From Formula 3 to AI Accountants: Ariel Harmoko, Co-Founder & CEO @ Artifact

0:00Hello and welcome to another episode of Riding Unicorns. Today we are delighted to be joined by Ariel Homoko from Artifact AI. Ariel, great to have you on the show. You've got a really interesting founder, sort of CV. So we're really excited to get into that. And also you're building at the forefront of AI in the accounting space. So it's going to be a really interesting conversation. Maybe in your own words, you could just introduce yourself. Of course. Thank you for having me, James. I'm really excited to be here. I'm Ariel Homoko, co-founder and CEO of Artifact AI. I was born and raised in Jakarta, Indonesia.

0:35parents divorced when I was less than two years old was brought up with my mom and my mom remarried when I was six and at the age of eight my my stepdad brought me onto the racetrack he was previously a rally driver and he put me on to one of his best friends go-kart rode me around the track didn't know which one was the throttle which one was the brake crashed into a tire wall and And, you know, my mom was like, stick to academia and like this sport is not for you. But my dad was very persistent on me becoming the next Lewis Hamilton. So I did professional racing religiously, started off with go-karts, traveled across the globe and eventually, you know, race across Europe.

1:20Grew up with Lando Norris, George Russell, three times world champion and eventually reach Formula Three. decided to stop racing. And just growing up, I was always passionate about the maths and sciences and being around mechanical engineers trying to make the car go faster, really draw me into like the world of mathematics. So I wanted to get as close as possible to where math was born, which was Cambridge in the UK and convinced my parents to let me move to the UK when I was 14 years old. I attended a boarding school in the UK and, you know, COVID hit, felt really vulnerable. My grandpa was diagnosed with Parkinson's, so I wanted to do something in the world of med tech and early diagnosis.

2:03And this was the era of convolutional neural networks and deep learning. So I came up with this idea to early detect malignant or benign cells in brain tumor MRI scans. I sent that across 20 different professors all over the UK. 19 out of 20 got back to me but one professor who I really admire Professor Alan Blackwell from Cambridge and he invited me over to the college Jesus College and then I was onboarded as the youngest machine learning researcher over at Cambridge partly working as a ML researcher but also working in a spin-off lab as a engineer we were trying to early diagnose celiac disease with the NHS they did really well I think they eventually sold the company but it was during that time, I think, you know, figuring out the idea is like actually providing value, creating value from scratch, really inspired me to become an entrepreneur.

2:56Didn't know what to do. Did a bit of work at JP Morgan, deployed the first internal GPT for software developers and equity derivative traders. And I think this was March, 2024. I got into this program called EWOR, E-W-O-R. my coach and the CEO of Ewar Daniel, he told me, why not you become a potential CTO for this German-based company trying to automate accounting in Germany? So I went and, you know, the tech stack was a human accountant and a spreadsheet, and they sold directly to businesses. And she wanted to build the whole tech stack from scratch and offered me 5 % to do it. So it wasn't a great offer but you know i just sat down next to accountants in germany seeing the kind of work they do and you know mundane data entry and reconciliation and that frustrated me and i went back to the uk called up my best mate from jp and yeah we started that company and now we're building artifact aerial brilliant story so let's dive into the racing driver part of your background first time firstly i'm a i'm a huge car enthusiast and if i wasn't six foot five which you probably weren't aware i would have probably gone down that route as well i'm still a keen go-karter uh so we'll have to get out on track sometime but um we we've also had lando norris's dad on the show we've also had a premier league footballer we we're kind of fascinated on the show about the links between sportsmen and women and entrepreneurs and in fact anyone at the top of their game no no matter what the industry is.

4:31Tell us a bit about why you think either your personality trait lends itself, lends themselves to both driving and being an entrepreneur, or what it is about operating at the top of your game in any of those areas that lends itself to success. Yeah, absolutely. I think discipline is a big one. During my time in racing, I didn't know how much sacrifice and how much time and effort that I would have to put in myself personally. Most of the work actually is behind the scenes, right? So you need to exercise almost every day. You wake up at 5 a.m. You're already in the racetrack doing practice runs with the engineers and the team is there to help you build the engine and also make the car really go faster than your competitors.

