How CRM Vendor Attio Turned Intercom's Fin AI Agent Into an Always-On Sales Rep

AI agents are everywhere in the CX conversation, but this CX Today case study gets specific. Nicole Willing speaks with Will Jones, GTM Operations Associate at CRM vendor Attio, and Paul Adams, Chief Product Officer at Intercom, about how Attio deployed Intercom’s AI agent, Fin, as a sales development representative (SDR) to handle inbound sales engagement.

Jones breaks down what drove the move, including the push to stay ahead of changing buyer expectations and the reality that static forms miss demand outside business hours. Jones shares what changed day to day after launch: always-on coverage, automated follow-ups, richer insight into what prospects actually ask for, and faster routing of qualified leads into the CRM.

Adams explains why sales is a natural extension of an AI agent that started in customer support. Customers experience one journey, not separate departments, and shared memory and a single source of truth are critical to avoid the classic “repeat yourself” handoff problem.

Both guests also tackle the hype question directly, arguing that AI is a bigger shift than the internet or mobile because it can do the work, while warning that success requires effort, experimentation and iteration.

For more Customer Experience tech news visit https://www.cxtoday.com

Transcript

SPEAKER_00 0:07

Welcome to CX today. I'm Nicole Willing. There's a lot of talk about how AI agents are starting to show up in real customer environments as day-to-day operational tools. But what does that look like in practice? There's less clarity on what actually changes once these systems are switched on. So today's conversation is around a life case study. ATIO is a CRM vendor, but also a customer using Intercom's AI agent Finn to handle inbound sales engagement. So to discuss the case study, I'm joined by Will Jones, GTM Associate at ATIO, and Paul Adams, Chief Product Officer at Intercom. Welcome guys, thanks for joining.

SPEAKER_02 0:44

Thanks, Carol. Thanks for having me.

SPEAKER_00 0:47

So can we start with you first of all, Will? Can you briefly describe ATIO's growth journey and the kind of customer conversations that you're handling?

SPEAKER_01 0:56

Absolutely, yeah. So as you'd introduce ATIO to CRM, we've raised about 116 million with a recent $52 million Series B led by Google Ventures Point 9, Redpoint, and others. We've got more than 8,000 paying customers, and this number is growing rapidly. And in terms of what the conversations that are handled are, we've got conversations anywhere between small businesses looking to buy a couple of seats to large enterprise customers that are shopping for really specific features like SAML, SSO, et cetera.

SPEAKER_00 1:25

And then what problem specifically were you trying to solve when you were kind of looking for a solution, you know, for AI support for these inbound conversations? And why was it an AI solution rather than maybe a more traditional sales automation?

SPEAKER_01 1:39

Sure. Yeah. I always like to frame this as though there wasn't really a problem with our sales motion. It was more like we were trying to look ahead of a curve. And, you know, it's, it's, it's like we we just don't want to be playing catch-up here. Um and AI agents are becoming ever more important in everyone's uh workday. Uh and I like we believe at least that AI agents are bec are going to become table stakes on websites. Um, because quite frankly, prospects are coming and they've got used to these AI tools that give them instant answers now and they're being less and less patient with getting information. And quite frankly, we we we exist in a CRM market that's really quite competitive. And having this ability to have like a 24-7 kind of inbound SDR agent on our website that's able to answer customer questions immediately and specifically means that we can kind of get away from this like more traditional kind of like static forms on the site that you fill out and and broadly like broad landing pages that are targeted to our like ICP segments. Um also we were just missing, uh, to be honest, an unknown amount of potential prospects. So, like a sales form is is always there, but it's got to be connected to a human at the end of the day. And if someone wants to get on a call at you know 7 p.m. on a Friday, it just it just not gonna happen. Um whereas Finn SDR was able to just basically give the answers immediately when needed.

SPEAKER_00 3:05

Sure. No, that makes sense. Um Paul, I'd like to bring you in here. Um, why is sales a natural step for you know an AI agent that originally came out of customer support?

