About the Episode

In Season 2 Episode 5 of Lending Made Easy, David Catalano and Bryan Peckinpaugh weigh in on how AI tools like ChatGPT can be put into action across different functions and lines of business for financial institutions. As artificial intelligence continues to revolutionize various sectors, financial institutions are harnessing the power of conversational AI to enhance customer experiences, streamline operations, and improve security measures. In this digital age, customers seek personalized recommendations and seamless interactions, making ChatGPT a valuable tool for banking professionals. However, as with any transformative technology, there are risks and considerations that must be carefully addressed. Join us as we delve into the possibilities and challenges of integrating ChatGPT into the world of banking, uncovering its potential and discussing how financial institutions can navigate the path to success.

FAQs About ChatGPT's Application in Banking

How can ChatGPT be used in the banking industry today, and what benefits does it bring to financial institutions and their customers?

ChatGPT can be used in the banking industry today in various ways. One potential application is using ChatGPT to identify the next best product for a customer based on their relationship, behavior, and the purchasing patterns of similar customers. This can help provide personalized recommendations without the need for human intervention. Additionally, ChatGPT can be utilized for fraud detection by analyzing transactions and identifying any anomalies or potential fraudulent activity. The benefits of using ChatGPT in banking include improved customer experience through personalized recommendations and enhanced fraud prevention measures.

What are the potential applications of ChatGPT and conversational AI in banking, both internally for brand perception and externally for customer support and interactions?

The applications of ChatGPT and conversational AI in banking are wide-ranging. Internally, financial institutions can use these technologies to gain insights into how their brand and products are perceived in the market. They can ask ChatGPT questions about their brand, products, and market positioning to understand how customers might perceive them. Externally, ChatGPT can be integrated into online chat bots or customer support strategies, providing customers with a conversational tool to interact with the financial institution. This enhances the customer experience by offering more natural and interactive communication channels.

What are the risks associated with implementing AI, such as ChatGPT, in financial institutions, particularly in terms of data security and the potential for incorrect or misleading information being generated?

Implementing AI, including ChatGPT, in financial institutions does come with certain risks. Data security is a significant concern since sensitive customer data is involved. Institutions need to ensure that the AI systems have strict access controls and safeguards in place to protect customer information. Another risk is the possibility of incorrect or misleading information being generated by ChatGPT. While AI systems are constantly improving, there is still a need for human supervision and training to interpret and validate the output of the AI tool. Financial institutions should establish robust processes for verifying the accuracy and reliability of the AI-generated information.

How can financial institutions mitigate risks and train their staff to interpret and utilize the output of ChatGPT effectively, ensuring accurate and meaningful interactions with customers?

To mitigate risks and effectively utilize the output of ChatGPT, financial institutions should focus on training their staff and establishing clear guidelines. Staff members need to understand how to interact with ChatGPT, ask appropriate questions, and provide the necessary level of specificity to receive meaningful responses. Training should also include educating employees on the limitations of ChatGPT and the need for human oversight to ensure accurate information is provided to customers. Establishing a feedback loop where employees can report any issues or discrepancies encountered with ChatGPT is also crucial for continuous improvement and refinement of the AI system.



Mitch Woods: Welcome to today's episode of Lending Made Easy. We're gonna be exploring the intersection of technology and banking and, specifically diving into the world of ChatGPT a pretty hot topic across a lot of industries today, but especially in banking. just because as financial institutions continue to digitize their services and provide personalized experience to their customers, the role of AI is becoming more and more important.

David, I'd like to start with you today. How could you see ChatGPT being used in the banking industry today? And what benefits do you think that could bring to a financial institution and to its customers?

David Catalano: That's a great question, Mitch. I, I think they could potentially use ChatGPT in identifying the next best product for a customer. And there are various solutions out there today that do that. But I think this particular solution may be conducive to that. Looking at the complexities of relationships and behaviors and what other customers that look like this customer have already purchased and identifying propensity and capacity.

You know, with respect to the solution set the person already owns the circumstances we understand as a bank and, being able to take in and consume all of that information and come out with recommendations that would make sense without human intervention. Simply letting the thing run. I could see that happening.

And then that secondary would be just around fraud detection. So think about that same technology, propensity, capacity, identifying new products, but flipping it on its head and saying, what's different about what's happening? All of a sudden, I'm using a credit card in Florida to buy gas. Well, I just used a credit card the other day in Carmel, and that's, you know, a thousand miles away.