5:24So it's within your DNA as a team to win and there's no space for second place in racing. And it doesn't matter how close you are to the person right in front of you in the finish line, that person is going to win. So that kind of mindset is so much, like very much embedded into my DNA. And I'm so thankful for my dad for actually like putting into racing in the first place, because without that, I wouldn't be able to have that kind of traits and qualities as a founder, right? And drawing that comparison now, especially building in such a hot space and you know you're trying to attract the world's top investors top talent and everyone they're chasing after the same set of customers I think execution velocity is everything and and your team is everything as well so very grateful for that racing experience which now I can bring to just this journey of building a company.

6:20And do you think discipline is something you can nurture or do you think you're born with that I mean you mentioned your your parents got divorced that you know any sort of early hardship is often part of it you know builds helps build resilience from an early age but how can you and others build that sort of determination and discipline yeah yeah I think that's part of it right like having a chip on your shoulder and growing up and being raised by a single mother and I've always wanted to make my mom proud and prove you know my dad wrong that's part of it but I think surrounding yourself in with other excellent people and understanding what greatness looks like in different categories that's always been able to put me in a position where I can learn from the best and I think I say this a lot to my friends and my team it's like the more you know the more you don't and humility is a big part of that because in order to strive for greatness, you're always continuously learning and being curious.

7:24So I think the advice I have for others is always try to put yourself in an uncomfortable situation and try to learn from the best and always be curious where you don't put a limit to yourself. And how important is love of the thing that you're doing in being disciplined at it? Super important. And I don't think you can be great at something that you do without actually loving it. You know, there are certain scenarios and edge cases there where typically outsiders in building a company can succeed. And we've seen a lot of examples in the past and of great entrepreneurs being first time founders or even founders not part of the industry doing really well.

8:13but with an extreme sport for example like formula one i think you need to be all in or else you won't actually retrieve the maximum roi but love is like such a interesting quality to have in my opinion because yes you can be dedicated you can be disciplined and you can also do well without actually loving the thing you do so it's kind of debatable there yeah there's an interesting trade-off as well you know too much discipline around something can probably actually make you fall out of love with it like you know if you're sort of too obsessed with it you know too much of a good thing so that there's a sort of there's probably a sort of balance to be found but um ariel we would love to why don't you explain exactly what artifact does uh and we'd love to hear the sort of journey from zero and we can go into a few few things along that journey Absolutely.

9:07So we're solving two main problems. One is the fragmented legacy stack that accountants have to deal with day to day. Second is the staff shortages that accounting firms are currently facing. So the industry has growing demands. Everyone needs accountancy. And not a lot of people are applying to become accountants or doing the CPA exam. So the staff shortages are clear. And, you know, BPOs are unreliable. Offshoring that, outsourcing that to BPOs like in India or LATAM, it's super unreliable. So we're building an agentic accountant that is ledger agnostic. So we sit on top of your existing stack.

9:49There's no rip and replace. And you can essentially serve a lot more clients without adding more human headcount. We do everything from the data ingestion, reconsolation, ledger posting, categorization, so on and so forth. So it runs in the background. And I almost describe it as like an outsource, like bookkeeping engine for these accounting firms so they can scale. And the vision is to make these firms like an AI native accounting firm in the future, where most of the margins and revenue are actually coming from advisory services, whether it's tax, whether it's CFO level advisory, that's where the real billable hours actually come in rather than mundane data entry.

10:33So that's artifact. And what is it like selling to these accountants? I mean, it's a fairly old school industry. People are used to doing things a certain way. and have done for many, many years. And I assume it's probably changed a little bit since AI has proliferated so much over the last couple of years, even since you've been going. Tell us a bit about what the reactions, the response is like from accountants. Yeah, I'll just start by saying, like, it's one of the toughest industries to sell into. And before we even raise any money, like, we spoke to several VCs in the UK that, you know, Ariel, like, this is probably the hardest thing you're going to do over the next five to 10 years.

