In this Connector Podcast episode, host Koen van der Huydonk interviews Tom, founder of 7Yoken, about the company’s autonomous accounting intelligence platform. Tom explains how 7Yoken’s API can replace days of manual accounting work with minutes by ingesting balance sheets, invoices, and bank statements in various formats (Excel, PDF) and producing full financial reports, KPI dashboards, and actionable remarks for banks, fintechs, and SMEs. The journey began in 2022 with a "certified autonomous bookkeeper"—an AI model that performs junior‑accountant bookkeeping tasks such as invoice processing, bank‑statement reconciliation, and accruals, enriched with Belgian tax and VAT knowledge. Building on that, 7Yoken is developing a full‑stack autonomous accounting intelligence that can generate profit‑and‑loss statements and balance sheets on the fly, and eventually handle personal‑tax filing. To make the AI trustworthy, they introduce a Digital Financial Officer (DFO) persona that interacts with users via chat, email, or report delivery, keeping a human‑in‑the‑loop. Tom highlights challenges of cross‑border accounting, noting that the system must understand local accounting rules (e.g., Portuguese, Argentine) while presenting unified reports to business owners. Technically, the solution runs on open‑source large language models hosted in Azure, allowing rapid adoption of new LLM releases (e.g., DeepSeek). The conversation also covers a new joint venture with Data.be—e‑invoice.be—a platform that captures invoice data at source, converts legacy formats, and feeds the AI engine, aiming to meet upcoming Belgian e‑invoicing mandates. Finally, Tom describes the ideal customers: small companies that still use Excel/Word for invoicing, and larger firms with legacy ERP systems that need a compliant API without overhauling their core software.
Chapters
Welcome to the Connector Podcast, an ongoing conversation connecting fintechs, banks, and regulators worldwide. Join CEO and founder Koen van der Huydonk as you learn more about the latest available trends and solutions in the markets.
Welcome to another podcast here live from Fintech Belgium in Brussels, and in collaboration with ourselves, the Connector, I've got Tom with me with a very special name of his company, 7Yoken. Tom, what do you do? What is your business?
Good afternoon. Well, our business is that we are creating an autonomous accounting intelligence to provide actually an accountant's days of work in just minutes. So basically we say 7Yoken, accounting done in minutes.
Interesting. Can you enlighten us a bit more about what that exactly means?
Well, for example, to have the example of fintechs and banks. Banks today, when they want to allow a company to have a credit, And they need to do a few financial health checks. And so basically you need to have an accountant's view to calculate if this loan is proper to give. And our API can just understand the data you give him, for example, a data balance sheet, or it can be in Excel or in PDF. And we are going to give you a report like an accountant would give you. So it would be the standard report with your basic KPIs, but also our opinions, our remarks, and And for example, some additional questions that we would ask the company about their balance sheet. So that's one of the tools or the results that our API can deliver to a person who is in need for an accountant but cannot have an accountant.
Can I say that this is a bunch of accountants that's behind this API, or is it like fully autonomous? I'm putting it black and white by purpose.
Well, The results. And so we are clearly, okay, what is the purpose of the API? But in the background, of course, there is an AI accounting intelligence model. And basically we are going for the autonomous AI accounting because he needs to know what to do based on the data it gets. So if we put a bank transaction to the AI agent, it's going to see, okay, I need to Check what's the amount, can I find a supplier, is already paid, where's the invoice, what is it about? And it's going to then check, okay, which APIs can I talk to, an ERP system, an accounting software, and it's going to execute his task.
So the application itself, Seven Joken, goes as far as actually executing. So you build a whole accountancy firm out of the box.
Well, out-of-the-box things. We started with this idea in 2022, and then back then we just have to accept the limits of the models. So we started first to create what we call a certified autonomous bookkeeper. You can compare it with a junior accountant. So basically it's all the bookkeeping tasks. So the transactional data, so the processing of the invoices, the bank statements, accrual, basically doing all the manual repetitive work. Yeah, that was the first level. And then of course we needed to add some more knowledge about corporate tax law, VAT law, and specifically for Belgium because I also own an accounting office. So we are using it in my accounting office.
That was my next question. You, you steal my questions, Tom.
Sorry.
No, that's fine.
No, that's the reason I was fortunate enough to, to, to have a good friend, Willem, who, who is actually an IT and cybersecurity guy. And so basically my ideas, he executes them in, in AI and development.
Mm-hmm.
