Breaking Customer Service Silos With Fin AI Agents

[00:00:00] AI Agents are only as strong as the data that they're given. When provided with an outdated data set, your agents could end up doing more harm than good. But not with Denodo. With an AI data layer built within your platform, your agents are provided with real-time data changes. So with Denodo, your agents can finally make the right business decisions. Simply visit denodo.com to learn more.

[00:00:27] What happens when an AI agent can resolve most customer questions before a human needs to step in? Well, Paul Adams is the Chief Product Officer at Fin and he's going to join me today to explain why customer service might be the first department to feel the full impact of agentic AI. Now,

[00:00:53] Fin, they were previously known as Intercom and today Paul will share how the company tore up its own roadmap and bet its future on AI. And we will discuss resolution rates reaching 70, 80 and sometimes over 90% and why sales and service boundaries make little sense to customers and how AI can remember customer history and business goals. He's got some pretty big things

[00:01:22] to say about that too. And Paul will also offer a candid account of rebuilding a SaaS company around speed, experimentation and making some unpopular decisions along the way. So if your support team is overwhelmed, this conversation will offer some practical reasons and lessons around why you should stop waiting for perfect conditions and get started. But enough from me. Let's get Paul right onto

[00:01:48] the podcast now. So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do? Yeah, thanks, Neil C. Hughes. Yeah. My name is Paul Adams. I'm the Chief Product Officer at Fin, the company that was previously known as Intercom. I've been here a very long time since we're beginning 13, 13 years. It's going to 13 long years, but some of them felt quite short. Yeah. Thanks for having me on. I'm excited to talk to you. Well, it's a pleasure to have you join me.

[00:02:15] I mean, 13 years is a long time, especially when you look at the explosion of AI just what three, four years ago in the mainstream, at least. Tell me a little bit more about that backstory, the origin story and how Fin AI become what it is today. Yeah. I mean, over 13 years, you've got to see many ups and downs. Intercom, I don't know if some of your listeners might be familiar with Intercom, but originally a kind of a breakout SaaS success. We were hitting all the

[00:02:42] milestones you expect from a fast growing VC back company, maybe 10, 12 years ago. And then as the company scales and grew, we kind of lost focus, honestly, went down the IPO route. And we, you know, we focused, we've tried to go up market and we tried to broaden the product offering at the same time, which in hindsight, wasn't the smartest thing to do. But then two things happened. One, Owen, who's our co-founder and CEO, he'd stepped out of the role for a couple of years. He came back

[00:03:11] and then about two months or three months later, ChatGPT showed up and we bet the entire company on AI. We changed everything, everything imaginable. And I mean, it's very literally, we ripped up our strategy, ripped up our roadmap. We had a machine learning team since about 2015. So we had kind of insight into, into the fact this was the real deal. And yeah, we took a huge risk and thankfully it's

[00:03:37] paid off. We've, you know, we built Finn, launched Finn very shortly after ChatGPT was launched. Finn's an AI agent for customer experience. We've had tremendous success since. Fantastic news. And fast forward to present day and everyone's not only talking about generative AI, it's all about agentic AI and agents now. So tell me more about how Finn has moved away from maybe some of the more traditional software categories and towards AI agents and what motivated that decision

[00:04:05] too. Yeah, we, you know, I guess, especially leadership level in companies, you make decisions based on what you believe. I'm of the opinion that you should have very clear beliefs about the future and their beliefs. They may not pan out to be true, of course, but we're like deep down the rabbit hole. You know, we believe AI is going to be one of, if not the most transformative disruptive technologies of our lifetime. I personally think it's far bigger than mobile,

[00:04:32] the kind of last technology shift we saw, bigger than the internet. We can get into that if you want, you know, but we kind of bet the entire company on it. And originally Finn was an AI agent for customer service and it did all the things you'd expect. It answered customer questions and then started to take actions. And over time, it got more sophisticated and more powerful. But then we started seeing people use Finn for other things like sales. And so suddenly these silos that have been

[00:04:59] inside companies for decades, you know, sales versus service versus success, like these, these departments sometimes don't even talk to one another, but the customer of course doesn't customer just experiences the company as a whole. And so we started to see people use Finn in ways that were more, I guess, customer first. And so we evolved our vision to turn Finn into a customer agent, as opposed to a customer service agent. And now we've like pretty strong conviction that

[00:05:25] businesses in the foreseeable future, some are doing it already today, will have a single customer agent. This single customer agent will do all customer communication, whether it's sales, service, so these things blend together. So it'll all, it'll deeply understand the business and context. It'll deeply understand the customer, their history, and it'll use guidance from the company to try and work out what the right thing to do. Sometimes it's to serve the customer, sometimes they want to try and cross sell to the customer. And so that's our vision and we're

[00:05:54] building it out. Awesome. And just to drill down on that a little, for anybody listening, hearing about you guys for the first time, they hear all the noise that surrounds agents at the moment. What does a single customer agent, what does it mean for how a business operates? And maybe if you bring it to life with any use cases or examples that you've got out there too.

