After exiting two startups, finance professional Daulet Aitmakhanov decided to build something closer to home, drawing on his own background in finance. His new platform, FinFlow, connects small and medium-sized businesses with finance professionals and gives those professionals an AI assistant to help them work faster and more efficiently. That means they can take on more clients, while business owners get a much clearer picture of where their money is going.
As part of Digital Business’s joint project with Astana Hub, “100 Startup Stories from Central Eurasia,” we spoke with Daulet about how well AI can support businesses and finance professionals today, why FinFlow wants to build its own small language model, and which markets the startup plans to expand into next.
“Around 80% of SME owners don’t have solid financial practices in place”
— How did you get into the tech business?
— I’m a finance professional by training. I graduated from Nazarbayev Intellectual School in Astana and then went to South Korea for my bachelor’s degree in International Finance. I chose Korea because I was able to get a full scholarship there. I applied to universities in the US and Europe as well, but those options would have been more expensive. So I went with Seoul, and it turned out to be a great choice, not just for the quality of education, but also in terms of safety, healthcare, and everyday life.
I came back to Kazakhstan in 2016 and eventually joined Astana Hub, where I was one of the first people working on the innovation and venture ecosystem alongside Magzhan Madiyev and his team. At the same time, I was consulting for SMEs and startups, helping them with financial management and modeling, management accounting, and company valuations. I also worked on several M&A deals.
I launched my first startup, Issu.kz, in 2018. It was a pure e-commerce business that offered subscriptions for beauty products like cosmetics and fragrances. To be completely open, we modeled it on the US startup Scentbird. Their service was doing really well, and we managed to replicate that success in our market. Three years later, the startup was acquired by a private beauty products distribution company.
After that, I became a partner and CFO at CTOgram, a platform for finding and choosing repair shops, service centers, and auto parts stores. It brings three sides together in one place. If something goes wrong with your vehicle, you simply submit a request and start receiving offers from repair shops. It works a bit like inDrive, where drivers send you their own terms for a ride. Then, once the vehicle has been diagnosed, you can order the part you need and, if necessary, pay for it in installments. We raised more than $1.5 million in investment, and the company was later acquired by Freedom Holding Corp.
— After two startups, you moved into a field much closer to your own expertise: finance. How did the idea for the project come about?
— While working on Issu and CTOgram, I continued consulting and working as a finance manager. I’ve always enjoyed digging into the numbers behind a business, understanding how it works, and seeing how things change over time. What I noticed was that around 80% of SME owners don’t have solid financial management practices or management accounting in place. And that creates very real problems, from cash flow gaps and mounting debt to rejected loan applications because of a poor credit history or weak financial health.
Another important point is that good finance professionals are expensive. Small businesses can’t always afford to hire one. On top of that, there’s a shortage of specialists. Kazakhstan has more than 2 million active SMEs, while the number of certified finance professionals is roughly 20 times lower.
By 2022, AI was really starting to take off, and combining a finance consultant with AI felt like an obvious next step. It was clear that the technology would make its way into accounting sooner or later, simply because there are so many routine tasks that can be automated quite easily.
That’s how, toward the end of 2023, my co-founders and I decided to start working on FinFlow. By early 2024, we were working on it full-time, building prototypes and finding our first customers.
“Business owners aren’t ready to trust AI alone with their finances”
— So, what exactly is FinFlow?
— Clients get both a real finance professional and an AI-powered platform that integrates with 1C. For now, this is the only system we support.
FinFlow looked quite different at first. We were focused solely on startups and planned to enter the US market first. But we quickly realized that winning customers there was a slow and difficult process, and, to put it mildly, startup founders were not always in a position to pay. So we shifted our focus to more traditional SMEs and started building a full-fledged AI CFO (chief financial officer). For a small business bringing in around 10 million tenge a month, hiring a finance professional on an 800,000-tenge monthly salary is a big expense. We started with an AI finance specialist on a subscription model at $100 a month. It worked as a chat interface where the business owner could connect their accounting systems and upload bank statements. As they interacted with it, the system would build out the company’s management accounting model.
