Thank you very much, Henry and Graham. Welcome back to the Government Staffs Community Coffee and Technology Conference. I think last year you guys were the last presentation. You closed out the conference, and this time we got to the second day of the conference. It's been action-packed, to say the least, so far. We're just halfway through the conference. Our registration stats are up. We're not a tech company, but we're up like 4% or 5%. We exceeded 3,000 registrations. Client activity, enthusiasm for digging into AI themes, very, very, very strong. We've had a record number of requests for meetings, that sort of thing. Thank you for—it's not possible without content. Thank you for being part of the content and presenting yourselves here. Thank you for having us. Yeah, absolutely. Henry, welcome back. Congrats to you, Graham, officially. I think I was making the call last time that we were all together, that, oh, you know, he's doing a good job. Maybe he should be CFO. Thank you, guys. Yeah, appreciate it. I asked you the same question, Henry, and the answer probably will be dynamic because the environment has changed. Where do you see ZoomInfo at five years from now? I'm sure that has changed if I asked you the question three, four years ago of a certain answer, which I did, I think, in 2022, 2023, 2024, 2025. Where do you see the company? Thank you. You asked me this question last year, and I didn't take the bait because we were in the process of changing a bunch of things and reworking a bunch of things. I thought it was more useful to tell people what I was going to do over the next six to twelve months. This time I will. In five years, I believe ZoomInfo will be synonymous with AI and go-to-market. That means a lot of different things. What I expect is, you know, we changed our ticker earlier this year from ZI to GTM. We expect that the products that we're building today and delivering to our customers and our data asset will become synonymous to building AI in go-to-market. Between now and then, we expect to accelerate revenue growth, expand our margin, and continue to be big buyers of our own stock, particularly when they're at prices that are below our intrinsic value, you know, like now. That's how we think about the future. How will we deliver AI to go-to-market? How do we become synonymous with, if I'm going to build AI in go-to-market, ZoomInfo is a necessary component to that. That's the vision five years from now. We're definitely seeing an evolution of the story as we speak. In the past few quarters, definitely the product strategy has changed. The business model is looking to change. You're a year into the new business model, if you want to call it a new business model. The upmarket motion is trending very nicely for you guys. ZoomInfo Copilot's ramping quite well. What are the primary initiatives to continue to do that will continue helping drive the growth momentum that you've seen off late in the last couple of quarters? Yeah, so there's a couple of sort of key things. You know, one, like you mentioned, we've really shifted the focus of the business upmarket. Today, the upmarket component of our business is 72%. It's growing 4% a year and accelerating. The downmarket portion of our business, we're driving more self-service, more PLG, more AI-driven service components down there to reduce the cost to serve and the cost to acquire. When we think about the growth algorithm, there are a number of things that we're focused on. I think the first one is just data. Our proprietary data asset is a necessary component to any AI-driven initiative that touches go-to-market. You see that in our Ops OS, our operations numbers, which are growing 20% year over year. What we're seeing there is customers coming to us and saying, particularly in the highest end of the enterprise today, customers coming to us and saying, we're building AI for our go-to-market teams, but the universe of data that we have inside of our CRM is not going to be enough for us to be able to properly deliver the insights and the automation that we need for our sellers, for our marketers, for our account managers. Why is that, Henry? What is the CRM data missing that you're able to fill in? Yeah, so first, if you go, there's so many things. If you go into your CRM data today, it is a static representation of whatever your seller decided to put in the CRM system. What you see today is a lot of AI solutions that are getting momentum where they rely solely on first-party data. Support use cases, chat use cases where I could put in my knowledge base, I could put in my past support tickets and my interactions with customers, and the AI will do a better job of answering the next incoming question from a customer than a human support agent. In go-to-market, my first-party data in my CRM system doesn't have all of my buyers. It doesn't have my entire total addressable market. The data that it does have is static. It's not being updated. It's not being enriched. If you just look across this room or in the conference, there's a large percentage of people who are in seat at new jobs, at new companies that weren't in seat at those jobs a year ago or six months ago. That change doesn't get automatically reflected inside of the CRM system. The CRM system is out of date. At the company level, companies are growing, they're shrinking, they're doing M&A. None of that gets automatically reflected in the CRM. Let's just pretend in a universe that the CRM did have perfect data on what was in