[00:06:15] Yeah. The biggest, I think that with AI generally, but certainly for our space, the technology is really far ahead of what organizations can organize themselves around it. So I often say to people that technology changes fast, but people change slowly. So inside companies, there's all sorts of reasons,

[00:06:39] barriers to change, reasons for slower adoption and things like that. And companies are set up in a way, like I said earlier, that hasn't changed in a long time. So sales and service don't really communicate internally all that much, or certainly don't think of themselves as strategic partners. But customers, A, don't care. They're just trying to get the thing done, get the problem resolved, or try and buy the thing they're trying to buy. And so I think what we're going to start to see is a necessity for businesses

[00:07:08] to start to break down some of those internal silos. And that's what we see. Like we have, you know, really amazing pioneering companies do like, you know, Anthropic use, Finn, with companies like Calchi and Clay, like Clay are another really, really great AI technology company. And you start to see with these companies, they don't really think about departments in the way that say a 10-year-old company might. The lines are much blurrier. And so I think to succeed in this world,

[00:07:35] companies are going to start to break down internal silos and start to collaborate more. I think roles are going to change. We see, we have Finn customers who are getting 70, 80%, even sometimes in the 90s, of their customer queries resolved by Finn, which means that people don't have to do that. And so people start doing other things, customer success, instead of just frontline customer service, which is better for the business. It's more interesting job for the person too. But you know, they're suddenly, they're adding value, they're kind of revenue generating

[00:08:04] functions, but they blur into the other worlds, like the success world blurs into sales. So I think companies are going to have to look at their org design and ask pretty hard questions about what roles are going to look like in the future, whether some of these roles will blur, some will be removed, they're going to have to add new ones and so on. And I'm curious, when you think you talk about companies like Anthropic, I would imagine very forward thinking now exactly what they're doing and how to get where they need to be. But when

[00:08:31] you start looking at some of the bigger and older enterprises, a little bit slower to adapt to change and they arrive at your front door, what kind of questions are they asking you? Are there any trends in the kind of problems that they're coming to you with and asking for you for help? Yeah, it's a fascinating question. Because sometimes some of our most pioneering companies are actually quite old. You know, like their banks or like utility companies, for example.

[00:08:59] And because maybe because they're so old, like 100 years old, you know, maybe because some of these companies are so old, they are actually more open to change somewhat ironically than a company that's like 20 or 30 years old. So often when companies come to us, Fin is a very disruptive product. And I think to embrace it and get the value from it, you're already in some sort of early adopter mindset, or you've already crossed some bridge

[00:09:25] where you've accepted that you must change, your company must change. Sometimes it's competitive pressure. You know, they'll see a competitor using Fin or something similar, and they'll realize that they're delivering a better customer experience. And so they'll want to compete on that front. I think it's also too important to highlight here, we're not talking about replacing people with AI agents. One of the great stories I read before you joined me on the podcast today with, I think

[00:09:51] it was one of your customers that they boosted their customer satisfaction scores with their AI agent. Yes, that was great. But it also improved the experience when the AI agent then handed off to their human agent. So it's the combination of the two, right? Are there any other examples of that kind of success that you've seen? Oh, many. Yeah. The economics of customer service are kind of fascinating. We all experience pretty

[00:10:18] bad customer service. And like, this is why we feel our mission is quite, you know, every company has a mission. Some are more relatable than others. We feel ours is quite relatable because most of us, when we experience customer service, do not have a good time. Right? You're like put on hold, or you're passed from team to team. The second person forgot all the things or was never given the things you told the first person, paid staking, you know? So the customer service world is not good, but it's actually for most companies, not because they don't care.