And of course, if you need to, if you want to make something like an autonomous accounting intelligence, you need to have the data. And yeah, fortunately, since 2012, I was trying to automate my dad's accounting office. And so we have a lot of data from invoices, from, from emails, We have seen almost everything. I'm not going to say that we don't have new situations, but we have seen a lot in those past 15 years that I'm working there. So yeah, the data is important and knowing what is happening in an accounting office, and then you can reach new levels of an accounting intelligence. So we first have the certified autonomous bookkeeper, but then now we are growing to having this autonomous accounting intelligence. So basically You can just give invoices, the bank statements, and we will create a P&L, a balance sheet on the fly.
So you literally go from what was maybe one directional, now is output. So do you go as output as far as filing those things to the authorities as well? What's the next step?
Well, the next step is the step we want to take because an accountant has different tasks. So now we just focused on the tasks that, that for me needs an employee, an employee. So I was looking for what can give me the most ROI. But of course, if we would focus ourselves on having a fully automated personal tax model, we would, we would be able to do this.
And you said a few times it's something you do in Belgium. Is it something that you can lift and shift into other countries, regions, whatever, continents?
That's basically the difficult part because, for example, if you want to apply Portuguese accounting, first of all, you need to have the knowledge. So I rather see it that we are going to provide more the situation where a company has a location in Belgium or in Europe, but also has an acquisition in, for example, Argentina, to give a really real-life example. And the problem is there that Yeah, you don't want to change the whole ERP system in Argentina, or also not possible, but you do want to be able to communicate like you would with a Belgian person. And so basically the first added value is not going to be for the accountant, but for the business owner who just wants to query the accounting data in Argentina without having to access the ERP system, knowing the language. And he basically is going to get the same reports like he would have in Belgium, but then from the accounting ERP system in Argentina. And that's really a powerful tool.
So you actually go beyond just having invoice information, but you literally go like financial information that comes into the system.
Well, let's just say the information that is needed to take the right decisions. And this doesn't always have to be accounting data, but most of the times it's somewhere in the accounting data. So yeah, you have different— I mean, if you want to, for example, if your core business is about supply, then you really want to focus on these KPIs of the supply. But as an accountant, you most of the time are not really always capable of knowing the details of this kind of industry. But the information lies in accounting data, in the invoices, in the in the bank transactions.
No, absolutely. In our preparations, you also mentioned the notion of a digital financial officer. Can you explain a little bit more about that?
Well, so instead of a CFO, we're going to have a DFO, or we have a DFO. So basically, this autonomous accounting agent needs to have a persona because it still needs to, or is still required to talk to a human for now. And so, yeah, that's true. So that's why we create the persona of a DFO. And so basically, the DFO can— you can just chat with it like we know we can email with it, but you can also just deliver reports. So basically you can say, okay, I want the report in Excel, PDF. And yeah, we don't need any UI for this. You just want the result. But we needed to create a persona to be able to, okay, how can we talk or how can we put some human in the loop? You don't want— people don't trust, of course, and that's true, the fully autonomous agent.
Well, I guess the reason why you don't want to have a full autonomous agent is because the notion of hallucination. And specifically with financial data. So how do you marry both elements?
Well, things in accounting, we are very fortunate that accounting always actually means a framework. And I don't want to talk too much about the accounting details, but you always have a debit and a credit. So it means that there's always some kind of checkpoint that you have to see if the LLM is right or wrong. But indeed, the Biggest problem is if he would hallucinate or if he would not understand or grasp the the idea of accounting. You know, he needs to actually have he needs to know has the notion of the parts of the accountant of the accountant. For example, he has to know the notion of a chart of accounts. He needs to know okay what's the difference between cost of goods sold or or depreciation. He needs to know this kind of concept, and if he doesn't know this. then that can really be a problem because then he starts to hallucinate. But of course, this you do by benchmarking and using test data. But indeed, let's just say that we are lucky that we can have checkpoints and of course a lot of data. So we see that today the models that we use are not going to hallucinate about their notion that they have of chart of accounts, for example.
I think also the notion that the models are changing rapidly. So how would you actually ensure that you as a company sort of stay ahead of the new innovation that happens within the LLM space?
Yeah, well, the part that we already like finished, like the Certified Autonomous Bookkeeper, for this we use only open source models and it's also only in our Azure AI cloud. So yes, we, we currently are able to run completely actually on open source models.
Yeah.
And it's— and I'm happy for this. It's only getting better. As you know, a few weeks ago DeepSeek released. So we're getting like a new wave of this thinking LLM. But for now, the tasks that we need to do or that we want to do, we can use completely open source models.
Yeah. Is that then the advantage of being in bookkeeping, which is a very old and traditional profession that sort of processes are clear? Flashed out, if you want.