there. There's a whole universe of things happening outside of the CRM that are critical for its seller, account manager, or marketer to know in order to architect and automate the go-to-market motion using AI. For example, company earnings calls. The company has an earnings call that says it's growing in Latin America, or it just did a RIF. That's an important statistic or important knowledge to know before I go engage the company for an upsell or a renewal motion. Company visits your website, researches the competitor, researches another solution, hires a new CEO, mentions something on a podcast that's relevant to your business. All of that context doesn't live inside of your CRM and is necessary for a go-to-market AI motion. What you have to be able to bring together is your first-party data. The CRM data is necessary, and it needs to be married to third-party data that gives sellers and marketers and account managers a view of the world that doesn't exist exclusively inside of the CRM system. How do you do that? I think first, if I go inside of any enterprise CRM today and I go look up the company Cisco, I'll probably get 17 different records of Cisco. There's a Cisco opportunity, a Cisco new business, a Cisco trial. When you are trying to build automation around when should I engage with Cisco, what should I say, who are the important people, I need one record of Cisco. One of the core engines inside of ZoomInfo is the ability to take data from a large contributory network of hundreds of thousands of users, a contributory network of tens of thousands of customers sharing CRM, marketing automation, sales automation data, and then a variety of data acquisition sources, data that we're acquiring, and bring that together to publish one Cisco with all of that information. In Go-To-Market Studio, what we've brought together is the unified go-to-market data warehouse that brings in that first-party data, your CRM, your calls, your emails, your transcripts, brings it all together and marries it to ZoomInfo to create effectively a virtual CRM layer that's self-reinforcing, self-healing, self-refreshing so that sellers, marketers, go-to-market professionals now have that foundation that they can start building AI automation and agents on top of, effectively. A data warehouse. How long have you been building this data layer? The last two and a half years. Yeah, and the key around that data layer is it's go-to-market. Is it not the old ZoomInfo database? It's like CRM interaction data with. With ontology that's AI-specific, that understands go-to-market, understands when there's an account and an opportunity, those are still one entity, that understands the go-to-market nomenclature, and then can bring that all together in one place. Tell us more about it because on earnings conference calls, we don't have the latitude to go to these kinds of elaborations. Tell us how you built it and how you used AI to. Mark's going to be on stage in about an hour. We're going to be talking about this, how they built the Terra Cloud and how they use AI agents on top, practically become, and Eric Yuan is going to be here on stage, exactly the same stage in about 30 minutes. We're going to be getting into this. Tell us how you built this layer and how defensible is the technology that you. First, let me just take you through the journey over the last three years. We've rebuilt our product organization. We've rebuilt our engineering organization. We've embedded AI throughout that organization. The way a product gets built today actually starts with an AI-written PRD that gets passed through engineering. Engineering's leveraging technology to drive velocity in the way that they're building. We have a data layer, a Postgres data layer at the core of this. The data from a customer's CRM data comes in. We use that technology from a matching and enrichment perspective to marry that to ZoomInfo data. That builds the layer that the go-to-market practitioner can build on top of. I'll give you just an example here when we talk about architecting a go-to-market motion. If I'm a customer, let's say I sell to roofing contractors. I need a database that gives me all of the roofing contractors in the U.S. I have to have that from ZoomInfo. I bring in my data so I know which of those roofing contractors I'm doing business with. On top of that, I want to architect something. I want to say, hey, I want to look at every single time somebody from one of these roofing contractors comes to my website or hires a new CFO or adds a vehicle to their fleet, I want to do something in an automated way. Sometimes I want to send that to a sales representative to take action. Sometimes I want to just use an AI agent to send an email or make a phone call or launch a marketing campaign. Once I have that data layer there, I can start architecting the signal that I want to take action on top of. You need to activate. That activation layer is where am I actually going to take the action? Am I going to have a sales rep work there? Am I going to have a marketing campaign automatically kick off? Am I going to have an AI agent interact with the company? One of the things that we realized when we built Go-To-Market Studio is if you actually go out to marketers and sellers and you ask them in your account-based marketing campaigns, what are the top 10 ways that you activate those campaigns? This is a study by Gartner. Five of the six top ways they do it is with sales, with an SDR, with an account executive, with