Banks Are Getting AI Brains, with Igor Tomych & Dumitru Condrea

IxD 2.09.txt

English (US)

00:00:00.040 — 00:00:10.560 · Speaker 1

We are in the same situation, like boom or dotcom boom. AI agent is knocking on your door saying, hey, I want to modernize your business. Subscription is a pain. We know about this.

00:00:10.600 — 00:00:22.520 · Speaker 2

If we have the proper mechanism, when we can distinguish the cash flow and PNL, it’s like a build in the house on the concrete floor, and then you can build a really tall and pretty cool building.

00:00:34.080 — 00:00:43.520 · Speaker 2

Hello everyone. My name is Igor Tomych and I’m CEO of DashDevs. And today you’re viewing a new episode of the Fintech Garden with my co-host Dumitru Condrea. Hello, Dima. How are you?

00:00:43.560 — 00:01:22.600 · Speaker 1

Hello, Igor. Again. I’m Dumitru Condrea, founder of Novafin, co-host for Fintech Garden Podcast, as usual. And today we will have a very interesting topic. Discuss mostly it’s not about like how to build something. We will not discuss today about product description. We will not discuss about today Compliance will take a totally new topic for us.

Even if me and Igor were working on this and diving in these topics for for quite a big amount of time, but we thought maybe this topic of today’s episode and today’s topic will be an inspiration for product owners, for fintech developers and people who are developing fintech products as business models, as an inspiration.

So Igor, what’s the topic?

00:01:22.640 — 00:02:42.690 · Speaker 2

Today’s topic is controversial, I would say because we discussed several ways to expose it, but the biggest way how it can be titled, I think, is digital banks are getting new AI brain, and it’s connected to the accounting, budgeting and your spending patterns. Because we’re already in this situation, when our banks or financial institutions who are holding the funds, they’re not just showing you whatever has happened, but also trying to help you with understanding what is going to happen with your money.

And it’s around budgeting for personal usage, like monthly budgeting or even sometimes weekly. It’s sometimes it’s corporate accounts where the long term today I think try to stay in the individual segment, but for me it’s really a central leap from showing you the current balance, how much you have on your accounts and the transactions, which happened with the ability to budget, help to understand your spending patterns and then predict whatever is going to happen.

From my experience, some businesses are built around this idea because one of the products that we work a while ago, it’s Chip Financial in United Kingdom, basic idea. And the motto was we’re going to help you to save money and saving money, I think, as essential functionality for digital banks, because if your customer is thriving, then it’s a good point for you because they have money to spend for additional functionalities or even premium tiers with your application.

So what do you think about this?

00:02:42.690 — 00:04:50.900 · Speaker 1

Igor, I agree with you that this topic is very controversial. I’ve never been, at least on my side. We are already facing some interesting requests and we started already to work with so-called AI transformation. Well, everybody are bored with the fact that AI is transforming finance. This is a much of everyone heard about, But things started to be more and more precisely.

Companies are thinking about AI transformation. Not like just a slogan or just like a fancy name. They started to think about the process itself. If two to discuss about this topic, we need also very clearly to separate that fact that exists. AI brain or AI transformation inside a bank, inside a financial institution and existing AI transformation.

In personal finance, AI agent can come as a financial consultant or a financial forecaster and help you to manage this. We will speak today about both things. But first of all, let me come with some numbers. Europe and the United States or North America per general, are just over budgeting. I mean, spends more then than they have a budget.

So it means they have some loans, they have some credits, they use credit cards, whatever. If you are used as a user, you’re used to spend more money from your credit card and then from a debit card. AI comes very much as a helper, and a financial forecaster helps you to, you know, some some patterns or to cut some of the expenses and so on when you are building AI agents and AI modeling transformation for financial institutions.

For a business, there is the same situation all the time exists of a budgeting all the time. The planning of a budget at the beginning of a year, or the beginning of an incremental period, does not fit with what arrived at the end. For different reasons, I don’t know. A customer decided not to pay the invoice or you lost some of the customers or people.

If you are a bank or a financial institution, people started to spend less or they just moved to your competitor or any other reasons. And this is why it’s good to have this kind of AI transformation and to keep everything on track, to keep the monitoring of your money if you’re an individual, or to keep the monitoring of the money and processes.

If you are a business. So basically today we’ll cover both sides from different angles. But in general we speak about the same patterns and the same functionality.

00:04:50.940 — 00:05:24.700 · Speaker 3

I’ve always believed that fintechs should even be limited by the infrastructure. Over the years I’ve seen so many strong teams with ambition. Ideas get slowed down by the same things. Onboarding accounts, ledger payments cards, compliance integrations, and the back office. All the parts nobody sees but every product depends on that.

