How Cleo Reached $400M ARR: Barney Hussey-Yeo on AI, Growth, and Building a Consumer Fintech Unicorn

Welcome back to writing unicorns today. We're incredibly excited to have Barney founder and CEO of Fintech Unicorn Clio on the show. Barney started building Clio as an AI product long before the current LLM hype cycle. So we're really interested to find out where that inspiration came from. And we're going to hear more about the journey of going to the US and scaling from zero to now $400 million ARR, which is amazing. So Barney, welcome to the show. I work at Red Bus Ventures with Simon, who obviously very happy investors in Clio. You were super ahead of the curve when it came to building with AI, particularly for consumers and on the application layer. So what was it that kind of gave you that foresight that this was going to be a massive opportunity? Yeah, well, thanks for having me, but I think the thing that I was really lucky to do is go to a really good CS department, do my masters in machine learning and I think being in a computer science department like a good one is probably like the most interesting place to get ideas from, because all of your professors are right at the edge of their fields. And they're all kind of like 20 years ahead because you've got to have this fundamental research that is nailed and done way before commercialization. So, you know, back then they were talking about bioinformatics, machine learning was my course, everything was machine learning. It's robotics, swarm robotics, so these are projects on swarm robotics, which is now like a bit scary because I know what is happening in the world of drones. But it was this fantastic place to learn what was coming. And if you look to the research happening computer science at the time, when I started in the early 2010s, it was all vision. So we just had neural nets, a weird beaten, I think it's the Alexa net benchmark. And that was the exciting thing, but just as I was kind of doing my masters and after natural language processing became dominant research helper and if you looked at all of the H and X simulation, it was coming from natural language processing. So I was obsessed with how machine learning could be applied to AI assistants and to language. And then, you know, everyone, you guys, the world kind of woke up when the 3.5 was launched and you had chat, but anyone kind of paying attention. And you about the transformer architecture, two, three years before that, everyone knew in machine learning, this was coming. It was just that was a catalyst moment for the media and kind of the lay people and it's now obviously the biggest thing in in the world and it's going to transform society, our economy where we live for good and for bad and it's amazing time to be building. I want to get on to your like insane revenue growth, but just before we get on to that, I think people need context. Explain like what you guys actually do. So clear as an AI assistant feel money very simply, we're trying to build the financial companion, the assistant, the coach for a billion people. We're building an AI that feels like a friend feels like a human. It becomes radically personal as to you as you interact with it and you speak to it and it learns from your data and from the context that you give it. Basically it becomes a always on, always available 24/7 proactive assistant that is kind of pushing you the exact right thing to do all about your finances every single day. So it's helping you plan for you really long term horizon events, whether it's getting a mortgage retiring, whatever it is, it's working back from whatever your goal is to give you daily actionable insights and advice to help make you a little bit richer every single day. Love that. Hopefully all of our listeners are going to be users and that much richer as a result. But yet so on to your like wild revenue growth, especially in the US. Can you talk a bit about that growth and the operational challenges that it comes with and maybe start with those I want to get on like the psychological sort of shifts you must have experienced, but just talk about the growth and the sort of operational tensions. So the central tension at the start was there was no revenue for five years we had made no revenue we literally just focus on doing the AI system just getting we kept the users. It was like a pretty out there strategy. I'm not sure I'd do it again was pretty brave. I was pretty young. But I was kind of adamant that we're just going to build like a world class product, make people love it and then we do revenue later. We're raising capital in the early days. It was the ZerPera right so capital was kind of like freely available. It was easy to raise. I never had to do a proper fundraising around while I went out and pitch people. It was always people wanted to preamp the round and just give me money. So I thought I was hot share. I wasn't it was like the macro and the ZerPera. It makes you a little bit overly confident. So we had the reckoning with the covid which is when I raise a series B and that was a tough round because it was just before covid we work had crashed. Uber just crashed like the entire landscape was changing from ZerPera to covid and I raised 25 million pounds at that point and I was like, oh