11:13Are you sure you're ready for this? Obviously with me having that previous racing mentality, I'm like, yeah, let's do it. But I've only realized that after building in this space for like six to 12 months. And yes, you require a bit of luck. Like we started off in the UK servicing. Our first customer was a top 60 accounting firm. And it's a very tight knit community. So everyone knows each other. And there are these like subgroups of accounting firms within the top 100 that are very much tech forward. And they've recently actually started hiring junior developers or even senior software engineers to join part of these accounting firms.

11:54So they can build NAN automations, lovable automations, automations which work, but it won't scale with enterprise data. So just looking back at what Palantir did really well with the whole FDE model, understanding that data integration mapping is super important selling to these firms because naturally your product needs to serve the clients of these accounting firms and close the books. So you have to consider different ledges, the different sizes of the business of these end clients, different industries, different chart of accounts, different tax rates. And building a product in the space is, I think, much tougher than building a product in legal AI because accuracy is everything.

12:41If tax rates are wrong or your account coding is incorrect, then it really just messes up with your books. So it requires an agentic system because then, you know, it's almost like a colleague for these accountants. And what the accountants have to do is just do review. That's where actually like the human in the loop only comes in. But to answer your question, it's probably one of the hardest industries to sell into. Yeah. And how did you go about this in the early days? Like were you, I mean, you mentioned FDEs, like were you being the FDE? Were you going into accountancy practices and sitting alongside them and watching their workflows and seeing how you could augment them?

13:25Exactly that. Yeah. I think FDE in the founder-led sales stage, it's another way of saying heavy customer discovery. And it's super important, right? Like you really need to know the ins and outs of the different workflows that each accounting firm has because different accounting firms behave very differently to each other. They have very much different client portfolios. So shaping that product and building that product roadmap carefully from day one is super important. And then it does require a bit of heavy service led FTE style motion initially, but I think it quite diminishes as long as you do incorporate the right self-learning algorithms.

14:06And what did that look like in practice? Did you, you know, you mentioned it was a top 60 accountability practice. Did you convince them to let you sit in their office or like what did it actually look like? yeah so just to reiterate like we sit on top of like the likes of xero quickbooks sage net suite and in the uk xero has a large market share so a lot of accounting firms are actually partnered up with xero and xero if you look in the website um they've got prizes for xero gold silver bronze for those accounting firms who are utilizing a lot of xero for their clients and so we looked up online and saw like, okay, we, the director of outsourcing within each firm is listed there.

14:50And I started just picking up the phone and we got lucky enough, like we pitched it as a digital AI accountant and, you know, with several ROI messages, but the product wasn't great at that point. Like it was barely ready. And we just got our first believer, which is the top 60 firm in the UK. And we work very closely with them. And ever since then, the company, you know, improve really fast in a very short span of time. And Ariel, most people will think of AI as like LLMs, like ChatGPT. It's very large language model focus. So a lot of document creation and things like that has been improved with AIs to date.

15:30LLMs are okay, but not brilliant really with numbers. So what was that challenge like? Have you had to go and look into different models and build your own customized different models? How have you gone for the infrastructure part of what you do? How have you pulled that together? Yeah, amazing question. And this is the biggest question we've had building in the early days because when we first started building this around September 2024, that was like GPT 3.5 and it wasn't great, right? Like even for standard language tasks. So I was all, I bet big on agents before agents was a thing. That's why I started this company.

16:13And I've always been a big fan of like, how do you augment and structure, you know, different agentic frameworks. I think the earliest breakthrough we've had as a company is understanding like the nature of the data that LLMs are dealing with. And it's, it's based on language as you know, it gets tokenized. It's not great with numbers, but what it's great at is actually code creation and code generation. So I think Claude had a big moment around that period, like between September 2024 and December. And data ingestion is so powerful with AI because, you know, extracting PDFs and all of that, I think it's, you know, it's quite solved now.

16:53The hardest bit is reconciliation, right? Let's say you have partial bulk payments and multi-currency and you're having to deal with thousands and thousands of like Stripe data. like that in itself it's like a research problem and we kind of accepted that early so we did a lot of things across self-learning there in order to improve a reconciliation agent per client of the accounting firm yeah and how how did you tackle that in terms of accuracy i feel like we can all be a little bit forgiving on like document creation if the wording isn't quite right or it uses us spelling instead of uk or something and you you still have to read through it and check it but you kind of forgive it because it's done a lot of the heavy lifting with numbers if they're not right you're you're angry right you're like this thing doesn't work so how did you get to a point where you're like okay we've hit an accuracy level where we don't just make people angry absolutely so the way you sell agents is very different the way you sell SaaS, then I think trust and within the UI UX and auditability is super important so that the accountants can actually trace the agent's behavior before they reach any kind of final output.