SPEAKER_02 3:16

Yeah, uh so yeah, Finn is a originally, it's about three years old, it's originally a customer service agent. Uh, that's kind of where we started. And we've a long background in building customer communication and customer engagement software. And uh Fin for Service has had phenomenal results. And so, you know, over the years, as we saw it grow and um, like Will said, like really transform customers and end customers' experiences. Do you have like 24-7, 365, any language, instant resolution, like customer problems are being solved and they can go on about their day. You know, no one wakes up in the morning hoping to talk to the customer service team as much as I love all those people and teams. Um, and so when we think about these things, we think very broad and we think very long term. Same as Will was saying, you know, you want to be ahead of these things. And if you look at what AI is doing, AI as a technology is quite different to something like mobile or the internet, in that these things kind of helped us do the things we do already better. But AI does work. AI actually does the work that people used to do and frees people up to do different things, more interesting things a lot of the time. But a lot of the way that we have built software to serve customers uh reflects internal company organization. So when you think about sales and service, you think about for two very different departments. You think about departments that barely talk to each other, um, never in mind, collaborate together. But customers don't think like that. You know, for a customer, if it's a company, it's a brand, it's a product, and they don't care about your internal issues or whether one department talks to the other. So in fact, when you think about these things, sales and service are um very naturally aligned. In fact, they're they overlap. And when customers are dealing with businesses, they'll move from sales to service very, very naturally. So, for example, you know, when customers are shopping, they'll ask about return policies and what how that all works. So they're asking about service when they're trying to buy. Or we often see customers start with service requests, like they have some kind of issue, and then the issue will get resolved. And very quickly they get talking about other things, like, oh, it turns out they bought the wrong product, but they still want the product. You know, like uh, for example, they bought the camera that didn't work, they want to send it back. They still want a camera. And so sales and service um move between each other very seamlessly. But in the in the kind of pre-AI world, you just can't train a human to be brilliant at service and brilliant at sales. It's a different set of skills. So you get up with these two departments, but you can train AI to do it, you know, and and with Finn, like we have a very strong conviction having a single customer agent for all customer communication because you can train Finn to be good at all of these things. Um, we'll have more roles for Finn, Finn will do onboarding and it'll do success and do all sorts of things like that.

SPEAKER_00 6:04

No, it makes sense because as you say, with that overlap, the customer experiences it as a as a single journey, don't they, rather than separate departments that there might be internally. Um so um, Will, when it when um since deploying for sales, what operationally has changed, you know, the your team in practical terms?

SPEAKER_01 6:24

For sure, yeah. I think it what's really interesting here is because of that 24-7 through 365 days coverage, you get some really interesting qualitative data out of it that you just quite don't get with static like forms and landing pages. Um one of those things which is really interesting is where since we've got this kind of way for customers to interface with us on the front end, we that they'll sometimes give us their email address, but they won't complete a conversation, let's say. Um and Finn is able to actually write automated follow-ups to them to try and re-engage them. And another really interesting thing is that you can actually almost find out exactly what your landing pages are missing. Because your landing pages, you know, they're broadly targeting like your your ideal crowd, right? But customers don't always use the language that you think they're gonna use. Um and what you can look, you can look in the kind of customer logs, um the chat logs, and you can see exactly what they're asking for. And that means that you can tailor your landing pages even more to what your customers are actually shopping for. Um generally, like we've we've had more and more leads coming in from the front end, and it's it goes kind of goes back to that previous point where we just we were just missing an unknown amount. Um and Finn is able to write directly to our CRM, of course. Um which means that they can the prospects can get connected with the deal owner instantly rather than having to be routed through kind of a traditional sales flow. Um another interesting thing as well is the training is permanent. So if you have really good context, like sales context, let's say from your past calls, etc., um, you can basically train Finn once and your kind of sales quality will always be the same, regardless of like what the customer asks.

SPEAKER_00 8:18

And uh Paul, you know, you you've referred to kind of this bringing together of sales and service and so on under one agent. Uh why does that shared context matter shared context matter?