[00:10:47] You know, a lot of, a lot of businesses are portrayed as not caring what customers, it's that they can't do it economically. It just doesn't stack up. Like if you've got thousands and hundreds of thousands or millions of customers, and they have lots of issues and problems and things to throw human capital at the problem, you know, it just doesn't work right economically. So a lot of customers of ours who are using Fin are underwater. They're just buried in customer queries and

[00:11:13] inbound customer problems and issues. And so actually what Fin does is gets them back above water and now they can deliver resolutions to most customers most of the time and actually deploy the people who were in the team to other things, you know? So it's not the case that we see, like you said, you know, teams being decimated with layoffs or we don't see that. Now maybe in the future, who knows? But so far we don't see that. We see people being redeployed to different types of

[00:11:40] roles, sometimes managing the AI systems. Sometimes it's things like, well, hey, the AI is only ever as good as the knowledge it has access to and whether that's actually accurate and up to date or the data connections it needs to pull customer data and so on. So people have to do that work. They have to design the AI and so on. But then sometimes it's things like businesses might have always wanted to have a phone channel for their VIPs. Hey, if you're a VIP, you always get phone, you always get through to a human, but that was just way too expensive to justify.

[00:12:10] And now it's not, you know, that will actually hang on. We could actually do things like that and create, you know, provide this white glove service. We could never could afford before. So actually we see people growing their teams more often than the other way around. And another big word this year alongside AI and all things agentic AI and agents is context. And just to drill down on the business value here that we're talking about, this technology can deliver.

[00:12:35] And it's almost like a version of customer support where you don't have to repeat yourself. So you could step away mid conversation, maybe come back three days later on a completely different channel, not have to re-explain yourself, pick up exactly where you left off. Just that single agent that remembers your situation, understands your setup, your context, and knows what you're trying to do. And that's one of the things that Finn memory makes possible for people listening, hearing about

[00:13:01] that for the first time. Tell me a little about that too. Again, we're back to the poor customer experiences. We all suffer too often, you know, people get handed off from team to team and you're a people can't look up all the information they need about you. So you've got to ask you things like your address. And obviously they know your address. It's buried in some system. System could be legacy 20 years old, very slow, you know, so as ask, it's faster and they'll write it in again and so on. And so people, you know, don't have access to all the right context and customer history and

[00:13:29] data and so on. Um, that's like kind of part of the problem, but AI changes that, you know, AI has not, not infinite memory, but for sure it has the ability to access systems, all sorts of systems, you know, big and small, well-engineered or poorly engineered, and it has access to, to all these things and it can pull from them quite quickly and synthesize information way faster than any human could.

[00:13:57] And so AI is way better at understanding the customer's context because it can do it far faster and far more powerful than a human ever could. So suddenly you've got this like just leap in capability and that AI, you know, modern AI, things like Finn can reason over this data too. And so you can feed it both customer context and this idea that FIN has memory or AI agents have memory, but then you can also feed it business context. Uh, again, like it's very hard to give a,

[00:14:27] 24 year old customer service rep, the business strategy and say, Hey, here's our 10 page business strategy. Here's our goals for the year. We're trying to optimize for revenue retention over growth or whatever, you know, are we, this geo is far more important than this, like no customer service rep, no leader, nevermind a kind of more junior person can hold that amount of context, AI cap. And so suddenly you have this again, leap in capability where the AI systems can both

[00:14:54] hold the business context and hold the customer context at the same time. And then you can give it goals. And so we've been moving FIN much more towards a goal oriented system where the business will say, here's the goals of the business. You now have the context from us, use a customer context in real time, you know, what, what reason over it and work out what to do as early, you know, we're still early in building out these systems, but so far we've really promising results.

[00:15:21] We see like we've e-commerce companies, for example, earlier, actually we've e-commerce companies seeing like all their metrics improve time to resolution has gone way down, but things like cart size has gone up, customer happiness and satisfaction has gone up. And it's because AI, and in this case, FIN is just better at doing these things. Wow. Just love that. So many great examples there. And I love the traditional metrics, but

[00:15:45] also combining that with the overall business goals, I can hear that setting off a light bulb moments around the world to business leaders listening and looking at the bigger picture. I mean, how do you see this approach potentially reshaping everything from CRM support, customer journeys and business roadmaps even as well? How do you see all this evolving and playing out over the next few years?

[00:16:10] Yeah, it's so hard. I think about the future a lot. I guess it's part of my job, like you do and talking to other leaders do. I think there are some things we can predict with high confidence. And then after that, it's a big giant, who knows, but you can kind of study technology. And if you study technology, the same pattern appears every time as in you breakthrough invention, the same pattern, you know, a breakthrough invention, slow initial

[00:16:35] adoption, mass market adoptions, people see value. And then the innovation cycle ends. All the companies copy each other. There's no real differentiation. And then a new cycle starts every single time. So we're kind of at the end of the mobile cycle, the iPhone, Android phones, they all look the same. iOS, the Android operating system, they're saying we've, I think we've a new that the day we're recording is actually we've an Android, latest Android announcement. It's going to be the same as the one I've an Android user. So the innovation cycle is over. Everyone's