What led me to launch fintech were attached apps. I wanted to give banks, fintechs and crypto teams a stronger starting point, something faster than building from the scratch, but more flexible than the rigid white label platform. Learn more at ABC.com.

00:05:24.780 — 00:08:44.280 · Speaker 2

I think it’s great to start with the basics in accounting, like a PNL cash flow and balance sheet. So usually whatever you see in your screen in the mobile application as the available funds, it’s a balance sheet because that’s what gives you an understanding how much of the specific currency available for your usage.

But what people sometimes are missing because there are two dimensions. One is the cash flow. So it’s basically that transactions that you see in your application, like today I paid for the coffee and I’ve paid for the subscription for Netflix. But when it becomes more complex, when you are paying for the things per year, because sometimes you might have subscriptions, which actually you’re using for the year.

And it means that even though you spend a specific amount today, it doesn’t mean that it’s distributed or that’s a loss for the specific month because you paid, I don’t know, $120 for some subscription service, and then you should technically account for $10 per month as a loss, not as a cash flow, which happened today for $120.

And I think it’s essential, really essential thing, because if we get the basic math and basic understanding of the things, how we are spending money and how we’re dealing with the real cash flow and our balances, and how actually it’s affecting the PNL, it means a lot because it can be two different processes.

You have some, I don’t know, savings, and you’re buying something that is going to use for specific years, like subscriptions, renting tools for your work, professional work and stuff like that. So two examples that I think are really good, showing the difference between PNL and the cash flow. Because from personal perspective, when you have savings, sometimes you can buy the things that you’re going to use for professional services and stuff like that.

And the current month will show that you are actually in loss. You spend more than you have, and technically it’s correct if you’re using the cash flow. But if you plan the losses across the next year for professional services, once again that you do, you will be in net positive. And that’s essential thing.

And the same sometimes happening where totally opposite situation sometimes is happening with the companies when they technically they profitable. But there is a or cash issues because a specific invoice have not been paid in time from the client. And then you have issues because you are you were expecting a specific payment to pay to your workers and then it’s not happening.

So if we have the proper mechanism, when we can distinguish the cash flow and PNL and our prediction application are properly showing you what is happening and highlighting what kind of the issue is happening. It’s like a building, the house on the concrete floor, and then you can build a really tall and pretty cool building.

But if you build it on the swamp, there is no math and good understanding. What is the cash flow and what is the PNL? That’s the moment when you will be grounded in information. Nothing will be understandable and in the end you will lose your building. So I think when we’re talking about the AI brain for the applications or AI systems, we should always be sure that the basic building structure is there.

So we know what is the cash flow for the specific individual, what is the PNL? And keeping those updated and properly assigned is really complex process because it’s not AI, something that it can calculate. We should see and understand what is happening.

00:08:44.360 — 00:13:56.060 · Speaker 1

Some increments of showing the combination of PNL and cash flow management already as increments, already exists in some of the applications. Online internet banking applications. Some of the banking applications are showing you like, hey, in three days you’ll have a subscription to pay for a yearly subscription.

Yes, you need to have money. Just you have to fulfill it. This is the first cohesion of cash flow and the management and PNL management and already exist. And it’s very important to mention the fact that AI assistant is not a chatbot, because majority of users are thinking that AI is a systems chatbot, which for no reason appears somewhere in your side of your of your desktop screen or mobile screen.

And you need to explain him. You need to give him questions and there’s FAQs. He has to respond. This is not like this. I mean, AI assistant is in the back, AI assistant works. I mean, it’s doing the plumbing for a banking institution, for financial institutions who are providing no card solutions or investment solutions, services for the customers or wherever.

You have a huge amount of data there, and you can come with that data and just compile it in a specific structure with very clean and clear instructions and give advice based, as you said, based on the cash flow and the PNL information. You can give advices to specific user advices that will help you a to keep the engagement, b to even upsell something for example.

I know you’re a provider for IBANs and credit cards or debit cards and but you already created some investment hub inside of your application. And in that I can give an example or an advice as an AI assistant. Give an advice to the users. Look, I see with you for example, each month can spend. If you cut these expenses and this expense, you’ll be able to invest $200 or $150 or $300 to a specific asset.

Again, with all the legal background work, you know, that providing disclaimers, notices and so on and so forth. But I mean, you can you can make upsells. I’m just giving this maybe a little bit absurd example, just to give you an example of how this can be used. What is AI brain? The AI brain is basically all the information structured, polished and under some kind of SOPs or flows.

