Generative AI: Growth Strategies for Fintech Companies

Background: In this conversation, Alicia Roisman Ismach interviews Stuart Levine, a fractional FinTech consultant, about the impact of generative AI on the financial services industry. They discuss how financial institutions are adopting generative AI capabilities in areas such as lending, credit scoring, document management, customer experience, and fraud detection. They also explore the challenges of trust and accuracy in using generative AI and the potential for AI to replace core banking systems. The conversation highlights the importance of data privacy and the need for clarity in distinguishing between different AI technologies.

Takeaways:

• Generative AI is being adopted by financial institutions in various areas, including lending, credit scoring, document management, customer experience, and fraud detection.

• Trust and accuracy are key challenges in using generative AI, and it is important to have human oversight and complement AI capabilities.

• The adoption of generative AI in core banking systems is still in the early stages, and it may take several years before it becomes heavily used.

• Data privacy is a top priority in AI applications, and personal data needs to be anonymized and protected.

• There is a need for clarity in distinguishing between different AI technologies, such as generative AI and machine learning.

• There is potential for AI to provide valuable insights and services beyond traditional financial offerings, but there is a need for more exploration in these areas.

Transcript:

Alicia Roisman Ismach (00:06)

Welcome and thank you for joining us today. This is on our series of Atlantic FinTech webinars to bring more information and awareness to our founders, FinTech founders in the region of Atlantic Canada. And I'm very happy to have you today and I would love if you can do an introduction about yourself and about the topic we are going to discuss today.

Stuart Levine (00:09)

Thank you, Alicia.

Okay, yes. Thank you, Alicia. Thanks for having me. Again, I'm Stuart Levine. I am a fractional FinTech consultant, having spent over 40 years in the financial services industry, banking payments, and as well as wealth management and capital markets trading areas. I am currently helping FinTechs to scale and grow with everything from business development, revenue generation, partnership development, raising capital, M&A transactions, number of different areas involved with growth. I also have a deep and expansive FinTech partner ecosystem consisting of individuals who are former C -level executives at banks and fintech companies with expertise in everything from payments to lending to risk, compliance, regulatory issues, a number of other areas. So I work hand in hand with my partners in my ecosystem to help fintechs grow and scale, as I mentioned, as well as mid -size and community and regional banks as well with their talent management and other growth needs.

Alicia Roisman Ismach (01:45)

We are going to discuss a very specific topic that is very relevant for what we know is happening in the industry today.

Stuart Levine (01:47)

Yes: the topic is generative AI and financial services.

Alicia Roisman Ismach (02:00)

Yes, affecting a lot of plans and a lot of decision -making, especially in the financial services industry itself, but also in FinTechs that have been planning on their solutions, bring to market and grow. And of course, generative AI will have an impact on the technologies they use and the services they give. So...

I wanted to hear more about that perspective from you on how the industry and the fintechs may be affected or are already affected by this.

Stuart Levine (02:36)

Yes. So let me kind of break break into two parts. I guess the first one, let me talk a little bit more about financial institutions. And in my experience, short time being involved with generative AI, I mean, I I think AI really just was commercialized less than two years ago. So I think what we're seeing is a lot of POCs (proof of concept), some pilots.

You know, maybe a few wide scale, you know, production systems, but most of it is still in the early adoption phase, at least from my perspective. And the financial institutions, whether they be traditional banks, digital banks, payment processors, lending companies, wealth and asset management firms, insurance companies, all the different...

aspects or sectors within financial services. I think the larger institutions, banks specifically, are starting to adopt generative AI capabilities in everything from in the lending area, credit scoring, for example, document management and analysis, customer experience, customer service and customer support, pretty much cuts across all fintech sectors and financial service institutions.

Obviously, productivity improvements through automations and efficiency and middle and back office operations. Research, equity analysts at brokerage firms involved with trading and portfolio management analytics using generative AI for research on public companies, as an example.