We had around 150 clients and an MRR of about $16,000. But in March 2026, we decided to change the business model and shift toward building a tool specifically for finance professionals.
– What made you decide to take that route?
– We realized that AI isn’t going to replace finance professionals, at least not anytime soon. Business owners simply aren’t ready to hand over control of their money to AI alone, and AI hallucinations are still very much an issue. The cost of getting something wrong is just too high, both financially and in terms of the product’s reputation.
Instead of trying to replace finance professionals, we decided to make their jobs easier and faster so they could work with more clients. From what we’ve seen in practice, without FinFlow, one specialist can comfortably manage around three SMEs. With our product and the automation it provides, that number goes up to about ten.
AI’s role is to take care of the routine work. As soon as a contract is signed with an SME, the finance professional is faced with a huge amount of data: bank statements, Excel spreadsheets, financial documents, information from 1C, and so on. All of that used to have to be sorted manually by category, amount, date, and other parameters.
Say your client sells phone accessories. They might have thousands of transactions a month, each worth 2,000 to 4,000 tenge. Traditionally, every single one would have to be categorized manually in Excel, which is incredibly time-consuming and tedious. AI can do the same job in about 10 minutes. The finance professional then just needs to double-check the results, review the categories, and make a few minor adjustments where needed.
“Our clients include medical centers, pipe manufacturers, dairy producers, and many other businesses”
— Are the finance professionals full-time FinFlow employees? How do you work with them?
— They’re our partners. Under our business model, both the business owner and the finance professional, or a consulting firm such as FinPark, pay for a subscription. We also work with organizations that train finance professionals, and we help them find clients.
— Does FinFlow work for any business, or does it need to be customized for each client? Have you ever had cases where it simply wasn’t the right fit?
— A year ago, when we were still building FinFlow as an AI CFO, that was definitely an issue. The system might work really well for a language school, for example, but not nearly as well for a window installation company. Businesses in different industries handle finances, transactions, and accounting in very different ways. We solved that by changing our positioning. FinFlow isn’t a product for businesses. It’s a product for finance professionals.
Today, our clients come from all sorts of industries, including medical centers, educational institutions, oil pipe manufacturers, dairy producers, and retailers.
— What makes FinFlow different from other AI assistants? Why not just build an agent for ChatGPT or Claude to handle a finance professional’s routine tasks?
— Ours is a highly specialized product built for a very specific audience. Over the past year, we’ve worked with more than 150 clients and learned how to label local data properly. That means our model, which uses Sonnet 4.6, Sonnet 5, and Haiku, can understand the context behind a transaction and why it happened.
We’re also building our own small language model specifically for working with the finances of Kazakh businesses. That will give FinFlow the context from our labeled data and allow it to integrate fully with banks. We plan to train it on our own data, supplemented with synthetic data as well. At the same time, token costs should be around 10 times lower than with ChatGPT or Claude.
And there’s another advantage coming soon. We’re close to integrating with a bank that serves around 300,000 SMEs. Becoming part of its ecosystem will give us a major distribution channel. Claude and ChatGPT are cloud-based solutions, so they simply won’t be able to fit into every niche, not least because of security concerns.
— Building your own language model is expensive. Do you have the budget for it?
— Building a model is getting easier and cheaper as the technology develops. Chinese tech giants, for example, are releasing some of their work as open source, so companies can use it in their own products. People also often confuse small language models with general-purpose LLMs like ChatGPT, Gemini, or Claude that are designed to handle almost anything. That’s not what we need. We need a small model built to handle a narrow range of tasks, specifically financial queries.
Yes, it’s still not cheap to build. By our estimates, we’ll need around $500,000. We’re factoring that amount into our next funding round specifically to finance the development of the model.