Very true. But also it's worldwide the same concept. If you would call to India and ask for the business expense, he's going to give you a business expense. And apart from differences in accounting, for example, you have Belgium GAAP, or you have EIS or IFRS, they're all following the same balance sheets or annual accounts. It's across the world, it's the most the same thing. So yeah, that's the lucky part.
And while looking at innovation in our space, I mean, a lot of innovation or has been claimed as innovation is, for example, Peppol. Is that— do you see that as innovative thinking or the next steps? And how do you deal with it in your company?
Now you're asking a funny question to me because actually, you know, Toon van Acht, Very well.
Everyone does.
Well, we start talking with each other in the Digital Finance Summit, and basically we have started together a new company. So Seven Joken and Data.be, a new company, e-invoice.be. And we're actually launching next week. And the reason why we—
Exciting.
Yep, indeed. Exciting. It's true. And the reason why is because indeed our problem today is actually that We can only have a capable accounting intelligence for your company if we are able to have the data.
Data in, rubbish in, rubbish out.
Well, let's just say that an accounting software today, it actually, it's a limited version of all the data. For example, you just need a few data points from an invoice, but you don't need to know all the details and you don't need to know exactly Let's just say that if your car goes to maintenance, it's just going to be one chart of accounts saying maintenance of a car. Not going to say, okay, we changed the wipers, we changed the tires.
No, exactly.
But this is data that we need to create the context because otherwise how would I know that it's going to be a truck? So different business expense or a Porsche. So this is the stuff that we need. And yeah, when we were talking about this invoicing people, well then, then we are at the source. the incoming invoices, outgoing. We're also going to connect with the bank account. And yeah, these 3 parts together with our AI accounting intelligence is going to give also a lot of, to the business owner, like pre-accounting information without the need of going to have to contact his accountant or having access to his accounting software. But basically the first step is to make sure that Our experience in normal invoicing is going to reflect on our e-invoicing platform, which means that we're going to make sure that our goal of e-invoice is that you can just do the same as you are doing today. So for example, we have made an— yeah, I know that a lot of people who have a small company still work in Excel or in Word PDF, which doesn't matter. So we have made a converter, for example, so you can just stay working like you are working. And for the incoming invoices, we're going to make sure that you're going to receive the invoices like it used to be, like for example through mail. But also we're going to make sure that it's going to be human readable because an invoice, the whole point is it's going to be machine readable. So an invoice is just a text file, plain text file. So that means that there's not going to be a logo anymore. I mean, most of the people have the habit of putting trust in things that they already saw. But if you're going to, if you're going to open an invoice from, for example, insurance, nothing's going to say that it's going to be a game insurance. Just no logo, not the usual feeling. So we are going to make sure that because it's not obligated to add actually the PDF as a binary, so you don't need to provide the human-readable. So we're going to make sure that we're going to let the people have always the same feeling as they had before by giving it the same feeling. Yeah. And Yeah, let's just say we are really eager to see how it's going to evolve because for the first time, I think now Belgium is going to be the first one who's really going to obligate. But today we still need to onboard more than a million companies.
Big work ahead of us.
Yeah.
And also we still need to find out what the pain points are going to be. And for example, if you receive an invoice and you want to inject it immediately into your accounting software. Yeah. What if the person sends 3 times the same or also sends it by email? So this is all going to be practical issues. And that's why also we want to create this platform to make sure that we're going to capture these problems and solve them.
Wow, super cool. Tom, thank you very much for sharing all these insights. And it's nice to hear that there's Belgian companies that are so far ahead in sort of using AI and thinking about how to combine all different data sources. So I was very pleased to hear that. Maybe one last question before we close this podcast off today. Who should contact you? Who is your ideal future customer?
For Invoice.be, let's just say that our first profile is a person who has a small company and like, for example, a management company and just want to just work like he used to be. We also provide an API for the people who need, for example, you have, you have people, companies that have one of their departments in Belgium, but just don't want to adjust their whole SAP system just to assure the compliance for Belgium. So we offer for ERP, we have an API, and basically we offer for all kinds of companies a solution for the current problem. And most of the times we see that their problem is that they have kind of a legacy system because they are having a production company. They don't want to to change this and we can provide for them the solution with an API or customized solution.
Well, great. I wish you all the best of luck.
Thank you.
I'm sure we're going to see much more about this in the press anywhere.
Yes.
But for now, thank you very much for sharing your information with us. Thank you also to the audience and stay tuned for more news from this podcast.
Thanks for listening to another episode of the Connector Podcast. To connect and keep up to date with all the latest, head over to www.jointheconnector.com. Or hit subscribe via your podcast streaming platform.