There's virtual assistants. We've all heard about chat bots and things of that in the wealth and asset management space. So the list goes on and on. Actually, one of the bigger ones I left out, fraud, fraud detection and suspicious activity involved with payment transactions and regulatory compliance, know your customer, AML, BSA, all of the regulatory issues. So.

It's certainly being discussed and talked about from both a large institution perspective, as well as from a FinTech product company perspective. They're all starting to embed generative AI capabilities within their products and platforms that can't be leveraged by either consumers or the small business market or the corporate market. So that's kind of a, I know it's a... seemed to cover everything, a little kind of a broad overview of where I'm seeing activity and application use cases of generative AI.

Alicia Roisman Ismach (05:27)

One of the things I wanted to ask you that on top of this amazing capability, especially when we are talking about large amounts of data to bring some processes and sense within it, is that it also has a high tendency of not being completely accurate in its responses and can sometimes even create scenarios that do not exist.

And this happened already when it was tried, for example, for court services, for legal, or it has been also tested in some medical environments and created situations that the professionals must be there to make sure that it's not used incorrectly. And it's not because of the wrong prompt. It's really a problem that is still being identified. So my question is, how long until the financial industry really feels the trust to bring AI to be a full solution for certain areas in the financial industry? You mentioned, for example, risk or fraud, or you mentioned, for example, underwriting when we are talking about the loans and how... well is generative AI today when it's tested in financial institutions in the results it's bringing compared with solutions that involve, for example, machine learning and processes that are defined by developers or data scientists as compared with a completely autonomous AI engine that will take all the data and decide, you know, where is that out of it?

Stuart Levine (07:22)

Right, right. Great question. I think it's still very early days, as I mentioned. We're still, so everyone's feeling their way in terms of how to use it, monitoring the results. You had mentioned, and correctly so, there are still sometimes bad results from a generative AI, so let's say a Chat GPT query and sometimes you'll get back incorrect responses. So how do you manage that, particularly if you're a financial institution or a fintech company? So I think a lot of that lies in how the company is really effectively utilizing the capabilities and the power of generative AI. In my mind, it's always risky and dangerous to create some kind of a... generative AI application and think, okay, I don't need any human interaction. I don't need to have a person look at the results and make sure it makes sense. So there's the danger if someone thinks or company thinks that generative AI is going to replace human interaction. I think generative AI was really meant to complement humans.

You know, to help them with their job functions and make things more quicker, more efficient, you know, improve productivity. You know, so the way I look at it is it's complementary to, you know, what a human is doing in a particular job function. So, I mean, I hope that answers your question.

Alicia Roisman Ismach (09:00)

Yes, one of the, we have seen situations where they have deployed already, for example, in AI chat function for consumers to communicate with the company. And I don't want, of course, to mention any names, but it was taken down after a while because people were asking questions and getting really responses that were completely, not only sometimes incorrect, but they shouldn't have been given to the customer.

So this is of course a trial and error scenario right now, but on the other side, are we really applying AI to lower those manual side of the operations or manual situations where intelligence can really bring value? Or are we trying to apply AI just to lower the number of employees (in) the company. It's a completely different reason to use AI. And I'm not sure that the second case brings really the value that may be cutting some costs there, but does it really a value driven decision or is it more about the kind of cost profit decision that may not make real sense?

Stuart Levine (10:26)

Yeah, excellent question. I think there are some institutions that are looking at AI as a way to reduce headcount, for sure. I think that is something that needs to be looked at very carefully, particularly if you get into this, what we discussed earlier, that if you're using AI to make decisions, interact with customers without some kind of human complementing the function. So I think looking at it from a headcount reduction and cost reduction type of value is probably a little dangerous right now. I mean, could it ultimately result in... that and maybe lowering cost, fixed cost and headcount, maybe at some point in the future. But I think where we are right now, I don't see that happening all that much. And maybe others are seeing it happen and there are companies that are strictly looking at it from a headcount reduction point of view. Whereas I think that's, it's early and I think it's a dangerous way to go at this point until the technology is really proven and is more, you know, has a higher degree of accuracy and precision because as you know, you know, sometimes you get false responses from it. So, you know, I guess the answer to your question is it's the jury's still out in my opinion, but the way I look at it.