Even impersonal banking or personally I bring that helps you or helps the user. A better usage of application or a better usage of the time. Of the capacities of individual AI. Brain needs instructions. When you as a banking institution, a financial institution will build some kind of a brain for your company as optimization of your finance, or you’ll build a feature of AI brain for your users.

In both situations, you already have this data, if exists, a big possibility. You have an army of people who are working with this data and can triage this data and can paraphrase this data where I now arrange it in a logical and logical information. Logical instructions. So before jumping to any kind of AI transformation companies, individuals have to understand that this is very important to have clean the mess with the policies, procedures and things you are doing.

If your individual you still have a policies even if you are very not documented. Maybe for the majority of people they are not documented. You have some patterns you spend every month for the rent or for bank lease, or you spend every month to pay the bills. And there still exist procedures and policies.

So it doesn’t matter individual or company. First you need to polish that processes. And then when you have the data arranged, when you have everything put in place, then you can build the AI brain. So AI transformation starts from polishing your internal mess and procedures and policies. And when you have a PNL process, for example, some people already are making management of a personal finance using PNL approach, some are majority are using cash flow approach.

And if, for example, in both cases if you can combine them, perfect. But in both cases you need to have strict rules and strict policies. What you do in case if appears. Unexpected expense. Even if you’re in a budget for a company, then you don’t spend on another things. And if you train and if you explain to the AI to the AI brain how these situations have to be managed, then you can build such kind of a brain.

Now, if you are banking institution and you’re building a brain for that will be embedded in the companies in your application for personal usage of your users. When you build such a brain with all the amount of data you’re having from this brain, and you can gather this amount data very legally from other sides of of our businesses, other products via open banking APIs and so on.

You can understand the patterns and you can give this information to the users. And you can give to say, not like, hey, in three days you have to pay $600 subscription for, I don’t know for a product, but hey, you have each year $600. Let’s start. And today, like, I don’t know, 180 days before the subscription will bill you put aside some money and those money we in parallel we can invest so you can create different wallets or accounts for deposits where money will at least not be eaten by, by inflation.

So this is what it means. AI brain and embedded AI brain into natural products.

00:13:56.060 — 00:16:38.030 · Speaker 2

I’m really loving using analogies and in this case, the closest one that I can think of is a type in prediction. Because take a look on our mobile devices, we have keyboards which sometimes are limited with ability to type. But what mobile companies who are producing those devices are good with is to do the prediction.

And the prediction for typing is usually based on two level of inputs. When you type in on the your mobile device, it collects, analyze and create personalized suggestions and predictions of your type and patterns. And then the second level, usually those companies are using really massive, really huge amounts of data to build common prediction tables.

And I think if we come back to the financial institutions or our own processes, like we really need to have tools which are helping to provide us common patterns, like how actually are you doing in this specific country? Because certain events, financial events They might be connected to the calendar, the payment patterns, because.

Payments in the United States are biweekly compared to monthly payments in Europe and some other countries. So we should really use the data. Group that you belong to. But the second one is to create the process, which is personalized and predicting the patterns for the specific person, because we might find a lot of things like subscriptions.

And that’s really good. Step up. Right now, no one expects or no one is treating this subscription as additional functionality. Everyone would like to see how much you are going to spend in patterns, and then understand when you need to switch it off, but still really fragmented. Like what you see in your mobile infrastructure.

Like is there is a subscription page and then your banking show and other subscriptions which are going through the app. So there is no aggregated data. And another level or another part that I would like to discuss is open banking, because by the design it was created to allow you to do 360 aggregation, what you have with the financial cash flow.

And I think those tools right now are still available and the foundation is there, but we still not the same level as with the type in prediction, because banks might be a little bit slow with the understanding how it’s important. And I assume the AI trends and AI race actually will will bump a lot, because for building really high level quality assistant, you need to have really good data.

And for having good data, you need to be smart about 360 visibility with the high level patterns with the personalized patterns. So I think we just in the beginning where the speed with AI is changing everything. What we dreamed for last decade with the available tools might come into one place.

00:16:40.990 — 00:20:09.050 · Speaker 1

Thinking about the availability of the data. Some current researches showed that already 57% of Americans are using AI, mostly in a primitive way were grocer mode as a chatbot. But they already use AI for managing their personal finances 16 something percent from all present in North America. Financial applications present in North America already use.

Some embedded AI functions were fully automated, where they have strong automated AI base. And all these applications that you have in your phone, financial applications in your phone, most of them, they are not speaking to your bank directly. And this is by very, very interesting problem. They are not speaking with a bank, but they know a lot about you.