a seller doing some action. Copilot has become the interface for sellers to take action. When they architect that go-to-market motion, they can push it into ZoomInfo Copilot, where inside of ZoomInfo Copilot, a seller can start taking action against the architected plays. That all starts with a data foundation that has to be robust, accurate, and singular in its entity matching. We feel like we're incredibly well positioned to deliver that in the upmarket and throughout our customer base. What do you think about the role of third-party foundation models in this layer? Yeah, great question. I forgot to include that. In that same contractor example, one of the things that we're seeing customers really change their perspective on was historically, I would build out messaging based on like an industry and a persona type. I would say, OK, if I'm contacting roofing contractors with over five car vehicles in their fleet, I'm going to talk, and it's the owner, I'm going to talk, I'm going to send them through this track of emails that go out, or this track of calls that go out. It would be industry and kind of persona-specific, but there would be broad swath. What customers want today is row-by-row personalization in their interaction with customers. When I have that data foundation, I have all the engagements I've had with those contractors. I know when they visited my website. I know what they've said on calls. I know what they've said in emails. I now leverage the foundational models to say, row by row, across every roofing contractor, take into consideration all of the engagements that we've had, all of the things you know about that company, everything ZoomInfo knows, all the first-party data, and write me row-by-row personalization from a messaging perspective per customer. Create for me the ad in a unique way, row by row. You can do this in natural language. You can do this in natural. You can do this with a prompt. The prompt will basically open up a connector through an API call into the ZoomInfo Go-To-Market Studio. Yep. Produce the results. It produces it line by line across that universe. You create the universe. You get the talk track row by row using the foundational models. You could do other really interesting things, too. Sometimes there's a data attribute that doesn't exist inside of ZoomInfo that's important to a client. For example, maybe I want to know, I'm going to go outside of the roofing contractor example, but I want to know that a company is SOC 2 type. You must have done a roofing project. We did have a roofing project. It was top of mind. I did an HVAC project. Also a great use case for us. Let's say I want to know every company that's SOC 2 Type 2 compliant. I don't have that data asset that lives inside of ZoomInfo. We haven't cataloged it. I can use the model, the foundational model, to go out, look at those rows of data, and then bring in a data attribute. This is SOC 2. This is not. This is not. I've seen our customers say, we really want to engage with customers that have very complex go-to-market models. They have PLG and upmarket and downmarket, and they have multiple product SKUs. Go out, research the company, and then give me a score of 1 to 10 based on the complexity of their go-to-market model, and then use that in the messaging. Now you have this universe of companies you're going to engage with. You can create new attributes using the AI, and then you can create row-by-row personalization that goes all the way to create a unique campaign asset, create for me a deck that I'm going to present at that company when I go to meet them. That's where we're leveraging AI throughout that. This is a very different conversation than the one we had last year and the one that we had two years ago. I mean, here you have incredible innovation, AI infusion into your product suite, whereas in 2020 and 2021, it was just growth. It was so easy to come by. We're working harder now and not getting the—and it's not just a ZoomInfo specific issue. It's been Salesforce too, Workday. It's all over the place. What is your—you take a step back. We're working harder, we're innovating faster today, but the growth is not there in the industry. Is it because the investors think a little bit of AI is taking over, right? Is it that, or is it just that we overbought and the industry is just working through overconsumption? It's just a matter of working through that cycle, and then we're going to get back to some normal? It is that we overbought, and we're working through that cycle. Just last week, I went, I had two customer calls back to back. Both of them, the first one has 27 go-to-market technologies that they've bought. They have no idea what the things do. They have this one and that one. It does this little thing and that little thing. They're just begging us to help them rationalize that technology stack. The second call I took, exactly the same thing. We buy a little thing from them and a little thing from them. There are numerous problems with the situation that particularly software companies got themselves in by overbuying these technologies. The first one is, obviously, we're not using all of them. The second one is the promise of these technologies really was, help me prioritize my next best action. Who should I be prioritizing conversations with across my customer base? When I have 27 tools, each one of them with siloed data about my customer, I get no overall view of the customer. I instead have a little bit here, a little bit