How would, how could that be, you know, putting together different scenarios that don't make sense based on previous customer behavior. And I also know that that's done today with credit card companies, but I just see this as a way to really analyze transactions and look for propensities, capacities or the opposite of that.

The opposite being potential fraudulent activity. Bryan, what are your thoughts on that?

Bryan Peckinpaugh: Yeah, I think it, you know, probably important first, we're, we're using chatGPT almost as the noun or the verb here, right? Like we say, Kleenex for facial tissue. It's really interesting to think about where AI in general, like you're talking about, David can be applied across financial services, but also the underlying, you know, conversational AI that is kind of spearheaded by ChatGPT and others in the market.

I think there's a lot of both internal and external facing applications for these types of technologies. I'm sure there's plenty of bankers out there, marketers at banks that are interacting with these conversational AI tools to try to get a handle on how It thinks the financial institution is portraying itself to the market.

So, you know, asking It questions about your brand, about your products, about what It sees, and using that as a proxy for your general customer base. I would have to imagine there's people out there doing it. if not, that's one of the first things I would be doing. Making sure that it's interpreting my brand the way that I want to see it so that if anybody is asking it questions about me it's surfacing the information I would wanna see at surface. I think it's gonna play very heavily and already is in some instances in how the FI's interact directly with their customers. Think about embedding these kinds of technology into an online chat bot or weaving it into your customer support strategy, taking away the push one for hours, push two for what have you,and having an actual conversational tool to consume, either through an online channel or maybe even eventually through a phone channel and using that technology behind the scenes to guide them through an interaction with the financial institution.

So I think there's a lot of ways that it can be used. You know how exactly it is today. I think it's just early, early on. But certainly something we should all be keeping our eyes on, and I think lots of, you know, potential implications.

David Catalano: Yeah, I agree with you. Someone has to call and interact with this thing for it to work its magic. And you know, in my limited interactions with it, it didn't really have depth in the areas that I was asking it questions about. So building out that depth and I'm not sure how that actually occurs other than feeding it more data but building out that depth. So that does have robust answers to what, you know, what product do I need. You know, I'm calling the bank and interacting with this chat bot. It would have the ability to know everything I currently have and what other people that behave like me or have assets like me or liabilities like me, could potentially also have and offer those up as well.

 I could see that as being in, you know, a unique way to use the tool, but the knowledge of the tool has to be built out. Cuz if it's not built out, then you just get superficial, or at least in experience, you're getting superficial answers.

Bryan Peckinpaugh: Yeah, that, I mean, that's the, the crux of all these machine learning ideas is they're, they're only as good as. A, the questions you ask them and B, the prior questions and the data that's fed into it. So to your point, David, the more they're interacted with, the more they're able to reach out into new data sources, the, the better they're going to be.

So, you know, back to some of your earlier points, weaving that into the fabric of your tech stack becomes that much more important the more data it has access to, the better it's gonna be, the more predictive it will be. That starts to worry me a bit too, is, you know, do you think about the type of data you could be serving to it you got a lot of security concerns there and making sure that, you know, it's, it's limiting the pool to just your data, et cetera. But it's also that, like you said, David, an educational issue that needs to be addressed. You know, how do I interact with these types of things? What types of questions?

How do I ask the questions? How do I provide the appropriate level of specificity to the questions, to make sure I'm getting something meaningful back and they'll continue to improve over time. You know, these things will continue to get better. You know, think about the fact that it effectively wasn't here 90 days ago, and now you're seeing, you know, iterations of a ChatGPT-like experience popping up everywhere.

David Catalano: Yeah, when I think about Siri or, you know, the, basically the lack of a keyboard and being able to talk to a device. So if you're just going to be able to talk to your bank or your bank app, And you say, well, I'd like to apply for a boat loan for $50,000. I could see that bringing you all the way through that process to the point where you're teed up the documents to sign, you electronically sign them, and the money's put into the account.

I think an experience like that is, I don't see that being a tough thing to do. I could be missing something, but, basically it's a voice command to your bank and the bank being able to navigate that voice command using AI.

Mitch Woods: Yeah. I think David, that's a, that's a great example of how AI could be used to really streamline some experiences for a customer. I think one of the scary things with that, and Bryan, I think you were kind of alluding to this, is if it doesn't know the answer, it's gonna make it up as well, you know, and present that as facts.

So there's still some adult supervision that's required with this, and I think to your point, training not only the tool, but training the people on how to interpret the output from that tool is, is important. So what do you see as maybe some additional risks with that for an institution that's gonna maybe dip their toe a little bit to the water of using AI and planning for, you know, augmenting their customer experience using artificial intelligence, whether that's on their lending business, customer service, really any, any area of the financial institution.