They know about your finances, they know about your spending, but at least about your your patterns, they know about some. Of course a lot of information is not available. A lot of information is tokenized. A lot of information is cryptic. You cannot just access and read everything. So just don’t be scared.

There is no way currently, at least for the majority of people, to know everything about you except the special services. But those applications, financial applications that you have in your phone, they are connected to some other aggregators. Data aggregators, I don’t know there like Plaid, MX, Yodlee.

Just imagine Plaid is connected to more than one. 1100 plus institutions around the world and 200 million users. So they know this data. So they have this data. So in a very short way explanation. You just can connect to Plaid of course if you have specific license and motivation blah blah blah. And you can take this information and imagine that you have this I mean imagine it’s real.

You have access for this data and you as a product manager or a product owner or a company runner, you have access to this data. So the most important thing is, as already mentioned, we have technologies, things that already work for, for, for years, like open banking data. There are aggregators you can gather.

And having this foundation you can build a lot of interesting things. Possibly the robo advisors is a niche that will grow very, very much. By the way, already exist some interesting startups and they already exist some interesting applications on the market who are already provided not only in MVP’s, but provided quite solid products on robo advising finance world that can be embedded to banking applications and can give an upscale and can promote, help your product or help you as individual to plan a lot of things.

Subscription is a pain. We know about this and even sometimes for me it’s a pain. I think I’m managing quite well my my budget and my PNL and my cash flows. It’s sometimes I can actually forgot about this even if I have some calendars, memo and so on. So AI will help me to manage this. And by the way, this is one of the iteration my personal AI brain building.

But I’m doing currently because again, coming back to the beginning at Novafin, we start to have some, some kind of such kind of a request for, for AI, digitalization optimization. And I thought, like if I’m doing for other companies the same thing, maybe it’s good, it’s time to do it for myself. But Igor, let’s move a little bit and discuss a little about the AR and the fraud war or AI.

Fraud war I, for example. Now generative AI already created possibilities to build fake individuals. Were criminals are using generative AI to invent people and start to loan money. How can AI brain build it in a financial application for personal finance work embedded in an internet banking application, for example, can help companies to build to to have this war against the AI fraud.

00:20:09.210 — 00:21:12.930 · Speaker 2

Maybe before jumping into the fraud, I would like to ask you a question because while going through the episode, we discussed like a there is way and there are the tools which will allow us to understand where our money is and what is the happening or going to happen with the funds. And the thing is going to be the separate applications, new type applications, because ten years ago budgeting on the mobile applications was the thing like Mint started and then Mint has been acquired by Intuit.

Right now it feels like banks or digital services or digital financial services. They sit in on top of a lot of data, but it still lacks because it’s like a 360 overview. You need to aggregate all accounts, and instead, is it going to be the super apps that we see on the market, Revolut or others, or it’s a new type of AI, brain applications or even ChatGPT will.

Will be the aggregator for your financial activities and then suggest, hey, you actually can do this and that, and maybe then the revenue for GPT will not be in ads, but into the financial services, because it’s kind of advisory. So what do you think about that?

00:21:12.970 — 00:25:15.150 · Speaker 1

We are in the same situation like dot coms boom or pre dot coms boom. So the beginning of the 90s, we absolutely are in the same situation. At the beginning of the 90s, you know, sales agents were knocking door to door selling computers. We know some good movies about this by the way. So just take a look. It’s identical.

Now sales agents are knocking or AI agent is knocking on your door saying, hey, I want to automate your business. I want to automate your personal finance or personal things you have to do every day, and so on. Your routine. And most of the responses are like, yeah, we already have it. We have Excel, we have databases, we have Power Query.

We have I know AWS, we have a lot of things instruments. Why I need this? I mean I can develop I have Revolut looked like super nice super app with marketplace refunds. We have possibility even to apply for a chargeback. We have a lot of things. Why I need this to make continue this analogy of an. It is beginning of the 90s.

I think people will try different kind of applications. We will have a boom of applications a row by advisors like a separate application. We’ll have a generic payment application as a separate or embedded into existing application, like a technology, like a technical vendor. We will have applications who will develop brain builder onboarding for for businesses to help you to build the brain in seconds, or at least in two days.

Not maybe not in seconds. We’ll have a lot of plenty of applications for me currently, it’s quite hard to say what will be the next boom or where would be the trend, because this will happen in 4 or 5 years. So we’ll have a bloom of different kind of applications. Businesses will try to find the correct niche to find the specific niche where it goes.