my god. We need to build a real business model and burning two million a month. I've got no revenue. I've got no business model. It's time to go. So that was like a 180 shift. Just as lockdown changed our lives, it definitely changed the business as well. So it was a pretty intense kind of couple of years but we successfully made that up. Just from a feature to a kind of global product with lots of product lines that are kind of tens of millions and revenue each and so much more diversified and real enjoying business today. Yeah. So you switched on charging for certain products. What were those phases of growth? But you know 400 million hour or so like what were their distinct phases. Do you think they took seven years to get to 100 million. So it's not like an overnight success like seven years to get to 100 million hour. That's actually kind of standard. If you look at the data. There's some AI companies these days that do it like two days but that wasn't what was happening back then. But seven years of long time you're grinding out to get to your first hundred not profitable in eight point during that journey. Really grinding to get to the first hundred once we got a hundred though. The next hundred was in like 19 months the next hundred was in like 12 months the next hundred was in like not. I mean it's just it's an exponential and it's a machine and it just keeps growing to the first hundred million is absolutely savage. So the hardest thing you're ever going to do in your life. It's brutal in every single way. They're kind of subsequent hundreds exponentially easier to go on. And Bonnie the products now available in the UK which is very exciting. The UK launch is off to a good start. If things are sort of easy and now then they were when you're any had a month's runway. How are you preparing for the shift from being slightly under the radar because your customer base and revenue was very US solely US really to now. So you will very quickly be seen as one of the UK Fintech icons and that shifting awareness around Clio in your local market. How are you preparing for that personally. We've got this great talent brand we're backed by the right DCs like Clio is always in the press and stuff. No way for good reasons. And people know we're like a top tier company and I think it largely comes down to like if you look at LinkedIn you just look at who actually joins Clio. Everyone's really been successful in their careers. It's a very high quality group of people and very smart people that's highly research focus. So it's just in the UK. It's like has a great track record for hiring. It is she's pretty easy for us. The harder thing is going to be kind of you know we're going to do Australian Canada and France and so they're going to be harder markets for us. And that's where the language is culture regulation that all comes into play and is a little bit different. The UK and the US are fairly similar in some regards. It's going to be more challenging doing Japan or later than it is here. No, I'm excited for all those challenges. We want to be global. I don't want to think about just two markets. We want to think about a billion people and being something's pervasive around the world. Awesome. And with consumers becoming more and more used to working with LLM's or chat AI products. How has that helped the Clio story as well as people become very accustomed to using products like this. And also what are the sort of data modes and defensibility that you're baking in to ensure that Clio remains as the leader in the sort of finance advisory space. Yeah, it helps on helps on acquisition. Sooner's chat you can came online or cost of acquisition came down because people willing to use these products and kind of got more custom to them. So if you look at our growth since 2022, it's just been like a rocket ship proper rocket ship and it's become a lot easier for us to do. So, you know, I've got to thank them for that. And obviously, you know, when I compete with me through the journey, but you know not to get too technical and nearly on like the data. We've been doing this for 10 years, right? And we've had billions and billions of conversations with users. We spent millions and millions annotating and getting evil sets for what works and Clio is not like chat to you in terms of all of our interaction really comes from push. So we're a proactive service. So we've had recommended systems and behavioral scientists and people working to decide, what do I say to you at what points in your journey to help you take the next back action. So there's massive data, though, but this massive kind of products advantage and it's compounded over time and that just really helps us build a world class product. So you've got that and then you combine it with all the financial products we've built and successfully scaled. So we don't just give advice, but we take action. We can help you get cards, debit, credit cards, find our pay later, savings, wealth management products like the sweet is there now. So we can monetize success than anyone's come in, which if you are starting from scratch, you're a little YC startup. It's much harder thing to do and get a real economics pretty quickly. There's a contradiction I want to explore because you come across as