there, a little bit here, a little bit there. One of our foundational premises around Go-To-Market Studio with that foundational layer is, let's consolidate all of that data in this one place with a go-to-market lens on it so you can make sense of the interactions across your tool set. Historically, I would just use ZoomInfo. There was an insight somewhere else, but I didn't know it. I didn't have visibility into it. We're trying to help people get value out of those tools. Still, customers have a tremendous amount of go-to-market tech sprawl that they're trying to consolidate. A new leader comes in and they have a different vision for how they want to consolidate. They maybe push out one or two tools. They realize the wrong one or two tools. They bring back one or two tools. We're still kind of fighting through the overbuying that happened. Got it. Graham, how do you keep up with this guy? He's going 100 miles an hour. Do my best. Of course, the first question is, I guess you got the inside draft on what the CFO job is because you've been prepping for it. Now, taking over the job function completely, what is the mandate? Yeah, the way I view this is, you know, the end of last year, maybe the first quarter or two of this year was really a story of stabilization. Now we're going to go build on that stable foundation. I think our upmarket opportunity is increasingly evident, and we're set up to go continue to capture that. Beyond that, I want to be focused on optimizing our downmarket business, getting that to a place where it's a smaller and healthier version of itself, and then effectively managing the balance sheet. I think if you do all of those things, and we've been doing all of those things for a couple of quarters now, you're going to get to a place where we have an opportunity to meaningfully re-accelerate revenue growth, expand margins, buy back shares, and essentially create this compounding effect on free cash flow per share. What is the, I guess, the trade-off is, I mean, although upmarket is growing nicely, you don't want to lose the downmarket. What is the trade-off between what is an acceptable level of growth you want in the downmarket versus the upmarket? How important is that trade-off to your go-to-market? Yeah, you know, I'll start with the upmarket. It's growing 4% right now. I think as we get into 2026 and 2027, we're really focused on getting the upmarket business to a place where it's growing high single digit, low double digit. I think we can get there just by getting growth in that segment to 100%, sorry, retention in that segment to 100% plus. We're pretty close to that already. Downmarket. NER, right? Net expense ratio. Net retention, yeah, net dollar retention. The downmarket business, I think we've been very cautious about not relying on that as a contributor to our near-term growth. It's more sensitive to macro and other trends. It's still a really valuable business. We want those customers from a logo acquisition perspective. Some of them will graduate into being healthier upmarket customers with better LTV outlooks, as well as contributors to our data asset. I think about the downmarket business right now as 28% of the total mix of our business and declining. Next milestone is let's get it to be about 25%. Over the next few years, get it to 20% or a fifth of our business. I think when we're at that point, it should be closer to 0% growth, maybe down a point or up a point in a given period. It's less dilutive to overall growth. Downmarket business has lower margins by a pretty significant gap relative to our upmarket business. We have this unique opportunity to shift our business upmarket to a place where there's better growth opportunity, better margin opportunity, but still be competitive in acquiring downmarket business. How do you delineate upmarket versus downmarket? It seems like a binary thing, but I'm sure that there's a lot more shades of gray. Yeah, it actually is pretty binary. Upmarket is what we traditionally refer to as our enterprise and our mid-market customer segments. It is any customer of ours that has 100 or more employees. We define it as upmarket. Any customer with 99 or fewer employees, we define it as downmarket. That is, so 100 plus. Yeah, that's not enterprise market. I mean, that's like upmarket meaning. Right. Yeah. Henry, what are the biggest wins you have secured in recent times that give you conviction that this new AI strategy is indeed working? I know you talked about the roofing, but any other, like a Fortune 1000, Global 2000? Yeah, we talked about on our last call, we closed our largest deal in the history of ZoomInfo. Yeah, largest TCV. Largest TCV deal in the history of, largest ACV too. No, close. Have the largest TCV deal in the history of ZoomInfo. That deal was, you know, one of the things that we saw with ZoomInfo Copilot was, and this was a Copilot almost wall-to-wall deal at the company. It also was our operations data into the company, also our marketing solution into the company. It was a platform sale. I think that the thing that we've seen and we saw there was when one of the things that happened from an overbuying perspective between 2020 and 2022 was customers brought on account executives, account managers, and CSMs onto the ZoomInfo platform. If you went across the user base and you talked to SDRs, CSMs, AMs, and AEs, and you asked them, is ZoomInfo indispensable to your day-to-day workflow? SDRs would have told you, absolutely, it's indispensable. Account managers