18:12So that's how we tackle that. They'd rather work with a vendor like us rather than having to build this insane system in-house because most of them have tried but failed. And they're all looking in the market to actually invest in the right horse. and especially in a very competitive market like accounting AI where there's a lot of noise, they don't know which one to choose, right? So you need to kind of like explain the ROI very well and give them a customer journey from now until the next 12 months. And Ariel, we talked about selling to this very conservative user group. How do you scale GTM in a group like that?

18:55you know it's fine when you're fde and you're very high touch sales process it's founder led they're getting to speak to the the top brass um but like when you're actually looking to build a go-to-market motion like how how are you starting to think about that if you've got to that point yeah yeah yeah we have a very lean gtm team i think pairing up with a great technical se technical CSM and even a technical AE those are the like the three things you need in the early stage at least to actually get from your like 1 million ARR to like 10 million ARR it is very customer service led because the product needs to very much serve each end client of these accounting firms the biggest unlock has to come from product because the more you kind of like build it, making sure you build the right interface so that accountants can actually fine-tune agents with your product, the better it gets, rather than you having to manually ask them, oh, what workflows do you want us to build for you?

20:00And that's not scalable, right? So you kind of want to have a base fundamental of self-learning that's really powerful. Yeah. And then you mentioned a technical AE there. Is your AE literally from a technical background and engineer yeah i'm not an accountant by background but i think this is where the outsiders kind of win it's like you're not you're not saying yes to the status quo and you're always kind of challenging the narrative of the space but it also means like it takes a lot more time for you to integrate and earning the trust of the industry and speaking the language so it took us a while, took me and my co-founder a while to like eventually attract a big four, you know, ex-KPMG or ex-Grant Thornton people to come join us who are also engineering by background.

20:50And these are like the sweet spot, like, you know, they can communicate to the core product team very well. Those pre-sale insights, they can even build custom workforce for them and they speak the language of accountancy. And accountants are just very much more comfortable with other accountants, right so when you have this logos up you're kind of having a good conversation already so yeah and you mentioned a new product there it's just sounded almost like a well you said n-a-n so it must be a sort of like lovable for accounting where you just say what you want to do it builds the agent and you deploy that against your stack and then it works is that kind of the future and Is that where do you think most AI companies will go, where they'll build things sort of almost custom, they'll build the rails, then they build agents, and then they build like a prompt interface that then anyone can deploy any agent, they've got loads of examples of different types of agents and structures, and then it becomes like a super agent for that vertical.

21:51Yeah, yeah. We haven't seen that much in a domain, but like Harvey and Lycora did really well with kind of like building almost like a chat GBT layer for law firms. And I do know like they have workflow builders now, but they don't at least like market or brand it like that. OpenAI has kind of initially teased their agentic builder, agentic kit. That's quite scary because, you know, the reason why they kind of build that was because they did this FDE led motion with really large enterprise and collecting those workflows. And they're going to productize these workflows, right? So we're seeing that more and more now.

22:31I do think the horizontally scaling workflow builders like Lovable or NAN doesn't fit for a very specific vertical, like legal or accounting. And I think you're absolutely right in saying like these vertical AI companies, whether it's us, whether it's, you know, someone new in a category will have to start building that defensibility and moat first through whether you're building custom workflows and augmenting yourself within these large enterprises, or even building, you technology with really powerful models that with high accuracy and only then you can start building like an agentic layer which you can serve to your customers i do see that as a trend for sure i was kind of interested with jumping around a little bit but i want to go back to go to market for a moment and just ask about internationalization because i've invested in a couple of tax business tax and accounting businesses and sometimes they're great businesses in their local market but they really struggle to internationalize because there's different tax and accounting systems in every country how have you tried to prepare yourself for that going forwards yeah there's a lot of example of this like in germany you know there was a lot of accounting tax startup that raised like more than 50 million dollars over the past five years and still managed to fail like just because the government has control of data of like this incumbent accounting software.