pretty laid back. And yeah, I imagine you're not to build such a successful business. And you sort of talk about how it's kind of easy now. But then, you know, pace of execution, I'm sure is still super important to you. I know you've talked about sort of Claude code and how AI is helping you and probably your organization. But like, how are you thinking about speed of execution as a CEO now? My kind of default pace is pretty fast. I talk fast, I move fast, I get everything done pretty fast. I like being intense and moving fast. It's kind of all of my leadership team, a shipping code. And it's not just shipping code to like our production app. It's shipping code to make the organization work really effectively. When you have 550 people globally, you have information just all these different places, granola notes, notion, emails, slight mode dashboards everywhere. And being able to pull that all together to visualize that, to automate, to make everything run really smoothly, I find the like building the OS of the company today is like so grassifying and so interesting. So I don't have to have all these individual meetings where I gain context and have pre-reads and it's slow. We can now have agent agent conversations, we're exposing all the data in really interesting and unique ways. And then we're building agents and tools to be able to interact and to do stuff with them. Running a business right from the FPNA, are you hitting management cases, down to the very individual engineer shipping. You can go all the way down the metric stack to the execution stack. You can have it all instrumented. And you know what's on track, where it's going off track, you create visibility on your business these days. So I think the companies that go exponential are going to be the ones that take full advantage of this and get full on it. So if you're not 24/7 in a code code, you don't spend 30k a month like I do on code code, you're probably going to get left behind. It's great for me because I'm like native to it, but I think it's going to be a real challenge for the legacy businesses. Imagine you're running BT or something. How hard it would be to change that organization and make it AI first. But yeah, it's going to be a big divergence, I think, in output in these companies. Yeah, that insane that you're spending 30k a month on Claw code, but I love it. What do you think your business would look like today if you had not been using AI as power users internally? You know, yourself, you talked about your senior management over the last 12, 18 months. Have you got any grasp of that delta? This only got really good since Opus 4.6. So Opus 4.6 was like November to December time and the Clawed CLI. It genuinely, maybe it's 4.5, but it was like genuinely, there was this kind of like our hard moment in coding where it's like, oh, this actually works now and this really works. So it went from being this kind of per-programmer thing, which you'd see and you'd kind of work with it and you still have to be a software engineer, essentially. So this thing where it's like, oh, it can actually do tasks end to end. And if I chain and I do recursion, it can do really meaningful like long term tasks. So it's probably only the six months we've seen this acceleration. I've definitely seen an acceleration of my token spend and all the companies that can spend. But it just means we are shipping a ton these days, like way more than we ever have. Yeah, we've had the kind of the cultural adaptation to it as well because we've got 170 software engineers. Not all 170 days people jumped on AI and people wanted to hand right code, filled months, long as they should have. So it's taken us time to be able to get all those 170 to be fully AI native, like all the tooling that we have, Cleo, everything is, you know, MCPEed and got all the right kind of platforms for it. But that was a bit of a cultural shock. And the people that aren't and haven't been using it, they're the people that we've had to be exiting from the business now. So like there's a real productivity gap between the people that have gone full into AI and the people that are still hand reviewing code, hand writing code. And I think you're going to have, you know, you know, a couple of years, you'll be like a marketing tagline. We still write our code by hand. And they'll be like the bespoke software agency is still hand coded because it's just changed so much. And if you think about the output of these companies is software, it's the thing that really does matter. And it's the thing that really does drive GDP growth. And I've got so many follow up questions. But one of them was around like token spending. You obviously mentioned some of your own token spending. How do you encourage adoption and use of AI whilst also keeping a control on sort of excessive use or maybe even personal use and things like that? We have a budget of two grand per engineer. And then it's like a grand per everyone else in the company and leadership can do whatever they want. I can't say something. But the business was really interesting though, Bonnie, even just having that policy is a head of where most people are. Because what I think I've seen from speaking to founders is that they have employees that are desperate to try and use