and account executives and CSMs, they would tell you, like, I use it here and there, but I wouldn't call it indispensable. What we realized was that it was critical to the future of ZoomInfo to get into the core workflow of account executives and account managers. When we rolled out the first version of ZoomInfo Copilot, our intention was that over time, this would be a solution that sat in the middle of an account executive and account manager workflow in the same way that it sits in the workflow of a sales development rep. What we saw, what we've seen now with our usage statistics on Copilot is now account executives and account managers have the same levels of usage in Copilot as sales development reps. We feel really good that we built a mousetrap that gets into their workflow. We're continuing to iterate on Copilot. There will be a big launch later this month that adds more agentic technology into the platform that gives sellers a view over their full territory. We think we're going to continue to embed within account executives and account managers, which is actually kind of an insane thought because today there isn't a pane of glass that account executives and account managers work out of every day. They work a little bit in CRM, a little bit on the internet, a little bit in ZoomInfo, a lot in Excel spreadsheets. That's where they track their renewal opportunities, their prospecting opportunities, a little bit in email. What we've endeavored to do with Copilot is to give them one interface that brings all of that together and allows them to ask questions of their territories, build decks and artifacts, build an account plan, write back to CRM. We're working really hard to build that pane of glass for account executives and account managers. Why is that new agentic capability? That new agentic capability is why. Why is it so used to the? The first iteration of ZoomInfo Copilot has agentic technology in it. The new version of ZoomInfo Copilot adds and expands that, particularly around building artifacts, writing back to CRM, building decks, building account plans, building a point of view for your next call, and who you should be engaging with. We're releasing that in two weeks. We feel really good about that tech. Back to the large deal, that technology has allowed us to expand outside of sales development reps and into account executives and account managers. From a seat perspective, we're now able to have real expansion within our customer accounts. That large account was account executives, account managers, customer success managers, and sales development reps. Got it. Is this, maybe I'm being leading, but is this casting possibly a halo effect on the core data platform that since you've added so much value around it, could there be a renaissance of the core data platform? I think there's two renaissances that we're seeing. One is high-quality data is incredibly important to go-to-market practitioners. We've had record win-back performance for the last four quarters in a row where customers went to a low-priced, lower-quality provider and then very frustratingly came back or got frustrated there and came back to us. We're seeing high-quality data become more and more important. The second thing we see is that we're hearing from a smattering of our customers, hey, the Google AI overviews and the LLMs have taken 30% of the traffic away from our website. We're not seeing the same amount of traffic to our website anymore. That's not converting into demand the way that it used to. What we're hearing from our customers is, oh, we're re-accelerating the hiring of our sales development reps to go outbound and re-accelerate demand that we lost because we lost this traffic to our website. That's what our customers are saying. Our customers are saying, we're hiring more SDRs now because we now have this demand gap. How do you make up a demand gap if you're not going to get it through SEO and through the traditional search engines? You have to go out and create it. You go out and create it with SDRs. We're seeing more and more customers now accelerate the hiring of SDRs to go outbound and create demand that is now lacking because the traffic to their websites has gone down. I thought SDRs could be replicated with the ZoomInfo Copilot. Inbound SDRs can be replicated pretty well. If you get a lead-in from the website, certainly somebody can communicate and schedule a meeting. Outbound SDRs, I haven't seen anybody replicate that with AI. Account or AI? People are hiring. People are account, yeah. That is actually. Yeah. Wow. OK, that's Graham, back to you. Copilot, now decent adoption, over 10% of the customer base, according to my information. What is the unit economics of the Copilot product if you do track it and the margin profile of Copilot overall? Yeah, first and foremost, we built Copilot to deliver durable value to our customers. The net revenue retention of Copilot will be the kind of primary metric that we track. The first year of Copilot being out in the wild was largely a story of migrating the existing customer base over to Copilot while also selling it as new business. We were really successful doing that. We were able to get uplifts on a per-seat basis from existing customers as they migrated over to Copilot. Now we're starting to enter that next phase where folks have been on Copilot for six months or a year, and we're starting to see renewal outcomes. We've been tracking the leading indicators of these renewal outcomes for a year now, seat