really kind of any industry. Once again, we're going to start as a voice call and we're again going to land on our global voice network and again go to the Five9 ACD. But in this case, the call is answered by a virtual agent, so no human agent. It's 100% software. And the virtual agent reaches out and consumes a variety of engines, and Jonathan mentioned this as well. So here for example, we need a synthetic voice so the IVA can talk. We need biometrics so the IVA can identify the caller based on their own voice speech recognition engine to convert what they're saying into some text that we can process. And then finally, an NLP or natural language processing engine, to attach meaning to that text and respond appropriately. And it is not uncommon at all for that to be split across 4 different vendors, for example. Now the virtual agent also draws upon information stored in the CRM and the marketing system. And imagine in this particular scenario that the virtual agent has determined that the caller is eligible for a free upgrade and wants to transfer that call to a live agent to discuss. Now, one of the nice things about our platform is that the live agent benefits from all the information already captured by the virtual agent, and the live agent is also assisted with artificial intelligence themselves, this time from Five9 Agent Assist. Now the way that we do that is we use Five9 Voice Stream to stream the audio of the call to the engines attached to the platform, and that allows us to suggest guidance and recommendation for the agent to use while they talk to the customer. And again, we could do this in a variety of manners. We could use a natural language processing engine or a large language model to do this guidance. And then finally, again, a large language model is used to summarize the call. Now often, there's steps after the call, so in this case, we're going to use Five9 Workflow Automation. Again, another layer, in this case, to send a confirmation SMS to the customer as well as a thank you e-mail. And once again, everything is written into the Five9 data lake and then the business can query that through Five9 Reporting, Five9 Analytics and well as web-based dashboards. Now they may even seek to use things like speech analytics, which allows them to look at sort of customer sentiment using that engine there, highlighted on the right-hand side as well. So as you can see, there are a lot of layers in the Five9 platform. And this really demonstrates why our platform is so powerful. Everything you need all in one place, handling the most simple interactions all the way to very, very complex interactions. And with that, I'm going to pass back to Jonathan again.