AI tools. They hit the free tier or the token limit or they meet to upgrade the package. But they haven't quite yet validated that this thing that they're playing with is going to deliver the outcome they want. They've got a choice to make. They have to expense it without knowing if it's going to deliver a result or they pay for it themselves. And then if it works, they then go and expense it. But there's kind of these structural barriers to adoption. But it sounds like you've solved that by giving everyone a play allowance. And then from there, they validate. And then if it works, then you can scale it up. Yeah, I definitely think there's a company to be built here like token economics, which is like the middle layer where it takes all of your problems. And then it kind of looks through what were you doing? Was it useful? Were using the right model parameters? Were using like too much of an expensive model? Could you use other things? How could you optimize all these? Generally, as you optimize a SQL query, probably don't actually, but as you optimize your code, there's ways of profiling it. And I think there's a similar thing to be built now with tokens. We're trying to build it internally. I think it's the company to be built. We look at people that are working with them. I also think particularly for non-engineering teams, for the sales and growth teams that are experimenting with 100%. Higgsfield for AI influencer creator and stuff like that. People are learning how to use these tools. It's not as simple as just typing something in, I really don't know if you've played with Higgsfield, but it's really complex. And you like, an hour's YouTube tutorial, and you're still not using 90% of the features. So people kind of play and then hit the barrier. And then it's not really like proven that it's going to deliver the value, but yeah, it is a whole business problem in its own right. Understanding where the usage is being spent, the productivity of that usage, and the translation from that tool through the rest of the business and things like that. Yeah, and so there's a big bit of phono in startups and the best companies are spending a lot on tokens. And here are the people doing 10k budgets. It kind of becomes a summary. And you're like, "Shit, I don't want to get left behind. Why is my budget 2k? Why is it not 10k?" Jo, I mean, so like the best startups are pretty well capitalized and they've been pretty just all in on this, which means that some of it is probably excessive and not that useful and not optimized. I think there's a rationalization over time with the business we talked about. But the next year or two, token spend is just going to, I would love to be in it early. I know a seed investor in Anthropic. I'm very jealous. I think it's a million dollar seed investment that's 100x already. But it's going to be a pretty wild couple years. Any favorite tools other than Claude? We use MetaDee, the Clio, Ham and Angel investor, but it allows you to look at all of your talent, all your interviews across the business. So it records it like granola notes. And then it kind of lets you compare to the actual performance in three six months time of those people and correlates what is actually happening in your interview process versus reality. So it allows you to pinpoint all these people are way too lean and it just makes everything way more efficient, having these notes and pre reads or generated for itself. I'm very bullish on that. Another one is stacks.ai, which is helping us close on a book within a day versus 10 days. You've done management accounts before. And then you've got to have all these people analyzing all the transactions and cash browsing it. Stacks just helps you get to management account closed and just automate your time financial stack. And it's like, you need two less people get the data in a day versus 10 days. There's loads of these companies being built at the moment, which I'm super, super bullish on. And you don't have to build everything yourself. You should still be buying SaaS software. It's still a good thing to do. But these ai-nated ones are highly useful. For companies that are well-capitalised and that are doing really well, I think founders and management feel it's fine to give big budgets and to spend on lots of SaaS. I think it's the middle of the road companies or the ones who are earlier on in their journey and still being super frugal, where they're perhaps making a false economy of kind of restricting token spend and saying, let's build this ourselves, let's not buy, which perhaps people have to just psychologically get over that ramp. Yeah, sounds so British. It's like a civil city, civil service. I really do. We should be optimising for success, not optimising to avoid failure. The people who have really taken risk and lent into the technological waves, they've been the ones that win. What would you advise younger founders, things that you've perhaps learnt on your journey? People often get caught up on improving the story in the pitch deck and they're like, oh, if I just tweak the way that I'm saying it, though, if I just make the pitch a little bit more concise and a little bit punchy and I'll change the narrative in this way. that way and then the VCs are going to give me money and