AI: Effective strategies to use AI holistically (plus a concrete path to ROI with AI)

In the age of automation and uncertainty, this session will explain how to implement artificial intelligence (AI) holistically to run your bank or credit union effectively, enhance key performance ratios and keep your institution financially healthy at all times, while at the same time recognizing digital & online lending  challenges with facing racial disparities with AI technology & low lending approval rates, how are digital banks bridging the gap? You will see a live case-study of a financial institution and learn how to do it yourself. 

Transcript:

Michael (00:06):

We hope you enjoyed the last conversation around generational AI and continuing in the AI track. We'll bring up the new panel for their discussion. So I'd like to introduce and please correct my pronunciation, Venezuela Carr, who is the CEO of the Vicar group. Then we have Akhila Rao, who's the Chief Digital Officer for Encore Capital Group, Midland Credit Union. And then finally we have Uday Akkaraju, who's the CEO of Bond. So I hope you enjoy this session and feel free to come up with questions after the group. Thank you guys.

Venezuela Carr (00:43):

Afternoon everybody. So nice to see everybody, everybody in the room excited about Artificial Intelligence. Anytime I'm in one of these rooms, I'm like, I'm not the only married in the room this time. I'm Venezuela Carr. I am the CEO of the Vicar Group. As Michael stated, I'm also the host of a television show called Down To Business would be where we talk about emerging technologies like artificial intelligence chat. We also discuss scams and broad litigation, true crime. So tune in to your local networks for that. So today we're going to be having a robust discussion about the holistic uses of artificial intelligence, which I'm very, very excited about. So again, we have and with us both having case studies that they are actively using for their companies. So we're excited to talk about it. I'm going to start with you Uday.

Uday Akkaraju (01:42):

Sure.

Venezuela Carr (01:43):

Just because Bob, I found extraordinarily fascinating when we talked about in our preliminary discussions, you guys, they are actually using this in ways that my technology and innovation advisor Corbin, who's in the room, I'll say that he had us embedded in using these technologies a year before they were public knowledge, Artificial Intelligence and Chat GPT. So some of the things he was telling me in case studies that could be used in banking and finance, they're actually doing Uday is extraordinarily build out this new program. So Uday, tell us a little bit about what you're doing with Bond AI.

Uday Akkaraju (02:22):

Sure. So as the name suggests, so we are an artificial intelligence company, but I think our AI engine is actually called an empathy engine, right? Because truly, I mean the thing is we have a lot of data, but the whole goal of our engine is to, can we really empathize with the consumer with the data we have and the data we don't have, and can we help the banks address those needs for the consumer in the shortest time? That's the goal for us. So that's what we do at the company.

Venezuela Carr (02:50):

And you know what I find fascinating about this? We had discussions with CEOs of large financial institutions that question the ability to use empathy and artificial intelligence. So the fact that you're coupling those I find fascinating as well. So tell us a little bit more about the empathy engine and how you're using it.

Uday Akkaraju (03:15):

Sure. I mean the thing is, I mean this is a question everybody asks. How can really AI be empathetic? So it's always the question, but the thing is, continuing from the last session, which we saw banks have a lot of data. Everybody has a lot of data, but can you really empathize with data? No. So for us, we really think there are two really important forms of data, of course transactions. And the second is conversations. So we take these two important. So when we go to one of our clients, the first thing we do is of course take the transaction data and crunch it through the empathy engine. And we know that we understand about the consumers probably only 40 to 45% because there's a lot of missing elements. And the second thing we do is we actually implement a bot. So now what happens is through the bot, we are not competing in the bot space, but we are using the bot to actually have conversations with the end consumers of the bank. So by combining conversations and transactions now we are actually really able to understand the consumer and understand their needs. So we are trying to empathize with them their needs. So for the bank, it's empathizing with their consumers. For us it's empathizing with the banks. So that's why it's truly an empathy engine. And we've actually had great success. I will share some statistics as we go, but I think going in that approach, balancing between your transaction and conversations is the right way to do it.

Venezuela Carr (04:36):

Absolutely. Now speaking of empathy, Akhila, your company uses this in a similar way, but a completely different way where empathy, and I'm intrigued by using it in your industry kilos in collections and how do we embed empathy in collections? Let's talk about what your company is doing with artificial intelligence.

Akhila Rao (05:03):

Yeah, absolutely. I mean, when you think about collections, I think the number one thing you want to think about is empathy, right? I mean these consumers have gone through a lot in their lives and we're here to help them feel better about what's coming next to them and how do you help them in that financial freedom or setting them up for that success. So the way we use artificial intelligence in doing that is helping us identify our segments of population so we can have the right conversations with the right people. Similar to any consumer. There are people who know exactly what they want and they want it to be very quick and easy. And then there's people, like we were in the track yesterday, humans like conversations with other human beings and how do you make that facilitate using as much of machine learning and artificial intelligence as possible and trying to get them to the information that they're looking for as quickly as possible. So that's how we use it.