Thank you, Callan. So with that, let me talk about how this forms a competitive moat. If you could just stop the share, Callan. Much appreciated. Thank you, sir. All righty. So how does this form a competitive moat for us? The reason really comes down to 2 core concepts, channel and agent ownership form a competitive moat that favors platform plays over point solutions. So let's understand that in some detail. Let's say a point solution comes along. They've got the world's greatest chatbot or voicebot. In order to do that, to have any sort of bot, you ultimately have to plug into a channel. You have to be able to receive the voice calls, receive the chats or the e-mails or the webs integrations. And you can imagine that the voicebot or chatbot might do that themselves. But historically, actually, what's usually happened, and I'll explain why, is these things don't plug directly in. Instead, they sit on top of exactly the type of platform that Five9 has that provides that connectivity and everything else. And the reason for that, again, is all of the capabilities that are inside of that platform that are necessary for every voicebot, every chatbot or a live agent, instead of providing that functionality one at a time for each of those, you provide it a single time as part of the platform. So again, if we go back to a specific use case, let's say, we've got a call that comes in here. And the first thing that happens is it arrives at our ACD, which is our routing engine. What should happen with this call? Should we send it to an agent? Should we send it to chatbot one? Should we send it to chatbot, two? Most companies will want some flexibility here. They wanted to say, maybe this is a high-value customer, a high net worth customer or a customer that based on data and integrations, looks like they have a really bad account balance and we want to send them right to a live agent. This kind of decision making is what our engine does, and you need that to sit in front of the voicebot or chatbot, even select whether to use it or to send a call to a live agent. Okay, let's say, the call then goes to a voicebot or chatbot. We still need much of the capability that Callan described, for example, around reporting. So one of the things customers want to know is what fraction of their interactions, whether it's chats or e-mails or SMSs, were able to be handled by the live agents versus which ones are able to be handled through these automations and bots? And the only way to get that kind of aggregate reporting is when the voicebot and chatbot sit on top of the platform, so the platform that can provide it. So we've got dashboarding. So let's say, there are supervisors that are still watching the system in real time, and they worry about things, oh, my gosh, there's too many people sitting in the voicebot or chatbot system where they're all hanging up. Something's wrong, maybe I should make some mid-day adjustments. That kind of real-time dashboarding is again provided by the platform that allows for the whole system to work as a whole. And then most importantly is the integration with data sources. This is the foundational part for any voicebot or chatbot. You can't do anything without access to data and APIs. And often in many of our enterprise customers, there are dozens of these integrations with all kinds of third-party systems in order for the overall contact center to work. And by having the platform provide those integrations and allowing voicebots and chatbots to plug into the VR platform, the overall solution is more reliable, more secure and more manageable for the customer. Again, this is why historically, for many good reasons, the market has preferred the voice and chatbots to plug on top of the platform. Even on top of that, you have the huge complexity of building the channels. Just the channels alone is a huge amount of work. It's deceptively simple. If you think you just drop in a few lines of JavaScript, you get a web chatbot, and web chatbots are actually one of the easier ones to do but still lots of work around customization and localization and flexibility. When you look at something like SMS, right, which is the way a lot of us like to communicate these days, there is a ton of work required just to build SMS integration as a vendor and a platform. You need to build up a huge inventory of numbers, you need regular numbers, you need them in different area codes and country codes, you need short codes also. And then to do that, you need global presence where you can -- and to get global presence all over the world, you need to interconnect with providers who offer those SMS numbers to you. And once you do that, you have to comply with regulations and rules around provider registries to prevent robotexting and malicious content, and that's a whole bunch of work and integrations. And then messages don't always get delivered, so you have to worry about delivery statistics and reporting and retry systems. And then there's these opt out things. You probably see you can type stop to discontinue, and that requires a whole pilot work in the platform to manage that and ensure compliance. And then you need language and personalization and building on top of all this. And that's nothing compared to voice, which is even more complicated with networks and telcos and all kinds of capabilities that I won't be able to get into. The complexity in building this is what makes the platform, and that's why voicebot and chatbot products typically plug in. So channel is #1. And then the second is the human, that most customers want some amount of agents in the system, whether it's more or less. Depending on the capabilities of the automation, they want the flexibility to route some calls to the human agent or go to the chatbot and then transfer it to a live agent in case there's challenges. And that transfer process can't be done effectively without being on top of the platform. The platform provides things like the transcript so far or the actions taken by the user when they were conversing with the voicebot or chatbot. All that is necessary for the overall experience to work. So when you put that together, in conclusion, these point solutions don't displace us. They require us. This has historically been in the case. We've got many, many third-party chatbots, voicebots that have historically run on top of our platform because -- and customers like them on top of the platform. Second, by owning the platform, it allows Five9 to offer differentiated capabilities to our own voicebots and chatbots that sit on top of the platform. And that's why it's often a customer preference to choose those products from the vendor of the platform. And third, owning the platform means that we can monetize those third parties when the customer chooses them, and that's another strategy that allows us to take advantage of our platform position in the network. So with that, thank you for your time. I'm going to hand it over to Dan Burkland, President of Five9. Dan?

Excellent. Thank you, Jonathan, and thank you, Ryan, and good afternoon, everyone. We thought it would be a great opportunity, rather than just hear it from Five9 on our technology, to have 3 Five9 customers share their experience with all of you. So I'm very pleased to be joined today by Joe DeLuca from Wyndham Hotels & Resorts; Ronald Benjamin from Chipotle; and Joe Friedrichsen from Blue Cross Blue Shield of Rhode Island. So thank you, gentlemen, for joining us today. I thought we could get started. If you can enable your videos, that would be super. I thought we could get started with some very basics and introduction. Why don't we start in the order I just described, and a quick background on you and your company and what you do there? And then we can get into how you're leveraging Five9 to help you with your customer experience. So first, Joe DeLuca, why don't you start us off a quick background on yourself?

Sure. Thanks, Dan. So Joe DeLuca, I'm the Senior Director for Enterprise Voice Contact Center Systems and Collaboration in Wyndham Hotels and Resorts. I'm with the firm for a little bit over 5 years, and responsible for all of the contact center technology and solutions for Wyndham.