Alicia Roisman Ismach (12:13)

I've seen also one of the things that I've seen is that the companies, fintechs, that are coming now with solutions to the market, especially those that are targeting financial institutions, don't necessarily have better prospects in engaging with financial institutions by having more AI, meaning some solutions are AI driven, are AI based, but not all solutions are AI based. Some solutions use AI to improve the results or improve certain services that they can provide, but are not necessarily an AI solution. What I have seen is that when the solution is an AI solution, a pure AI solution for banks, that doesn't... resonate with all financial institutions. It brings actually to some slower adoption by the market, like slower entry to market.

How far are today the financial institutions, especially when we talk about the large number that you have in the US, how far are we today to see AI as a normal type of technologies that banks are using, adopting and applying against just doing some pilots or some trials, but that are completely separated from their regular products they use and offer.

Stuart Levine (13:50)

Right. Hard to say exactly how far away we are from getting to the point where it's really heavily in use, adopted, AI technologies embedded into various areas within an institution. Because as I said earlier, I think we're still very much in the pilot phase. I think moving from POC to pilot stage is kind of where we are right now as based on everything I'm seeing and reading and talking to financial institutions. I think everyone's trying to formulate their strategies. So I think that's where we are now. As far as how long will it take to get there, to get into a full production type of mode, it's anybody's guess. It could be a year, it could be two years, it could be longer. But everybody is moving kind of slowly because of all the press and all that's in the meat about how dangerous AI technology can be and could result in machines taking over the world and all of that kind of thing. So if I knew how far away we are and when massive adoption will start, I think I can make some investments now and be a very rich person. So I don't, looking at my crystal ball.

I think, I think it just feels to me that, you know, once we kind of move out of the pilot phase and there are some successes and some results and, you know, senior executive leadership teams can see the ROI. you know, then I think you'll start to see a quicker, you know, an acceleration of an adoption of the technology and incorporated into their mainstream operational and production systems. So I don't know. One plus years is my best guess is the timeframe.

Alicia Roisman Ismach (15:45)

Yeah, we have seen that the financial institutions are not the first adopters necessarily in any technology.

Stuart Levine (15:52)

No, especially in the insurance sector and in capital markets, trading, wealth management, even some areas within commercial banking, retail banking, they're a little further ahead, but insurance tends to be a little bit of a laggard.

Alicia Roisman Ismach (16:11)

That's interesting, actually, something that I didn't think about at all. Do you think we are seeing a lot of solutions coming with AI that involve the interaction with the client or are related, as you said, for example, risk and fraud are in the back end, but are related to transactional activities that being between businesses or the consumers is not important right now. But, do you think that when we go further into the core system of the bank and the do you think that the future of generative AI could be also found in replacing these huge core systems of tens of billions of dollars that no one really is jumping into replacing so fast? Could generative AI come with a response to that problem that we are all knowing for decades and they bring a new era into course banking where real time and intelligence goes hand to hand in the near future is something that we could dream about.

Stuart Levine (17:19)

Yeah, so just to make sure I understand the question. So when will we see AI being incorporated into core systems is what you're saying? What you're asking. Okay, yeah. Yeah, right now it's very much a front end with obviously integration to the core backend systems in banks and... wealth management, asset management firms, et cetera. So I think over time, we'll probably start seeing as core legacy systems start being replaced, I think there will be from the product vendors or even some of the homegrown custom built core systems. More and more, I think we'll see AI capabilities and functionality as part of the of those software applications. But again, I think we're several years away from actually seeing that. I think some of the, in the banking sector, for example, if you look at some of the core banking legacy systems, Temenos and Jack Henry and nCINO, some of the product vendors have product roadmap plans, I'm sure underway to incorporate AI functionality into their next generation of core banking systems. So I think it's definitely coming in the future. I don't know how far away it is. I think once the technology is proven, then they can really release.