SPEAKER_02 8:30

Yeah. Uh it's kind of similar to what I said earlier. A customer is a customer, and you know, they like just like they don't care what um organization the person they're talking to works in, or you know, they've just their own problem they want to get solved, or they're trying to buy something, or trying to sign up for a plan or change something. Um equally, they don't really care what software you use. And we think that um to create these, like, you know, our kind of mission uh at Intercom with Finn is to create these perfect customer experiences. And to create perfect experiences, you need perfect being like the customer changed nothing about it. They oh, they knew who I was, they remembered me from last time, they've all my history, they knew exactly what to recommend, you know, the thing happened quickly, the problem was resolved, the action was taken. Like Finn can do very complicated things, take actions, write back to systems and things like that. Uh IDO being like a great example, like Finn will write back. Uh and so um, you need a single source of truth. You know, you need a single customer record. If you have like multiple agents, one for sales, one for service, one for success, and especially if these agents come from different companies, like company A, you know, you buy agent one off company A and agent two off company B, these agents have different context. They don't share the same context, they have different customer history, different memory, and um it'll end up being a big mess. It'll end up being just like the you know, a lot of customer space today is quite terrible because um you're handed off from tool to tool, from team to team, you lose context. The amount of times you have to remind people I I already told the other person, but here we go again. Yes, this is my address, or this is what my order number was. And so you're gonna re-re-create all that same crappy experience if you buy from like different vendors, yeah, because the memory and context will be different. And so we think it's really important that businesses have singular goals, and so they should have a single agent that can manage the entire customer communication with shared memory across everything.

SPEAKER_00 10:30

Yeah, no, that makes a lot of sense. And so, um, Will, you know, from your view as a CRM vendor as well as a customer, um, what does this experience really tell you about how CRM kind of is evolving with AI and maybe how it needs to evolve?

SPEAKER_01 10:46

Yeah, it's it's it's super relevant. Um, what Paul was saying there about having a single system of record, and that's really what we are trying to do at ATIO. Um, we have this concept of universal context where Attio is that central connection point for all of your business's context across all the different tools. And as I'd said before, Finn is actually able to write to Attio, and then Attio is able to write back to Intercom with data from the CRM. So they're basically cross-populating each other and sharing information, which is amazing. And then more broadly, um, we think you know, CRM should be something that's extremely connectable to AI agents. Um, that's AI agents that are built into the CRM, like AskATIO. Um, that's also MCP connections, so you can connect Claude, ChatGPT, Gemini, et cetera. Uh, and also having a great API layer, which means that you can basically write scripts that then perform actions inside of your CRM. And I think more broadly, we're we're also thinking about this more from kind of how can we make your CRM proactive as well. So not only kind of passively holding your data, but actively looking across all of your information and then suggesting what um next outcomes should happen or what next actions that you should take.

SPEAKER_00 11:57

Sure. That makes sense. Um and Paul, you know, obviously there's a lot of noise around AI and the implementation of AI agents, but beyond that, how do you genuinely see AI agents transforming the customer journeys over the next year or so? Um and why should people remain cautious within that?