then they're all going to come clamoring at my door. I'm really sorry to say sometimes it can work like that to be fair but like the reality is you've got to build and if you have not built a great product just got that build. Just don't talk to investors, don't talk to the VCs, play hard to get if anything tell them you're shipping, tell them you don't have the headspace for someone of such low IQ. Honestly that is what gets VCs going. Do the guy that just raised like a 4 billion valuation, the deep-mind founder that just came out raised the billion. He was pitching Tier 1 fund and he said all of the partners there, after like five minutes I just don't want to give any headspace to venture capitalists. Sorry, I need it for more important things. Can you talk to my chief of staff? He walks out the room, puts in his chief of staff, shocker that Tier 1 fund offered a turn she within like 10 minutes. Got rejected. Anyway just fucking build, just build a great product, don't care about investors, play hard to get if anything and get numbers, build something great. Honestly if you put numbers on the board, you build a great product, it ends up working out for you. Sometimes there's macro cycles, yeah but eventually if the numbers are on the board, are you going to get paid? So just print numbers on the board. So the first final question is our future unicorn prediction. So yeah if there was an early stage company that you thought had a good chance of going all the way who they be. I've got four. Is that all right? Can I do four? eloquin.ai. We're using it clear. So there's lots of these kind of agents for CS that are automating thing. The difference here is that they use computers so they can watch your CS agents do a task. It learns from that and automates it. It is automated like 30% of our most gnarly CS tasks where you had to like click around, move load of stuff around which working in production, it's working at scale. Amazing founder. I think that's going to be, that's already going to be huge. Jack and Jill eloquin.ai. Eloquin.ai. Jack and Jill it is an AI agent for kind of recruiting. So you have an AI agent that helps you with your job search and then you've got one for finding talent as well and it matches the marketplace. One more reasonable. They just raised 8 million for kind of pre-seed. The founders are truly exceptional. One of them was a professor of machine learning at Cambridge and they're building superintelligence for coding. So I think those three are all going to be multi-billion hopefully trillion companies in the future. And you said you don't have a fourth. Clove is back like sell. They are crushing but it's in my space. So I've got, you know, I know a lot about it and thought a lot about space but it's going after the ultra-high network. So Clio's at kind of mass market, mass market affluent. They're building like the personal wealth advisor for people with a million, 10 million, 100 million in capital and then the difference is they're using humans but an accommodation of AI. So like this says, if you've got all team of millions in the bank, you're probably going to want to talk to a human but you're going to probably want the intelligence of AI. So it's combining with the two best things and I think they're going to crush. Okay and then the final question Barney is our dinner party guest game. So if you could have dinner with any three people, would they be? I would love to go back to a credible leader like Obama. I think you could learn a huge amount from him. Again on the politics one, I'll be great to just bring back Churchill for just, you know, talk about what was going on there. I think that'd be an epic one. And then maybe you bring Alan and Choring. You bring him back and you would talk about AI and what's happening today and get his take on it because I bet you would have the most kind of interesting and insightful things to say. If you told him about the future, you'd probably like build upon it and actually come up with a new revelation even being a hundred plus years gone. So you might agree. Bit of a weird eclectic mix of people but I think we'd get on. I think you'd be a good conversation. Awesome. Thanks so much for sharing those though. They are all people that have been mentioned previously as you can imagine because they're pretty like go to names. Awesome. Barney, thank you so much for coming on. It's obviously great to have you on with what you've achieved. Clio is just a rocket ship now. You've went for the long term plan and it's great to see and obviously it's now available in the UK. So anyone listening can go and download it and give it a go. I've got it. It's really amazing how easy it is to connect all your bank accounts and get such interesting insights. So yeah, it's been great to hear how you actually have gone about building it and some of the insights. So thanks so much. Well, thanks for having me guys. It's been a pleasure and I'll speak to you soon. That's it for this week. Thanks very much for listening. To stay up to date with the latest episodes, please follow or subscribe on your favorite podcast platform. We also have a newsletter called Reading Unicorns, which is another great way to get every episode direct to your inbox. And we'll see you on the next episode.