Excellent. And Ronald?

Yes. I'm Ronald Benjamin, I'm the Senior Director for Guest and Employee Experience for Chipotle. I've been here for a little over -- almost 2.5 years, formerly with Delta Airlines, and I'm responsible for all contact centers for employees, guests and social media and -- for Chipotle.

Excellent. Thank you, Ronald. Glad to have you here. And Joe Friedrichsen, Blue Cross Blue Shield?

My name is Joe Friedrichsen. I work for Blue Cross and Blue Shield of Rhode Island. Been here about 7 years. I manage all the infrastructure that includes, in this case, the Contact Center as a Service. And prior to that, I was at Fidelity Investments for close to 20 years.

Excellent. Why don't we take the next question in reverse order. I'll start with you, Joe Friedrichsen of Blue Cross Blue Shield. Just describe, if you could, why you moved to the cloud, and in particular, why you chose Five9? And then we'll get into talking to each of you about how you've leveraged and how you're utilizing our AI and automation solutions.

About 3 years ago, we have decided to get out of the data center hosting business and move all of our technology into a cloud. Most of that moved into Microsoft, but again, we had to focus on what are we going to do for a contact center. We had 2 different technologies. Both of them were end of life, both of them would take significant upgrades to get to a supported version with no added value. So we looked at an RFP, looked for Contact Center as a Service. We knew we wanted to move to a SaaS model. We looked at companies like Genesys and NICE, Cisco and Five9, we did an RFP. I led that RFP along with some of the business teams. And we have decided that for us, Five9 was a perfect fit not only for the short term, but really, for the long term based off of the technology road map.

Okay. And Joe, while I've got you -- I know you've done some work with our AI and automation solutions. Maybe share that with the group on what that...

Yes. So earlier this year, we actually started with that interactive voice agent, and we actually rolled that out. We've rolled that out to our customers directly as well as the doctors who call us up when they have claims questions. We really looked at 2 different real use cases. One was the authentication or authenticating people through that natural language processing that John talked about, but also looked at how do we understand their intent of calling and then route them appropriately to either our own internal centers or to external providers for us. What we saw was in the first 3 months, we actually surpassed our 3-year goal of authentication rates, which was really remarkable for us to see that. And then from a containment perspective, again, like routing calls to the appropriate areas and routing them to external third parties, definitely saw that go much above where we were targeting. So our interactive voice agents really did a great job from a self-service perspective for us.

Great. Excellent. Good to hear. And over to you, Ronald. Tell us about your choice of Five9 and then how you've implemented AI and automation?

I have much of the same journey as Joe had earlier was we went through the exact same vendors because they are the best in the space. And we landed on Five9 because we feel like we could grow with you and you were growing, and you are a great solution and is tailored to what we need for what was going on. So it allowed me to have multiple vendors in multiple spaces because I have the employee experience, customer experience and social media. So all these significant areas that have multiple vendors using the same platform, it allowed me to leverage all those things and to exceed, much like Joe, the markets that I had in place for customer experience as well as average handle time and actually understanding what was going on in the business as well.

Good. And Joe DeLuca from Wyndham Hotels and Resort, you have a rather unique model in how you're leveraging the IVAs, the virtual agents. Maybe you can take us through not only your decision to move on to Five9, but then what you're doing today and kind of what your plans are moving forward?

Sure, sure. So I think the decision to move to Five9 was a pretty easy one. Did go through the whole process with an RFP, all the major players. Five9 checked all the boxes. And what was supporting our centers prior to Five9, I guess the best way to describe it was hybrid. So there was some on-prem in conjunction with a data center and managed services. Lots of operational complexity, to say the least. And the cost, of course. Obviously, that comes along with that was pretty extreme. And so Five9 allowed us and afford us the opportunity to streamline, reduce that complexity, and of course, the overall cost. And it made it much simpler. And there's a lot more to say about that, quite honestly. But if you speak to the IVA component, one of the first things that we started to do with the IVA was directly associate it with deflection and containment, right? So the low-hanging fruit. We had -- I know this had 40,000-plus calls per month coming in just to reset passwords for our loyalty members. And between the verification and validation of credentials and the entire process, it was pretty burdensome to have all that handled by an agent. So the IVA allowed us to automate that through an authentication process, right? Checked all the boxes as far as our compliance and the legal departments are concerned. And then we were able to move forward and really chip away at that overall volume on the password reset.