[00:17:04] copied everybody else. But we're at the beginning, the very beginning of the next cycle. And I personally don't think enough people have internalized how early it is and how big it's going to be. You know, I think mobile, you know, many people started their careers kind of post internet in the workforce today. I started at the very beginnings of the internet. And then I worked in

[00:17:31] the mobile team at Google when the iPhone came out as in the UX team. So a front door seat for better and for worse to the very crazy disruption that that created. But I say to people like, you know, imagine the world, it's hard to, but imagine the world before the internet and the world after the internet. And, you know, we're kind of in the late nineties still, you know, late 1990s, maybe 2000. So all of the greatest, most disruptive change is yet to come. And I worry that there's people who

[00:18:01] haven't internalized that idea. And I kind of saying like, oh, you know, we're three, four years in now. Not that much has changed in my work. When I think about it, I'm still using most of the things I do today. We have, I'll give you kind of one, maybe more concrete example of what might change. We have a product called Operator, which is an AI agent for customer operations. And initially we were worried, not worried, sorry, initially we thought that Finn could do customer experience. It will talk to

[00:18:27] customers as author problems, but it can't do customer operations. That's just too hard, too complicated, you know? And actually it's turned out that Finn and Operator can do customer operations. It can do incredibly complex things, but the user interface for Operator is unlike any tool that people have used before. It's very conversational. And I think increasingly it would be for voice first. People just use voice as the most, voice is the most natural interaction

[00:18:56] mechanism, hence podcasts. We're talking and people are listening, you know? And so I think a lot of software in the future would be very voice first. We'll just talk to our computers and then very conversational. So all of the UI we use, menus, filters, labels, dropdowns, buttons, that'll all just disappear. And so I think that level of change is coming soon, like, you know, two, three,

[00:19:23] five years maybe, where most of the software that we interact with will look very different to the software we have today. But you know, like I said at the start, who knows? You don't really know, but certainly that's the bet we're taking. Yeah. Incredibly cool. Of course, when you look back to Google there, the iPhone, the rise of the mobile world and digital disruption that followed cloud and then now AI, did you have any idea that

[00:19:51] you would end up where you are now when you look back and join the dots? No, I'll give you some funny stories. I worked on the very first mobile version of YouTube. So we were like working in the mobile team in London. The mobile team was so unimportant that it was in London. It wasn't even in Metinview, the HQ, you know? Eventually, obviously, Android took over and so on. And Google did a great job, you know, all credit to the Android team. They've built a brilliant operating system over the years. But the first version of YouTube on

[00:20:21] phones, most people in the company were like, no one will watch video on their phone. No one. It's small. It's terrible quality. It's grainy. And then you get into arguments like, well, the phones went bigger, the screens went bigger. And then there's an argument that like, there will never be enough internet bandwidth to support it. Just like pure hardware limitations. We'll never have enough computing and wires and chips to support crazy high definition global video

[00:20:49] over for the entirety of the human population. Turns out we do, you know? And you get that with AI too. People are like, there'll never be enough chips. You can't build enough hardware. We've been wrong with that kind of stuff before. So in the early days, people said no one watched videos on the phone. Work was another one. I worked on like the first versions of Gmail. It was at the time I was at the company. It's quite lucky to get this front door. Again, no one will do email on their phone.

[00:21:14] It's a toy. You know, you see this stuff all the time. The iPhone is a toy. It doesn't have a keyboard. People need keyboards to do proper phone stuff. And you end up being wrong about it a lot. So yeah, you know, it's going to be interesting to see how this plays out. But I think that for a long time, most of us have worked in a very stable optimization oriented economy in technology. And it's hard to break your mind out of that. It's hard to think, well, do you know what? Actually,

[00:21:45] how we interact with computers at a very fundamental level is going to change quite a lot. Wow. Incredibly cool story there. And if we go back, I mean, there'll be a lot of people listening, looking at your incredible journey. They want to go on a similar journey to you, following your footsteps. And those leaders listening that want to do that, tell me more about what it took to to rebuild a SaaS company to be fully AI first. Because again, it's something a lot of businesses

[00:22:12] are chasing right now. Any tips you picked up along the way, what to do, what not to do? Yeah, many. We're here all day. I think the biggest, honestly, the biggest one that I tell people is, the biggest one that I tell people is, the change that we had undertaken into Calm, which then later became Fin, was brutal. And I use that word like very deliberately. It was really

[00:22:38] hard. We had to change almost everything about how we work. People are change reverse. People don't like change. And we had to push unpopular change through. And some people quit. We amicably parted ways with a lot of people and said, hey, we are betting the entire company on this thing. We're changing things. We're deleting things you love. We're changing processes you created. It's fair. And this is true for me,