So it will be quite a Wild West. In a moment, we’ll appear. Crisis at crisis can be generated by like the crash of dot coms example. Or maybe stronger modeling will appear. Maybe some advancement in quantum computer will happen. Who knows. So there will be a crash where the market will reshape and. And only after that crash will be able to make a strong forecast.

What is the future of a so-called super apps where those super absolute will exist will be very sharp? My humble opinion is that companies who already have a super apps like Revolut too, perhaps where there are some very interesting strong internet banking applications in Ukraine, for example like monobank or PrivatBank.

They will have to adapt to this and to embed AI and AI will not be a feature. Like I already mentioned, like I don’t know, a chatbot pop up. Hey, we have AI transformation. Look at our chatbot. You can speak with it. now that AI will be behind. By the way, yesterday I dived very much into for those who are working with Atlassian products at less than hello and I tried their Rovo.

The AI things works perfectly, perfectly well. So for those who are working with AI in JIRA, I think something like this would be awesome. Bringing applications but more evolved of course, and more embedded. Be like a mix of I know a genetic agents like Hermes mixed with Claude Code or Claude will have behind all the data from the companies and will make the plan boring.

For my financials, I honestly would like to have an albeit as a user. Very happy to see in my one of my internet banking applications to see hey dude, you have subscriptions or you have I see that you travel a lot. I see that you don’t know spends money on inventory for making a shooting movies or podcast. Maybe you need into a year to change your camera or microphone or whatever I propose you to take as a goal to purchase or this asset, this kind of microphone or this kind of microphone.

And for this you need, I know, $1,000. And for this, let’s start to gather some money starting from today. And I will take every day $6 or $7 from your bank account, and I will invest it somewhere. I will invest it and we’ll do everything well. So in the moment, in a year, you’ll have sufficient money to purchase this.

For me, it will be absolutely the best best feature in one internet banking application.

00:25:15.190 — 00:27:15.040 · Speaker 2

As a closer, maybe a part or small questions around the fraud detection and AI or generated identities. I think it might be the separate episode for that, but in a nutshell, I’m a little bit skeptical because if we talk about the identity theft or fraud, it happening for more than 100 years or even thousands, because the ability to identify and authenticate the person always hard like think about 100 years ago, the proof of your identity were papers and we all known spy stories when the spies were supplied with really good quality fake documents to get behind the lines, and really popular in East Europe or even post-Soviet countries, that spies were identified because they’re staples on the documents.

They were not going with the iron like crust. And I think it’s a spear and the shield problem, because if the technology for fake identity is evolving the same way, the technology for protecting the customer or protecting or finding those cases is also evolving. And in this case, I really like the old, I think Rathergate case, when there was a scandal in 2004 that a specific I will not go quite deep, but a specific document.

It was released as authentic document from 1973, and then the community found out that it was fake because it was typed with the Times New Roman font, which didn’t exist at that time. It shows that when we think that a specific technology is helping and tracing the editing. Tracing the mask. Tracing the AI generation for pixels.

It still works, I think an ability to find out. It costs money, it costs investments in R&D. But it’s not just one way robot. So let’s maybe dedicate a separate episode to talk what is happening right now on the market. What are the even tools to be protected, or what are the tools that that are trying to use for stealing and faking people?

00:27:15.080 — 00:28:36.450 · Speaker 1

I thought we’ll have sufficient time to dive in also on this topic, because building an AI brain for financial company will give a huge boost for the dating fraud and AI fraud, even like we have current examples. This year, Mastercard and Visa started to work with the AI internal tools for the authentic fraud, and they are doing absolutely great.

They have a lot of success in this markets. Ah, I mean a lot of fraudsters start to, to, to run and to find solutions and to find how to avoid any kind of problems and in fines and that appears and it works absolutely great. So this is, by the way, one of the examples of how AI brain can help you to avoid fraud. But yeah, we we don’t have sufficient time.

So be sure we’ll dive into this topic in one of our next episodes. It was a very interesting topic. So me with Igor where all the time happy to consult you and to help you in building AI automation, an AI transformation to your company. Not like just inserting the chatbot application, but to come and to build it properly with understanding how it has to be done during this time.

Polishing all your policies and procedures, understanding what is your workflow, finding the leaks that you’re having in your budget or in your processes, and then just on that foundation to build a very well established brain that will help you to build AI transformation for you and for your customers.

00:28:36.450 — 00:28:43.730 · Speaker 2

Thank you for joining us. And let’s catch up for the next episode in one week. See you. And please subscribe and follow our social networks.