Bryan Peckinpaugh: Yeah, I mean there, there's a reason, you know, you look at our industry, you look at healthcare to highly, highly regulated industries. There's a reason for that. we're talking about very sensitive data. We're talking about very sensitive advice that you can get in those sectors. You know, I, I think back five years, six years ago, maybe more when the robo-advisor idea was really prevalent and, you know, trying to automate that or, you know, eliminate the need to have a financial advisor in every branch and, and try to provide technology driven, AI driven analysis of, what you should do with your money and never really took off. I don't think to the degree that people wanted to, because you're, again, you're playing right into what you were talking about, Mitch, of do I really wanna put those kinds of decisions in the hands of AI given its current state, you know, what are the fallout from that?

If I use a ChatGPT solution, and you know, Mitch, you ask it, how should I invest my $50,000? And it gives you bad advice. Where does that then fall back on after the fact? What does that look like? So we're gonna see it in drips and drabs, I think as we move forward, as they test out bits and pieces.

You know, it little lower risk if you do it from a customer support perspective and try to route, you know, try to provide basic answers and, or route you to the appropriate place as opposed to, Hey, I think you should put all $50,000 into the Shibu coin or whatever that, the one that, that, Elon Musk was tweeting about for a hot minute.

Mitch Woods: Uh, he probably knows a lot about, a lot more about it than I do so.

David Catalano: Yeah, that, that, that is interesting. I guess what I would say is, if I'm a bank, there are certain things or, or certain risk buckets that I would wanna use to test concepts, and some concepts will fail and some concepts won't. So these are bets, right? So what kind of bets are we making as bank managers, bank executives?

Where are those bets being made? And, you know, what could come about from those bets? So in, in our lending world, You know, there's a lot of people that don't use scored lending. It's been around for a very long time. We've had it since 2002. I mean, you could score a million dollar loan, using the SPSS score.

Many people don't even do that. Why? I, I just don't, I don't know. They don't trust it. Right. But could they do that with a small portion of their capital and have a portfolio of just scored loans? Sure. You know, we have another solution called Portfolio Monitoring, where you can actually monitor accounts, both deposit and loan for the same customer and determine, hey, I, I can auto-renew this loan and half my portfolio could be auto renewed and I could deploy all my portfolio managers toward the big loans that might have challenges and focus my energies in the right place.

But they don't do that either. So. I, I don't know. It's a, you know, it's a matter of where do you want to make your bets? What are those bet sizes? And what could potential outcomes be? But if you're not making bets, you're not progressing forward. I'm not, you know, I'm not sure what you're doing, just following along, I guess.

But somebody's gonna, somebody's gonna figure this out, and they're, they're gonna, they're gonna have it nailed. Maybe it's a website called Boat Loans by Phone or Boat Loans by App, I guess

Bryan Peckinpaugh: I don't know that sounds, that sounds a little dicey, but, you're, you're likely to see one pop up sooner rather than later. I would probably spin it on its head a little bit, as well. I think there's fantastic opportunities for efficiency in different areas of the organization. So, David is an example I'm thinking of a commercial lender leveraging it, you know, almost like a student would writing a paper to, to help with part of the credit memo. Hey, ChatGPT give me a writeup on the manufacturing sector in, you know, Buffalo, New York. Trying to leverage it for those types of activities to see if I can shave, you know, time off of certain tasks that just have to be done, that are more in line with what it is good at today. I think there's lots of examples like that across what the financial institution does to, you know, just help automate the administrative type tasks that are inherent to doing the business. That would probably be where if I was sitting working as a lender or somebody in the branch or, or what have you, those types of activities, you know, using it to try to automate a little bit of what I do from that day-to-day perspective.

Mitch Woods: David, any, any other final thoughts there?

David Catalano: No, I, I think it's interesting enough technology to try and deploy it in different areas of the bank. And if I was an executive at a bank, I'd be thinking about this. I just came back from an industry conference. These are financial technology companies and this was one of the biggest topics discussed.

So I would absolutely pay attention to this. it's, it's not gonna go away anytime soon, and there's no sense in being the last one to adopt it. You might wanna consider being a little earlier with this one.

Mitch Woods: Well, David, Bryan. And some great insights today, and really anyone out there listening, if you have questions, feel free to reach out to our team here or, or go to ChatGPT and, and ask it how, how it can help you as well. But thanks everyone for listening to today's episode of Lending Made Easy.