And I understand what you've discovered since is handling a centralized means through an IVA to now alleviate the load that you used to route out to the many, many properties and have the front desk clerks and all that labor force have to take mundane, repeated questions. What time's room service end? Is the bar open? Is there a swimming pool? All those kinds of things.

Yes, and that's the next level we've taken to. So we've gone from the likes of that low-hanging fruit with the password resets, really simple stuff to automate, to the next level where we're taking that sentiment analysis and intent analysis. The AI is delivering self-service, being able to answer calls along the lines of whether or not you're running a shuttle, what's your pet policy. A lot of opportunity from that perspective, and it's really taken a significant chunk out of the non-revenue generating calls going to our centers and ensuring that, again, they could be self-serviced or rather more according to the property to be handled. So that was a huge win for us.

Excellent. Excellent. Ryan, any questions from you or the audience that you'd like to ask to the panel?

Yes, for sure. Guys, thanks so much for being here today. I just want to say I love that breakdown by Jonathan and Callan because it just shows how complicated a contact center could be. I mean, you guys live this, but it's much more than just answering a question, right? And look, as a former contact center agent myself back in college, it does pain me to ask this question of you, which is, given this rise in AI, do you see a potential reduction in the number of agents in the future for contact center and over what time period? I love to hear your perspective about it.

I can jump in here. Really, at the end of the day, we don't see AI change -- eliminating our agents. When we look at it, we really look at it from an omnichannel perspective and what's that member's journey, right? Sometimes, they're on our website first and then they're calling us. Sometimes, they don't even go to the website and they call us. Really understanding that the AI can solve some of the basic kind of questions but then typically, our customers want to talk to a human being. The analogy I've always given is really around how when the Internet came out, everyone was thinking like retail banking is dead, right, that people don't want to go to a retail site. And what we found is people want to go to those retail sites not for the basic stuff, but for something that's very specific to them, very unique. And so -- and what they found and what we found the same thing is our Net Promoter Scores go up as we provide those great services, both from a self-service and then easily transitioning that to a human-assisted experience.

And if I could just add to that. I think that it's a misconception that an AI is just going to automatically remove jobs out of the equation. I think, quite honestly, that while there is opportunity for -- to make some kind of reduction, for the most part though, it really just allows the agent the opportunity to focus on other areas. It would be more productive, so it really optimizes the agent's time more so than replacing the agent outright.

Ryan, I'll go one step further to say, I think it will grow. And that's really being optimistic because it now allows us to focus on getting more customers through our venues. It allows us to do -- to work on the things that we really need that human touch to really work on. So in actuality, if we can increase the footprint that we have in our businesses, give a higher level of service, it actually opens the door for us to increase our market share and actually hire more people with better intel, better customer service to really take us all to the next level.

And building off of what you were just saying there, Ronald, we talk about agent happiness. And one of the things that every call center suffers from is the high turnover. What we see AI doing actually is helping us with that agent happiness, giving that agent the information that they need in order to service that customer, understanding what they're calling about, what they did on the portal, a whole history. Bringing that all together so that when the agent gets the call or the contact, they know what's going on, and they're not struggling for the first minute or 2. They can just help them. And then with some of the Agent Assist, you can have AI guiding them through the conversation, which again helps us from a training perspective that we don't have to train our agents to be the best at everything. They have like a coach on the spot helping them through the process. So really, what we see from an agent perspective is the turnover rate should start to go down because the agents are happier with the tools they have and the information they have to service the customer.

You guys are making some great points. I guess, I've been thinking about how AI could impact like the labor spend and the cost perspective. What I'm hearing from you is like more the efficiency and how you provide a better customer experience, and you're almost more interested in how you can drive more revenue opportunities. Is that a fair characterization about how you guys are viewing the AI potential in the contact center today?

Absolutely.