[00:23:07] a ton of things that I had created at Intercom over the years, systems I had designed, processes I was really proud of. I had to suffer. It was too bad. They need to die. They're not compatible. We optimized for speed to market. So we obsessed with speed to market. You need to just get everything live as soon as possible in the smallest form possible to learn. So it's just this world in which I would

[00:23:32] encourage everyone listening to get things live, get them out into the real world. You can debate around the table and in the office all day, but you're never going to actually learn who's right until it goes out into the world. And so we had to kind of go through this brutal transformation and optimize for speed to market and just get things live. And we learned hard, failed in a lot of ways, made a ton of mistakes. But because this time is so disruptive, so new, so much change,

[00:24:03] you know, it's kind of the journey. It's a necessary journey. You're going to get loads of things wrong and you'll only really know if they're wrong until if you're out there in the market, you know, just get things live, make unpopular decisions that you think are right. They're some of the biggest things we had to go through. And for anybody listening, hearing about Finn for the first time, you're used by over 12,000 of the world's most forward-looking brands. Looking at leading resolution rates too, I think 76%

[00:24:31] is the standard across there with many seeing more than 85%. For anyone coming and wanting to work with you or finding out how you might be able to help them transform their customer experience, do they need to be in a certain place before contacting you or where do they need to be before they can go on that journey with you? Yeah, they don't need to be anywhere. You know, honestly, this is another kind of, I asked one of your questions earlier too, like another common thing. The only place they need to be is like in

[00:25:00] the right mindset, willing to say, let's do it, let's go, let's do it. I'm here, I'm here for the change. We need to change. And then like champion the change internally. Again, there'll be so many ways in which people say no without saying no. You know, they'll say like, well, not now, or it's the wrong time or, well, I promised the board this other thing. So I need to commit to that first. So they need to be in the right mindset. But after that, they just need to do it. And you know, one of the biggest things I hear from many companies, not just with Finn, but AI

[00:25:29] transformation generally is we're not ready. You know, our data is a mess and, you know, we need to clean all that up first and get our own house in order. And I think that's fundamentally wrong. You're delaying learning, you're delaying getting things live and you're delaying how quickly you can learn. So, and it turns out the data thing, I think is a bit of a red herring for many companies matters for some, for sure, you know, regulated industries, things like that. But most companies can get AI products like Finn live with messy data and see huge return, huge value. You just have to go

[00:25:59] for it. You just have to try, start small, get experiments live, real ones. And so with Finn, we're ready to help anyone who wants to do that. They just have to get in touch, Finn.ai, get in the right mindset and we can help anyone, any business, any size. Awesome. And I would urge anyone listening that wants to learn more about how Finn AI and a combination of their human teams as well to resolve every customer issue, no matter how complex,

[00:26:26] I know there's a bit of a myth that every business thinks, well, we do things differently and our industry is a little bit more complex. It seldom is, most businesses are very similar once you get down to solving those problems. I'll urge them to check out the show notes. I'll include a link there to everything, including your socials as well, which should feature some of the new things coming out. But more than anything, thank you for sharing your story and talking about all this stuff in a language everyone can understand. I really appreciate your time tonight. Yeah, thanks, Neil. Thanks for having me. It's fun.

[00:26:54] I think Paul's story today, I think shows why adopting AI is an organizational challenge, as well as a technical one. An AI agent can remember customer history, consult business goals and resolve routine requests at speed. But companies must decide how human roles change around it. Paul is seeing people move into customer success, knowledge management, AI design and higher value

[00:27:22] service rather than simply disappearing. And his advice today was quite direct. Start small, put in real experiments in front of customers and learn before your internal debate becomes a full-time occupation. And waiting for every data problem to be fixed may feel responsible, although it can also delay useful evidence. So a big thank you to Paul for joining us today.

[00:27:47] And you can learn more about them at fin.ai. But over to you. Could one customer agent improve continuity? Or would your organizational silos prevent it from succeeding? I think Paul's gave some fantastic tips and advice today, but I'd love to hear your story. As always, techtalksnetwork.com. You'll find 4,000 interviews, different places you can meet me on the road this year. I'm going to

[00:28:13] lots of conferences. And one of the reasons I attend these is not just to record interviews from the show floor, but to meet you. And I'm trying to meet as many of you as I can while I'm on the road. So please send me a message. Other than that, it's time for me to go now. So thank you to Paul. Thank you to each and every one of you for listening. And I'll be back in your podcast feed tomorrow. Bye for now.