Venezuela Carr (05:59):

And we wanted to really differentiate and make a distinction between artificial intelligence and machine learning. So Akhila take that and run with it.

Akhila Rao (06:08):

Yeah, I am going to try doing that and as a creator of a product, you can maybe help that more. So one thing I noticed is that when we talk about AI, there's so much, it's become this buzzword that everybody wants ai but they don't really fully understand what it means or what does machine learning mean? Or just that a product has ML and AI capability, suddenly it's the next best thing. But essentially I think as we've heard in a lot of these tracks, it's only as good as the data that we understand that we have within ourselves, whatever data we're collecting. I gave an example yesterday about in my world we collect a lot of information through our phone calls. Every phone call is recorded, every consumer interaction is recorded, and I can harness that data to help them figure out the next best action.

(06:57)

So that in my mind is the core of the solutioning here is can you take a consumer from point A to point B where they want to get to in the least amount of steps as possible, but allowing them to determine or allowing the company to determine the next best action for them. For example, if a consumer in let's say is looking for an auto loan, they're not really looking for an auto loan, they're trying to buy a car, how do you get them from making that right decision about what cars I should be looking at with what I'm qualified for from a credit standpoint and how do you mix the two together? And I think therein lies a little bit of both. I think I look at artificial intelligence as an evolution. You understand your data, you understand your consumer journeys, you build out paths for them to see how many ways they can branch out from each possible outcome. And that fed over time is machine learning, which eventually can then make the decision, which we call artificial intelligence. That's how I see it evolve, at least in the use cases that I've implemented.

Uday Akkaraju (08:11):

Absolutely. I mean now we have generative AI to add to it. So I think you've mentioned about AI and ml, but I think generative AI is actually creating net new content, whole net new content out of what is happening. But I think the example I the last panelist was about the soccer, if you asked Chad GPT who won it has data only 2018 and it gives an answer, which is not true or it gives an answer about men's soccer. But if you see in banking's data, I think you'll have lesser bias because compared to a large language model which is spread throughout the internet, there's a lot of bias. So many people are contributing it. It's very, very important for creators to actually look at bias. But I think in the banking world, when we're looking at transactions, I mean there is bias. I'm not saying there is no bias, but I think there is less bias. So it's an advantage the banks compared to the large language model providers, if they really use AI to address their consumer's need, I think it's going to be a huge win in the banking industry.

Venezuela Carr (09:03):

Yeah, exciting stuff. Uday, you are using your technology not only for banks and their consumers, but you're using it for banks and their employees in other industries, right? Tell us a little bit about that. I found that fascinating as

Uday Akkaraju (09:18):

Well. Yes. The thing is, I mean we actually are a typical B two B provider. So when we sell to banks, we help banks implement these AI models, get to their end consumers. So when we were doing it, I mean we also had interest from employers, large employers, small employers. It's like we need to take care of our employees, we need to take care of their financial health. And this is actually a great way, your empathy engine is actually a great way to do it. So we actually went ahead and implemented with employees. So the thing is, I mean, because the employer is rolling out directly to the employee as a non-traditional benefit, we see more engagement. That is what we see missing from the bank side. So banks have to really and consciously roll it out, explaining the benefits. I mean, we see a little gap there, but I think banks have to roll it out in a way we see the employees rolling out actually a benefit. We are caring about you. So I think we see that difference. It's working great in both sides, but I think there's a lot to learn from the employees to the bank side.

Venezuela Carr (10:17):

Absolutely. I do agree with that. And for those of you that are in decision making positions in the financial sector, a product like this to be able to provide all kinds of insight training to employees and be able to engage with them at a different level and in a level that's personalized, I think is important as well.

Uday Akkaraju (10:40):

Exactly. See, the thing is that, I mean there must be many banks in this room, but the most, I mean, everybody wants to implement it. We have done a survey and say eight out of 10 institutions want to implement an AI system today. But I think the most difficult thing is though, they have the intent, the biggest roadblock or the challenge is preparing the data. So every bank has data, for example, they have their data with their core providers and they have to have permission, they have regulation and all that stuff. So there is a complicated preparation stage to actually implement an AI system. That's why you see almost 85% of the AI projects don't add value. 78% of the AI projects don't have ROI. So because of that particular phase, we realized that. So as part of what we do, and we actually also encourage our partners, is we offer that service for free. So if a bank can actually have the data preparation stage done well then it's all r o I for them.

Venezuela Carr (11:40):

And we have employees at the Vicar Group and team members that are embedded in financial sector operations. I too woods was as well. And we know that sometimes things are presented to the employees, but there's really no true engagement. It's a check in the box. I think this makes it a little bit different because it's a more customized, personalized approach.

Uday Akkaraju (12:00):

Absolutely. That's why when I said about transactions and the conversations, so the transactions will help you identify the need. All of us, were doing all propensity, right? It's still propensity when you just look at transactions, but when you pair up the transaction with the conversation directly with the consumer, you validate it. It doesn't become propensity, it becomes need-based. So when you do that, there is engagement, there is trust, and that's why I'm saying marriage between transactions and conversations so important. It it's understated.