Perfect. You guys all mentioned having IVA or other AI solutions with Five9. Love to hear how you're finding the payback and justifications around what you've implemented so far and kind of the ROI that's driving at this point.

I can jump in again to start us off here. When I talked about we're reaching our 3-year goal from just an authentication perspective, that's a dramatic operating expense reduction for us. We were forecasting that to take 3 years, and seeing it in 3 months is just remarkable. So we actually don't have to hire as many people as we used to. We stripped about 10% of the duration of a call, so the average handle time by utilizing this. And this was just our first phase, which is kind of, I think, what Joe was talking about, the low-hanging fruit side of the house. Now we're going into a much more complex phase, much more API-driven phase of helping our members understand their benefits, where they are with their benefits, and what's the next logical step for them to process a claim or anything of that nature. Much more complex, but what we've seen so far with the conversational design that has gone on to really drive that best member customer experience. We're looking forward to deploying the next phase of this later in this quarter.

I'm following the footsteps of Joe DeLuca. We are moving down the pathway of using the IVA to do password resets for us. And with us, we are on a path -- growth path at Chipotle. So as we add restaurants, add staff, their mandatory training is every 90 days with those things. And what happens is they forget and they call in, what's going on with the staff? So this drives us, allow us to continue the growth trajectory without having to hire more people or spend more output with other -- those who are answering the phone and do those things as well, but also give us actual intel. I mean with the -- everything you have there to make that experience what it should be and also a great coaching tool. It comes back to tell us where we need to put more money, training and put our focus to really make it a better customer or employee experience and make sure that we are not missing anything. It just helps all the way around. So I see it as iron clad our feature, really. Making us better for the future to make sure that we are -- our investment maximizes that, and then we also become the best we possibly can be.

Yes. Just to jump in. I'm so happy you brought that up because it's not like a contact center, set it and forget it, right? These are evolving, living, breathing systems that like you already have a use case in a marketing program. So it makes sense that like as you see more efficiency, you look to see where other areas you can do more with. Guys, just as we wrap up here, I'd love to hear just your thoughts on like what's coming next in the contact center? How you can see this technology impact your business over the next 3 to 5 years? And if there's any use cases or anything particularly interesting do you think you can adopt to make your contact center better?

I think I can start that off. So I think what's next, undoubtedly, it is -- it revolves around the use of AI. But without a doubt, the solutions that provide all of those insights, the things that Ronald was talking about, that's invaluable, right? Understanding why folks are calling and being able to utilize that information to further target and focus and personalize the experience, right? And I think solutions that also really achieve that personalization is what you're going to start to see across the board. And of course, obviously, it ties into a lot of other aspects of different benefits that are easy to achieve such as -- on banking, just the experience itself with automation and not having to rely on a human, right? And that's the easy stuff, but to take it to that next level, like I said, the insights are really what's invaluable.

I'll tag on what Joe says. The 3 Ps for me, too, is predictive, personalized and proactive care. Those 3, AI will catapult my 3-year plan into probably about a year to help me internalize those things, set systems in place and actually go after to make our business grow more profitably and also give us a higher level of satisfaction, not only for the employees but also for the guests.

And I'll wrap this up by talking about really from our perspective, again, long term, AI really would impact our back office more so than our agents. So I think about our knowledge management and how do we take all that knowledge that we have and being able to share that with our members when they attach to our portal or when they're calling into our IVA. It's really that knowledge management that in every contact center I've ever worked in, is problematic and takes a lot of effort to manage. But now, there's the possibility that by pointing the technology at our data and then using it through APIs with Five9, that we have much better outcomes as far as quality is concerned. And again, I think we'll -- as I listen to this conversation, I hear what others are doing, really, I think the opportunities are really unlimited as far as how it can drive up Net Promoter Scores while reducing our cost if you design it the right way, which is also something that we really haven't talked about. But you've got to really design this for the right experience so that you're not creating customer abrasion.

Well, Joe, Ronald and Joe, thank you so much for taking time today. I know this was invaluable for our audience to hear directly from you as customers and how you're leveraging AI and automation. I know, Ryan, that you had several questions you wanted to ask of the Five9 management team. So thank you all, and back over to you, Ryan.