23:59I think Europe is definitely really hard to crack because of, you know, each country has its own regulation. That's why we kind of like started off in the UK and the UK has very much similar, you know, ways UKGAP and USGAP and IFRS. So we decided to, before launching in the US a couple months ago, we localized the product for the US market, whether it's retrieving and preloading context of the industry of different US GAAP regulations. And we do have a plan in terms of scaling very well in the US and start acquiring, whether it's AI native accounting firms or taxpayers and accounting players in EU eventually.

24:40But if it doesn't get there and we start competing with the likes of maybe even Penny Lane in France and they're entering the Germany market, we already have to start thinking about preloading context very early on of different regulations of these EU countries. Ariel, you mentioned acquiring companies there, I think. Are you talking about becoming a roll-up? That's one way of doing it. I think when trying to scale in Europe, the roll-up space is so interesting and private equity has been very active in accountancy. So most of the top 10 or top 20 firms in the UK and US are actually owned by private equity now, and they're acquiring smaller firms day by day.

25:25So we're seeing a trend there. You're seeing like Thrive Holdings and OpenAI partnering up so that, you know, OpenAI can help with domain specific models. I think, you know, it's not right or wrong. We've seen Harvey One as itself, like as a software company selling to legal. we have yet to seen any much great successes as an AI role at play so TBD but yeah that's one avenue we might explore as soon as we start scaling after the what would have to be true for you to go down that path I mean I led the pre-seed of LawHive which is an immensely successful legal legal tech which now acquires small legal firms and embeds a bunch of technology into them to massively improve their margins and does a bunch of other stuff as well but for for law hive that's the most efficient growth channel you acquire a legal firm and all of their customers start coming through law hive like have you thought about the assumptions what needs to be true for for artifacts roll up strategy to be the focus going forward yeah i think it all comes down to like margins and, you know, which accounting firms you want to target first.

26:42And it's all about like sequencing. I find that, you know, with bookkeeping firms, the margins are just very low. And it's a much a commoditized business. So I think it's good in a sense where you have you can start acquiring them and just get a bunch of like, you know, clients on board into your platform. But the question is more like when and do I want to become an AI native accounting firm versus just a software business that scales and sells to even finance teams directly eventually right and unlock a larger time it's more like you know having a lot of overhead on people management and all of that I think if I would have started my company again now like I would have just thought build an AI native accounting firm from scratch so maybe acquire a decent top mid-market firm layer AI underneath and before I acquire any smaller firms I make sure the standardization process is very well so these clients can be onboarded into like artifacts platform immediately and we can scale that in that sense so because we are we are talking to some firms who are just massive like 10 ,000 15 ,000 clients through through roll-ups and now they're having issues with standardizing the client's like system of record and so it's such a mess and you know that's that that in itself as as a software service like that can be like quite a successful software business yeah if you standardize that yeah i mean it's such an interesting debate for any company sort of disrupting a service business and service industry like which model do you go down do you build a full stack ai native law firm accounts of practice do you build a marketplace that provides some tooling to freelancers or small firms?

28:27Do you just provide the SaaS? And if you can reduce the operational complexity to a level where it makes sense and you can extract the extra margin that otherwise the people you're selling your SaaS to can extract, then it's really viable. But building a playbook for roll-ups at scale that doesn't become a huge distraction is a real challenge. Absolutely. And you've seen it with the company that you invested in, right? I mean, they're seeing great success or it's harder for them to scale in that roll-up game? Yeah, no, they're having huge success with it. So they do have a playbook. And I think it will be worth tens of billions, that business, because it's such a huge market.