“Delayed reporting is a major pain point for business owners”
— How did you find your first clients?
— We used performance marketing, organic channels, and referrals. Partnerships with financial companies have helped as well. We experimented with different ad campaigns and landing pages, tracking the funnel at every stage to see what worked.
It was important for us to understand what business owners were actually struggling with, and delayed reporting came up again and again. You close the books for July, but the report may not arrive until the second half of August. For several weeks, the owner doesn’t have an accurate picture of expenses, balances, profitability, or cash flow.
That’s when we started offering a product built specifically around that problem. The low price also helped win clients over. With the previous version of FinFlow, the subscription was $100. Today, the average subscription is around $600, but that now includes access to a finance professional as well. Even with the higher price, sales have improved. Business owners value being able to speak to an actual specialist, and the subscription includes weekly calls.
— What determines the price?
— We have two packages, priced at $600 and $1,100. We segment clients quite heavily by monthly turnover: up to 15 million tenge, up to 50 million, 50 to 100 million, and over 100 million. Each group needs a different approach because the challenges are different. For example, a business turning over less than 50 million tenge a month may be perfectly well served by a finance professional with less experience. Once you get above 100 million, the business model is more complex and the owner expects someone with much deeper expertise. So the difference in price really comes down to the size of the business and the experience of the finance professional.
— How many clients do you have?
— We only switched to the current model in March this year. We now work with five finance professionals, who manage just over 20 clients. We’re growing around 40% month over month. That sounds like a lot, but early-stage growth always looks like this. You might start with five users, get to ten soon after, and suddenly you’ve grown 100%. What we’re especially proud of is our 100% retention rate. Retention was much weaker with the previous version of FinFlow.
We’ve already brought MRR back to where it was before, at just under $20,000. The company is still operating at a loss for now, but according to our plan, we expect to break even in October or November.
“We expect to raise $1.5 million at a valuation of around $10 million”
— A year ago, you raised $500,000 from Orbit Startups. Have you raised any additional funding since then?
— No. We’re preparing for a seed round this autumn. Over the past year, we’ve focused on developing the product, strengthening the team, which now has 12 full-time employees, and repositioning FinFlow.
Now we feel we’re on much firmer ground. We understand what makes our product valuable and, just as importantly, how to sell it. In the next round, we expect to raise $1.5 million at a valuation of around $10 million. We’ll be focusing mainly on international investors. We recently took a closer look at the investment landscape in Kazakhstan and realized that the market is going through something of a “venture winter” right now, with far fewer deals than last year.
— Are you looking to expand into other markets?
– We’re interested in the MENA region and plan to launch there in 2027. I tested the UAE market last year and found that the customers are quite similar and face many of the same problems. The main differences are in regulation and tax laws. We’ll need to adapt the AI model, of course, but it won’t require a complete overhaul. When I presented FinFlow there, the feedback was very positive, and local businesses showed a real willingness to work with us.
Another area we’re developing is partnerships with banks. We’re already working on our first project of this kind with a Kazakh bank that serves 300,000 SMEs. FinFlow will be integrated into the bank’s ecosystem under a revenue-sharing model, giving its clients access to our solution at a lower price than our standard subscription.
— Which accelerator programs have you taken part in?
— Two. The first was a MENA-focused program supported by Astana Hub and White Hill Capital. That was actually when we decided we wanted to expand into the region. The second was Orbit Startups. Before investing, they require startups to go through their accelerator, and it was a really strong program. It’s somewhat similar to Y Combinator: they don’t just put money into the company, they also help you develop the business and grow as a founder. At this stage, it could be useful for us to join a more specialized program focused on entering a particular market or, for example, working with banks.
— If you were launching FinFlow today, knowing everything you now know about the industry, what would you do differently?
— We would have gone straight to the product model we’ve been using since this spring. If Claude Code is for developers, FinFlow is for finance professionals.