Thank you. Excellent. Guys, appreciate the time. Just as a transition, just following off what Joe said. Yes, I always complain about how our first 2 weeks in the contact center, they just gave you a binder to study student loans for 2 weeks and then it's like, all right, you can hop on the phones. So like, there's got to be a better way to do this. So -- but it looks like we're kind of in the early stages of that for sure. Dan, just because we came off the customer panel, I'd love to start with you. Just as activity begets activity in the contact center and there's new technology, and you're probably receiving a lot of inbound questions, love to hear kind of like what are the pain points or interests of your existing customers today around AI? And what are they thinking about buying in the near term just in terms of AI and automation?

Yes. Thanks, Ryan. And you all heard some of those examples just in the last few minutes. I'll kind of break it into 3 categories. One is, and what you heard first was customers are experiencing this end-of-life scenario where their products on-premises are legacy. They're older. The manufacturers have said, we're done innovating there, we're done enhancing them, you've got to move off of them. And they have very few choices. They mentioned the 3 main providers, including Five9, that are CCaaS providers or cloud platform providers. So that's the first step is get off the legacy, move to the cloud. And then it's the ability to leverage the AI and automation that's truly only available on a platform that's cloud-based. And that's something we see as a potential catalyst and an accelerant to companies saying, aha, it's going to take us a year or 2 in some cases, depending on the size of -- and scope of their company to make that migration. And they're starting to recognize, especially as companies like you just heard from, have already started implementing these solutions that they better hurry up because they've probably got a year or 2 to get to the AI point. And so we see that as a potential accelerant.

Excellent. Yes. And that touches on my next question either for Mike or Dan. We talked about how like in the past, some large contact centers would never go to the cloud, right? Like they would just sweat their assets for years and years, and that's like what are going to make these guys finally move? So I think, could AI be the catalyst? Like, could this be the thing that finally catalyzes like, okay, well, we're never going to move but now, we need these features? Maybe for Mike, do you expect the rate of cloud migrations to change or maybe see interest from those customers that you felt would just stay on this asset for a while given this change in generative AI?

Yes, Ryan. It definitely is a catalyst. As Dan said, we view it as an accelerate to the process that has been happening for many, many years, right? We saw for the last, I'd say, decade, and I think we'll see it for several more years this mass migration off of on-premise solutions to the cloud, as we've talked about many times. This is still a market that is 20% in the cloud and 80% on-premise, so it's a massive, massive opportunity. And if you think about -- we debate about agent counts and what that's going to do over time, but don't lose sight of the fact that this is a massive market that is going to the cloud no matter what. And this is -- this AI revolution is just going to excel, and that is going to benefit Five9 and other players in CCaaS.

And Danny, you talked about like your pipeline and how things are going well. Would you be able to characterize like how AI conversations are impacting that? Like, are you seeing like shorter deal cycles as a part of this or maybe more -- or the activity or more testing? Like what -- has that made any changes in terms of like your near-term opportunities?

Yes. A couple of things to unpack there. One is the fact that the doubling in the pipeline of the high-end enterprise, our strategic accounts over the year, somewhat driven by this AI and automation revolution. But also by the fact that you mentioned earlier, why enterprises were "reluctant" or willing to sweat more -- one more year. A lot of the large enterprises needed to wait, and we saw this very same thing happen with CRM, right? Until you can go prove yourself at the highest point and most complex and global organizations in the world. As soon as we demonstrated that with a couple of the large Fortune 50 accounts that you all know of at this point, with tens of thousands of agents in a very complex environment, their criteria was we got to make sure you can scale, be reliable, protect our data, deliver all the functionality you have today and give us an easy migration path through the tools and expertise to move off of one very -- this is like -- I think Mike's used the term, it's like heart surgery. You're basically taking -- you can say heart surgery or central nervous system. We have tentacles into all the back-office systems. And when you go off of the prem systems into the cloud, all those integrations need to be rebuilt. So hopefully, we can just replicate all the feature functionality that comes standard and move that over to our platform and spend our time on that new, innovative and complex side of things. So as soon as they saw that we could check all those boxes and deliver at scale with all those capabilities and to give them the platform to then innovate on top of, that was key for the large enterprises to start on their process. So we're seeing an acceleration there already.