Those new versions of their products. It is down the roadways. Again, I don't know how far, but I think just about every banking or financial services product vendor, I'm sure has it on their product roadmap plans to just make AI technology a part of their core offerings. That's what I'm seeing.

Alicia Roisman Ismach (19:29)

It's just that we really can't know at this stage how far it is. And it really could be a game changer. The moment you see core systems that have been developed, implemented or embedded with generative AI that could completely change the way the financial services are delivered.

Stuart Levine (19:47)

Right. Yeah, absolutely.

Alicia Roisman Ismach (19:48)

My question now would be more about the data that is used for these AI applications and systems. And we have seen that in other applications, there have been a lot of discussions around copyrights and the use of data.

When we talk about individuals using banking solutions, we expect of course bank secrecy act and the privacy of the data of the user will prevail even with AI solutions. So it will need to be anonymized and you cannot share information in a way that may surface in response to another person in the bank, it's very important to keep things separate.

What happens when we are talking about business information and the use of AI could bring a lot of value on one side to improve your business, to improve the management, financial side, the management side of the business for businesses using their banks to manage their finances. But on the other side, we know that some of the competitive advantage of a business sometimes is in the way of they manage things and the AI uses enormous amounts of data to get to the result they need to provide. Do you see that the privacy or the ownership of the data?

Will become such that the businesses will be also compensated for the data they provide to the systems that end up being trained with that data or you see more of the same where that is used and they will try to keep it as private as possible and that's the situation that we will see for a long time.

Stuart Levine (21:45)

I think more the latter. I think data privacy right now is such a big issue. And some of the AI -based systems that I've seen so far, which hasn't been all that long a period of time, a lot of the data is anonymized, as you said. PII (personally identifiable information),

being removed, you know, if that data is being incorporated into some kind of an, you know, an AI process. So I think data privacy will always continue to be, you know, a top priority. And, you know, I think the companies, you know, banks, fintech companies, you know, when they are involved with using

a consumer's personal data or maybe a small business customer or a corporate client's data, that has to be, if not the number one concern and priority. It's certainly in the top two. So I think it'll continue to be a very, very big issue, particularly when that personal data is being utilized.

in large language models and things of that nature. It really has to be addressed for sure.

Alicia Roisman Ismach (23:15)

And another question I have for you is that we talk about financial information and the privacy, but many of these applications of AI, and you mentioned that when you were introducing the topic, are related to documents, to document the understanding the documents, managing them, and also making decisions based on data from documents. Now, documents specifically when we are talking more about the bank, services to companies or businesses of all sizes. Documents have more information than just financial information.

And that can bring also to understanding of the customer by the bank in a completely different light of what they are doing, what is their activities or how well they are doing those activities, how they can grow, even provide services that are beyond the financial services. Have you seen any interest or exploration of financial institutions to provide more services that are related to data that is beyond the transaction because of the amount of data they will have once they are capturing full documentation, especially that is commercial documentation of course, because they will learn about that business, something that is much beyond revenues and costs.

Stuart Levine (24:40)

I have not yet seen that. Wouldn't surprise me if the financial institution is doing everything they can to understand the data, the customer data, and try to capitalize on it. Maybe if it's consumer -related, could be used for marketing purposes, for promotions, based on the data that they capture from the documentation that the consumer is completing. So, I mean, that's being done today. And I think it will continue to be done, particularly leveraging AI technology even more. But as far as for commercial purposes, like, I don't know if you were thinking selling the data to third parties or anything like that. I just don't see that happening on a large scale and still addressing the privacy issue. So I don't know if you're seeing it, but I haven't really seen any examples of that.