29:11But you need some real operational wizardry to scale that. I agree. And Ariel, tell us a bit about your fundraising. So, you know, the fundraising journey for a SaaS startup is one thing or an agentic startup is one thing. That for a rollout business is quite another. But where are you on your funding journey? Yeah, you know, we raised about$5 million in total. You know, we've been operating for almost less than a year. And then we announced our seed back in March and we did like a public launch video in August. Currently, you know, gearing up towards an A round. but you know conditional on several things i think we're very happy right now with where we are but also we want to prove like with a successful new product launch as well i'm sure you guys know there's a lot of noise in the market so having that kind of new product launch publicly and successfully can be the greatest differentiator for us and it's all kind of me speaking to other people about how powerful the accuracy is and the self-learning component but unless it's you know publicly out there and completely self-serve for accounting so that's that's definitely the game changer and yeah i think you know we're probably on track to raise an a by early key one and then uh scale the company after that very exciting are you doing that now in the uk or us you mentioned that you'd moved recently yeah we have an engineering hub in london the quality of talent is just amazing.

30:47I love the UK, pretty much grew up there. But GTM wise, we're scaling it in the US for sure. We have, well, it's a fine split now. We still have more UK customers, but I would assume in the next six months, most of that revenue is going to come US. So definitely going to scale that GTM team in the US. And will the fundraise be US, do you think? I think, you know, it'll be an interesting mix. Maybe, you know, we find a, you know, co-lead between both US, UK, someone with a strong forehold in EU, but also started investing in US companies. Could be in multiple ways. But if we look at the market, at least like from a US perspective, it's like almost five times larger than it is in the UK.

31:35So it might be worthwhile to spend a lot of time with a US VC who has those enterprise kind of help and connections. Yeah, awesome. Well, it's very exciting. And you've clearly proven that when you set your mind to something, you will find a way to make it work. Despite the obstacles that get thrown at you, it's about breaking those down. And you seem to be very good at the sort of first principle thinking of just getting through those and moving through it. and it's really interesting and exciting so it's been so fascinating to hear about that we're just going to move on to our final two questions ariel so the first one is a future unicorn prediction so this would be another relatively early company that's you know you think has a lot of potential to go all the way yeah if i had to choose one right now um i think my future unicorn pick would be open router they're basically like strive for ai models a single standard compatible API that routes traffic across hundreds of LLMs and clouds.

32:39They raised a series A of$40 million from Menlo and Resin and Sequoia and doing millions of inference right now. And I think they're almost like over 100 million run rate. So that would be my pick. Nice. If only they had better signaling from their investors. It's a pretty good investor lineup there. and then the final question is our dinner party guest game so if you could have dinner with any three people who would they be yeah the first one would be richard fineman you know co-created qed explaining how light and matter interact he contributed to the manhattan project and invented like fineman diagrams and very much like this person who just think from first principles there's a famous youtube video of him like explaining when you slip you know when you're on ice just explaining about friction i think that first principle is kind of like thinking i i adopt day-to-day in building this company but also in my life and the other person would be leonardo da vinci i think he's a great renaissance man polymath you definitely need to be able to you know get your hands dirty on different sides of the business for example for you to learn really like the ins and outs and for you to have a contrarian view of the market.

34:02And I think being a contrarian in life in general, I think that's super important to really drive the business forward, right? He also inspired Luca Pacione, who is the father of accountancy and published the first book on double entry bookkeeping. The last person I would invite to my dinner would be Ayrton Senna, who is a three-time Formula One world champion, legendary in wet. I saw that onboarding of him in Monaco in the 80s, I believe. Just an insane driver and obviously very unfortunate of him to crash an E-mola, but he lives in all of our hearts and most of the F1 drivers also aspired to be him.

34:47So yeah. Awesome. we've had Senna once before from George Richardson at AeroCloud Leonardo da Vinci we've had about seven or eight times I think he's one of the top picks which is completely understandable because there's a lot of creative inventors and stuff come on the show and then Richard Feynman is a completely unique pick so it's really great to hear a new one like that and I think yeah you've obviously learned a lot from his way of thinking which is really interesting well Ariel look thank you so much for coming on Riding Unicorns and telling us your story definitely one of the most unique stories about your journey into entrepreneurship achieved a lot at very young ages through through your life but this is a a big opportunity and um you seem to be tackling it with a lot of gusto so we wish you all the best with it and thanks again for coming on thank you so much for having me james and thank you hector as well thank you thanks ariel 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 please tell your friends about it and engage with us on social media and we'll see you on the next episode