SPEAKER_02 12:16

Yeah. Um I mean, there's a lot of AI is hyped, you know, in lots of places for lots of reasons. People have vested interests in it being big and successful. But like in my career, I think this is the single biggest technology change we'll probably ever experience. You know, like I started my career in the early internet. Um, I saw either front door to mobile. I worked in the mobile team at Google when the iPhone came out and we were building Android and the first versions of apps. I think this is way bigger. AI is just way, way bigger than either mobile as a technology cycle or the internet itself. Um, because AI does work. And AI can do things humans used to do, and AI can do things humans can never do. Like I said earlier, you know, there's no version of a human that is excellent and expert at sales and service and success. Like those things have a very domain-specific set of skills and knowledge, but AI can do it, and so it's creating these experiences that are um that were impossible before. You know, you could never have a mission that says we will deliver perfect customer experience to every single customer every single time, 24-7, 365. You know, that's that's like impossible, not even like a dream, you know. And so AI can do this. And when people put in the time and the effort to invest in doing it right and setting out it properly, and persevering at times and trying things, you know, a lot of successful AI implementation requires experimentation. We we build FIN with a very experimentation, you know, experiment-first mindset. We experiment constantly different ways that the Fin can work, whether it's at the models layer or you know, further up the stack at the application layer. And so I encourage people to um invest the time and believe what's possible. We have so many, we have hundreds of stories of true transformation in companies. Um, like you know, Will said there, like you know, ADIO has seen really great results with Fin for Sales, and Fin for Sales is very early. Fin for service, you know, we have complete transformations, amazing business growth off the back of Fin. Uh like leading AI companies like Anthropic and Clay and Adio as well are using these products as their kind of first step, uh their front door. And so the reason that that that the best AI companies are doing that, so it's real. You know, the hype is justified. But in terms of remaining cautious, you have to put the work in. You know, like people are looking for some kind of silver bullet magic thing. They do an AI project, an experiment, to prototype, it fails, they kind of write it off and say this thing doesn't work properly. People are worried about you know, AI doing jobs that they have. So if you're a software engineer, suddenly clawed code is a lot better at your job than you are. So that's obviously worrying. Um, but we have we're like fully clawed coded here, and um our software engineering team have never been more productive, you know, and it's exciting. So people should be cautious to kind of write it off. You have to put the time in and you have to put the effort in, and you have to be prepared to fail. Experiment, fail, try again, and and you will see success. If you persevere, you'll see success.

SPEAKER_00 15:23

So that's right. Um, and that leads nicely into um my final question uh for both of you, um, but maybe stick with your pause since you are on this um practical way, ways that people should be thinking. So for our audience who is, you know, a CX buyer thinking about you know implementing agents, if you could offer one piece of advice for how to approach it, what would it be?

SPEAKER_02 15:47

Don't wait. You know, like don't wait. People I hear all sorts of reasons. Like people never say no. You know, people aren't gonna say, like, no, AI is a fad. Like, we're we're beyond way beyond that point. But they'll say things like, our data isn't ready, or oh, we first have to sort out this organizational thing between these two teams. Um don't wait. The longer you wait, the more your competitors are gonna invest in this and they're gonna see huge success. Like we've again, so many examples. Um I'm sure Will has with audio too, so many examples. Like if if you're kind of thinking about it, you're already way too late. So I would like really, really implore that people get on it, try it, take risks, experiment, um, and stop waiting.

SPEAKER_00 16:37

And Will?

SPEAKER_01 16:37

Yeah, I mean, for yeah, yeah, absolutely. Um, I think what I really liked about uh Paul's previous answer there was like the the effort point, the the effort point there. And really, like my one piece of advice would be like get the context kind of synthesized for an agent. Um as I had said before, ASIO is your single system of record, and that's what we're we're hoping to be for businesses. Um and what what I was able to do with um with Finn is we have a pretty extensive help center, but that help center content was was uh basically the service agent's training set. Um so not great for like um explaining how the product works, but not perhaps so great at explaining the the unique selling points, et cetera. So for for me, it's like putting the effort in to start with to get an agent like Claude, for example, on research mode to go ahead and scrape all your publicly accessible information about your customer case studies, your features, your help center, your product pages, and synthesizing that all into kind of like these context documents for your agents. Um and then basically just iterating and then using your internal systems of record like ATIO or your your own CRM solution to look at your sales calls and think, right, what's the what's the tipping point here? Like what actually sells customers on our product? And the more you feed of these pieces of information to Fin SDR, the better it will act on your website. So getting the context right, I think, is my number one piece of advice.

SPEAKER_00 18:07

Absolutely. Well, that's good advice and good a good place for uh people to start when they're thinking about this. So thank both of you, Will and Paul, for your time for joining us and giving us your uh experiences and perspective.

SPEAKER_02 18:20

Amazing. Yeah, thanks for having us. Thanks.

SPEAKER_00 18:23

And to our viewers, uh you know, to continue exploring the implications of a genetic AI and customer experience, you can find related stories and videos on cxtoday.com. And to keep the conversation going, join our LinkedIn community. Thanks for watching.