Yes, I love hearing that from some of your larger customers that it's like, look, we're really excited about the opportunities today and the future improvements today, but we're really future-proofing our system here. And it's not just about this year, but the next 5 years. And I think it's a good time to tag in Barry. I know we're still early in this AI revolution, but I'd love to hear about how you think about the potential for AI to be monetized today and how do you envision pricing given AI capability?

Yes, absolutely, Ryan. Thank you. So just as a reminder, for some of the audience who may not be intimately familiar with Five9, we get $200 per month just for access onto the platform that Callan and Jonathan described, all those features and functions. And now with respect to AI and automation, the best illustration, of course, is the IVA example, which is what the 2 Joes and Ron talked about so enthusiastically. And we sell those on a per port basis, and that's $400 per port per month. And that's only one of the portfolio of 8 different products that we currently have with more to come. So it only goes up from there as we embed throughout the whole, or more precisely as our engineering department and product invest throughout the whole solution or the AI and automation. And by the way, this has been a passion of Five9 for -- we've been talking about this now for 4-plus years. We're a software company. We believe that AI and automation drive efficiency and productivity gains. And the more software that we sell, the more revenue we get, and our customers are also happy because they get a, you just heard, significant and tangible ROI for what we're delivering. So it's early days. But as Mike pointed out a little while ago at the beginning, we look at this concretely and confidently as a TAM expansion, as a mix shift towards automation.

No, I appreciate that perspective. And look, I have to talk about some of my own research. We've published a note today talking about some estimates around pricing and agency growth. But maybe for Mike or Dan, we'd love to hear your guys' perspective. We keep hearing like, what if half the agents go [ live ]? What could that mean? It seems a little interesting, given how like for years and years, we always talked about how contact centers kind of only expand. But love to hear how you just think about changing agent growth rates and maybe the overall agent opportunity in light of AI.

Yes. Sure, Ryan. I'll start. Dan, feel free to pile on. But there's an ongoing debate, as you stated, Ryan. It's -- and again, I think if you look at industry analysts, they'll tell you that the agent counts are not projected to go down even in very recent reports. If you talk to -- you listen to the customers we've talked to today, it's not a massive decline if there is a decline in terms of agent count that they're anticipating. But let's just, for the sake of argument, assume that there is a more significant decline. The good news for us is remember, Five9 is a software company. We provides solutions to our enterprise customers to process interactions. Think of it as interaction capacity. And those interactions can be human to human on one end of the spectrum. They can be fully automated on the other of the spectrum. Most of them, in my opinion, are going to be in the middle in this blend, somewhat assisted by AI, but probably still involving a human. We can debate whether that impact on agent count is going to be 0, small, medium or large. At the end of the day, we monetize any interaction, whether it's human to human, whether it's fully automated or somewhere in between. We're a software company that provides solutions. We provide the platform, which Jonathan talked about as kind of the control point, which you have to have. But understand, it's important for everybody to understand that we have products and solutions for all types of interactions, not just human to human.

And that interactions hub or that plane is pretty complicated, right? Like walking through the job. An example, like when people always ask me like, why do you contact center deployments take 6 to 9 months? It's like, well, it's not like they're just sitting there waiting for it to get spun out. So Mike, I would love to kind of hear your perspective on just like, what goes into a contact center deployment that requires this significant customization? Like what back-end systems do you link into? Like why is professional services such a moat for Five9?

Yes. I mean, this is a huge one for us. It's a huge reason why many customers choose Five9. Quite frankly, it's our team of experts. We've got close to 500 services people that are absolute experts in this field. They -- most have come from the legacy Avaya, Cisco, Genesys in terms of their backgrounds. As Dan said, this is very much like open heart surgery. This is a major technology transformation. You talk about ERP or CRM transitions and transformations. This is, in most cases, more complex than those transformations. It requires expertise but it also requires integrations. You mentioned it, Ryan. Most of our large enterprises probably have 30 to 40 back-end systems that we integrate to, 30 to 40. Talk about complex and sticky and requiring expertise. And so at the end of the day, we talk about this a lot. Five9 really brings the power of technology, but we also bring the power of our people. And that's a big reason we win in these large enterprise opportunities is because of our expertise and our cultural value-driven behavior in terms of customer first. We do whatever it takes to drive success for our customers. And it's about driving business outcomes as well as obviously, joyful customer experience.