utilization, frequency of usage. We have more bespoke metrics that track specific actions and conversions. All of those were positive and historically have led to more positive retention outcomes. What we're starting to see now at the very end of Q2 at a smaller scale, and then in Q3, it's starting to ramp up some, and Q4 will step up a lot. We're starting to see these first renewal cohorts show up on Copilot, and we are seeing early signs that the renewal outcomes are better. If you think about the unit economics there, they're pretty similar to the rest of our business, high margins. If we can actually go and realize better retention, better customer lifetime value from Copilot customers over time, that can deliver improving margins. Got it. Anybody wants to jump in with a question? We've got a couple more minutes. The other thing I wanted to ask you was, curious, since you've been in the industry for a long time, foundation models. First, there was one, and then there were two, three. Now it's like I can't keep track. CROT is now a big deal. Microsoft is in with AI. What do you make of the flooding of foundation models, seemingly all trying to do the same thing? Is this focusing an opportunity for companies like ZoomInfo that you think that the cost of compute and the cost of access intelligence is going to come down? What do you make of this? If you project this trend out, what does it mean for your company and more broadly the software industry, the applications? Yeah, I think, one, you do see cost of the models coming down exponentially over time. Are you getting the economics? We are, absolutely. The way we kind of think about all of the different models is there are some models that are of lower cost that can do a certain task very well and maintain that lower cost. There are some tasks that need a newer model. We're balancing the different models across the different tasks to get the best cost-efficient outcome. There are some things we're using Anthropic for. There are some things we're using OpenAI for. There are some things that we're using Google Gemini for, depending on the task and depending on the cost related to those calls. I think that's one thing. The second thing is it also means you should be building on the edge of the models because when a new model comes out, things that you weren't able to do with the existing model, you're going to be able to do with the next model. You should also always be pushing the boundaries of things that are, you know, almost not possible with the existing model or not possible because the next model that will come out will solve those things. We're seeing a lot of that. It was very difficult to build artifacts using the first iteration or even the second iteration of these models. You couldn't build a deck. You couldn't build an account plan. Today, out of the box, you can build an account plan if you give it the right context and the right prompt. We're continuing to evolve the functionality of the product as the LLMs evolve as well. Yeah, it's hard to forecast the future. I mean, clearly, we did not see this coming a year ago. Nobody saw these things coming. It's going to be equally futile if somebody tried to paint a picture. The seeming contradiction therefore might be your question. Where do you see the company five years from now? I guess it's easier to see five years out than it is to see six months, nine months, 12 months, I would trade. As you connect the dots, clearly the cost of inputs is going down. Intelligence is more accessible. As an adjunct, what do you make of this software is dead hypothesis, which I don't believe in, but I'm curious to get your, but maybe you do. Yeah. Software. What do you make of it? I hope I wasn't sitting in this seat if I believe that. Someone with conviction that that's not going to be the case should be sitting here. It's meant to be provocative. Yeah. Swing back in. I think that there are a lot of things that AI is going to give people capability to do that they don't have the capability to do today. When I think about software being dead, I think number one, that vastly underestimates the workflow domain knowledge necessary to build great software platforms for specific use cases. When we think about the software that we're building, we have to deeply understand our customers and their workflows in order to build the right types of software for them. That's going to be, you know, we have really great people who are really focused on that problem set and are constantly evolving that. It's got to be maintained. You have to go to the new model, and you have to spread out the costs across different models. I think all of that work that has to go into building a great output from these models is going to continue to exist. People will rely on third-party vendors to do that. That MIT study that literally everybody in the world is talking about said that when you partner with third parties to build these solutions, I can't remember what is 3x more likely to be successful than if you try to build something in a homegrown way. That's the same thing with software. It's the same thing we've always seen when people try to build something in a homegrown way versus leveraging a vendor with expertise. I think that construct will continue. May your words prove to be true. From my lips to God's ears. Yeah. On that note, I wish you really well in achieving your vision in 2030. Thank you once again for your investment in Goldman Sachs. Of course. Thank you, Chris.