Alicia Roisman Ismach (25:54)

Yes, even, you know... even giving the clients some ideas on how to grow their businesses because the AI have analyzed all their activities beyond, as you said, their revenues, including the type of agreements they have with their clients or how many employees they pay for payroll. It can bring them a lot of insights on how to grow their business. I have not seen anything like that, but it will be interesting because there will be a lot of cross -information between the different organizations using AI and servicing the public, consumers or businesses, that will have information that was not originally the type of information that that service provider needs to provide their service. So we will see a lot of cross -services.

Stuart Levine (26:47)

Yeah, yeah, yeah, I think, yeah, like I said, in the corporate banking, maybe in the small business SMB market, right, if those businesses have a loan, for example, with a bank, there's a lot of information there, financial balance sheets, income statements. So yeah, the banks could certainly put those documents through some kind of an AI engine.

Scanning those documents, looking for insights, mining the data so that they can then come up with recommendations to those clients and say, hey, based on your cashflow, have you thought about doing this? Have you thought about this service we provide that could help you with your cashflow needs or your working capital needs or whatever it may be? So yeah, I think for... those types of internal use within the financial institution, I think that we will start seeing more of using AI. But as far as trying to commercialize the data to other parties, I don't see that happening.

Alicia Roisman Ismach (27:58)

I see also one more trend before we finish our conversation today is machine learning and AI being used for the same solution. Some would say this is an AI solution and they will say this is a machine learning solution. And they, or they will call generative AI to a solution that is much more limited in scope, that does some very specific services. It seems to me that also the industry needs some preparation to understand what type of solution they are really reviewing or really testing or really piloting. And to start differentiating between the type of technologies that involve a lot of intelligence in them, sometimes algorithmic, sometimes it's learning, sometimes it's generative AI, but the mix is pretty high right now. We mix the terms and call everything everything.

Stuart Levine (28:59)

I totally agree with you. I think that, you know, I've been involved with machine learning for probably over 20 years. It's, you know, it's an older technology. You know, Open AI and Chat GPT and everything, you know, that Google is doing and Amazon and Microsoft and Meta, you know.

Those are sort of the next generation of AI technology. And what really made OpenAI, Chat GPT popular, a little less than two years ago, when it first was introduced was that end user interface, the ability for any individual just to say something or type in a query and get back a result. On the back end,

The technology, the large language models, and the other processing on the back end is really still using, I think, the older machine learning technology, which has been improved and enhanced over the past 20 some odd years. But the real distinction to me with generative AI is that front end, that user interface.

I think there's a lot of confusion in the terminology. If you ask somebody what's the difference between machine learning and generative AI, I think they struggle to explain the difference. Unless you're a data scientist and you really know this stuff. But yeah, so I agree. There's definitely confusion. There's a lot of hype. And I think it's going to be a little while before everybody really understands the distinction between the older machine learning technology and what we have today.

Alicia Roisman Ismach (30:47)

Yes, and not everything requires AI. Some things we still need to use the full scope of technology, not only AI specifically. And one last thing is that I really enjoy looking at what's going on from one side, but from the other side, I think that we are not looking at the use of AI broad enough.

Alicia Roisman Ismach (31:17)

And I'm talking about mainly entrepreneurs and fintechs. I'm not talking about us as the people looking at the industry. I think that they are all focused on very, very specific use cases that have been already explored or people are already exploring again and again and again when there is so much that could have been done that will bring tremendous value to the industry, even when we are talking about financial inclusion or we are talking about the underbanked. I see a lot focused on the same, as you mentioned part of it, you mentioned wealth management and you mentioned savings or loans. I think that we both have seen a lot in each of the areas that are the top popular ones, I would like to see a little more on the part of things that bring value to humanity within the financial industry. So let's see and hope to see more of that as well.

Stuart Levine (32:12)

I totally agree with you. Couldn't agree more.

Alicia Roisman Ismach (32:31)

Stuart, I enjoyed our conversation and this is just the beginning. I really appreciate your time and we'll see you soon again.