Yes. And I love the idea that, like, the efficiency gains don't just stop there, right? Like, they keep wanting to do more and more and like, the more capabilities of your products you come out with, the more they can add to their own platforms in terms of software adoption. Callan, if you don't mind, I had a question come in from an investor that I thought might be interesting. Just would love to hear your perspective maybe perhaps on how you think the evolution of voice AI versus the chat-based GPT conversations could shift over the next year or so given the events in technology there?

Sorry, that one was directed at me?

Yes.

Yes. So how will the shift take place? Well, basically, I think the way to think about it is if you think of LLM as an engine that allows a new class of interaction to be unlocked, then you're likely to see it used in those sorts of interactions. Now it doesn't really matter whether we're talking voice interactions, chat interactions, e-mail interactions. Basically in all of these channels, we essentially convert to text. That's one of the first steps that we do. And then once you have it in text, you can apply either an NLP model or an LLM model or any other model that comes out for that matter. So I think what we've seen over the last few months is that there's certain classes of task that are now much more easily solved using an LLM than with traditional techniques, and the big one that you'll see is summarization as a task. We already have in market essentially a post-call summarization product that essentially is powered by LLM that does exactly that. If you look at summarizations as a task or a class of interaction, that was really hard to do with just the earlier generation of technology, NLP and things like that. So that's the way I think people should think about it as essentially unlocking new types of interaction automation using LLM. And then we've seen this time and again, as Jonathan went through in his section of the presentation, as the new eras have rolled out, we see new types of interactions being automated.

And Ryan, if I may touch on that from a customer's viewpoint, what I'm finding is the customers want to offer -- and when you talk about voice versus chat versus e-mail versus chatbot, they want to give choice to their customers. They don't want to push them towards one or the other because remember, at the end of the day, the customer experience and that NPS is so important to them that they want to just give that alternative. A convenient time, one version or one channel may be far more appealing to you than another. I'm going to use my PC and a website to do certain things when I'm here in my office versus when I'm driving down the road or in some other capacity. So you've got to really giving choice and helping those customers but not forcing them into a self-service option. We've all seen how that's been responded to. So that's one from a customer standpoint. And then as Mike mentioned, the beauty of looking at this not as a seat thing, I know we've always used seat because it's an easy thing for people to understand because historically, it's been agents. But think of it as we're handling all the concurrent customer interactions, and those aren't going down. As the business grows, those go up. And so we're looking at it from a port -- each port or simultaneous concurrent transaction, and there's going to be a whole lot of mix. Some come in and do a little bit in the IVA. We saw that from one of the customers here that's just doing voice authentication and some intent and then getting the calls routed to a more specialized agent, and then helping them with more. So the key there is we're an interaction hub or an interaction management platform, and we'll take all interactions from whatever channels. We'll handle them from each end of the spectrum and everything in between. And the more automated -- more automation enters the equation, the more software enters the equation, and therefore, more dollars enter the equation.

Excellent. I think that's a great place to wrap things up. You never know what new technologies will pop up. Maybe next year, Michael, we're talking about like AI and VR and contact center in virtual reality, but we'll see. Mike, do you have any closing remarks there? I appreciate the time.

No, I just want to say thank you, Ryan, and thank you to everyone thait joined us today. Hopefully, this was insightful for everyone. We sure as heck enjoyed putting it together. And thank you to Joe, Joe and Ron, for joining us as well. We really appreciate you making the time to be here and to tell our story.

Yes. Thanks so much for providing us folks access. It was great to hear from the entire Five9 team. And for the investors on the line, if you do have questions, we can get them in front of these guys and we'll sort them out with some answers. But guys, appreciate the time. Thank you so much.