Venezuela Carr (12:29):

And Akhila, Uday mentioned two words that I think are important in your industry as well, engagement and trust. So tell us a little bit about the feedback that you're getting and what you're seeing as a result of using AI for the debt collection process and how you're actually helping people because you're actually doing a service for that person. So talk a little bit about that.

Akhila Rao (12:54):

Yeah, I mean absolutely right. So I think about when you're in banking, and I've spent a fair amount of my time doing that. Your primary goal is to acquire new consumers. So it's a whole different strategy on what do you want to do, how do you get more people to come to your, let's say site or your branch or what not to look at the latest products that you have and how do you get them to actually take it? Right now in the collections world, it's slightly different. I already know who my consumers are. My stress step is to get them to engage with me. And like you said, one is trust, and the other thing there that you mentioned is it's need based. So establishing that need for our consumers to say why should they work with the company and what's in it for them. I think that's what we spend a lot of our, and I think it's about how there are a lot of companies using AI within the search side of things and how do you make yourself more front and center when people are looking for solutions to engage with you. I think that's really been the focus in which we've been working on.

Venezuela Carr (14:03):

Absolutely. And can you talk a little bit about how your use of this artificial intelligence and machine learning allows you to distinguish the best approaches to reach and engage with that consumer?

Akhila Rao (14:19):

It's still a work in progress. So we are thinking about when you think about communication, there's a lot of things that you can do in terms of emailing. I think in one of the segments we heard that if the direct mail response rate is less than 1%, the email response rate is further below because how many of us have five different email addresses to give out so that you're going to get a marketing communication and you never want to look at it again? So how do you prioritize on top of it? You have the big domains, whether it's Gmail or Yahoo, they have their own mechanisms to see whether or not your email should even be on the right tab. For example, if you get on the promotions tab within Gmail, it's very hard to undo that. How do you stay on top of it? So I think there's a lot that you could use in terms of testing through that and figuring out how do you stay top of mind for your consumers. So that's another way, a simple way of engaging. It doesn't need a whole lot of artificial intelligence in doing that, but a lot of rule-based content-based things that you could do, which then you can evolve into something bigger.

Venezuela Carr (15:30):

Awesome. In the artificial intelligence and machine learning conversations, we hear sometimes a lot of fear, which I understand and we hear about the potential for replacing humans in doing certain jobs, which is real. It's a true thing. But what I love about how you both are using it is it's used to amplify how our employees engage in what we do. I know that you are using these models with some of your organizations. There was something that you mentioned in our preliminary discussion about they don't threaten how the models don't threaten, but give us just some case studies of how you're using it with the financial institutions you're working with.

Uday Akkaraju (16:17):

Yeah, absolutely. I'll give a use case with one of our clients, but I think see predominantly historically when companies or industries use technology, I mean the productivity increases. So when the productivity increases, the cost of goods and services decreases as a result. I mean your pay increases. So it's been historic. We are not reinventing the we. So there is a pay rise, the wages increase, there is economic development and all that stuff. So in our case, what happens is when we work with a bank, when we take the data to ingest, the first thing we output and show them is how much revenue they're missing on an average. If it's between a one to 10 billion bank, we see about six to $7 million just left on the table. So if they implement a solution, there is more revenue. If there is more revenue, they can create more new jobs and they can invest more in growth. So there is the fear, like any new technology, but I think we are proving with data that actually it's actually going to help us, augment us better actually in fact, create new jobs and increase the pay.

Venezuela Carr (17:21):

Absolutely. We talk about how we can amplify things. You guys, I was in LA last week in a completely different room of people. It was all Hollywood executives, television producers from every major network you guys all know, and some of the case studies that corpsman bought to us as our technology and innovation advisor, they were using expertly. And not only were they using them, like Uday is saying, it was actually creating revenue to create more jobs for things that artificial intelligence could not do and could not touch. So it's not just revolutionizing the banking industry, we know it's revolutionizing every industry, but it is amazing to see what's being done with this. And I would encourage you, even as we are approaching and looking at how we're going to be using this to see the good, what can we use it for? How can we use it? How can we harness it to actually accelerate potential, not only with our operations, but with our employees as well, which is what you're doing. Akhila, tell us about some feedback you are getting from your team members about how this is helping them.

Akhila Rao (18:27):

Yeah, definitely. It helps them focus on some of the more complicated things. I think that's the beauty of it, right? When you make the more mundane, more regular things, easy for your employees, it definitely allows them more room to do some of the more complicated things. That's the feedback we hear from our consumers too in finding it, but it's always finding the right balance. I know that when you think about some of these, whether it's generative AI, chatbots or any of the chatbots, it's hard for it to know what exactly the consumer wants. And that's a big pain point if you keep asking the same question and if the conversation is becoming more complicated, when is that handoff? When do you actually introduce a human to actually do that work? So I heard in a couple of different panels where we're checking for responses, we're seeing whether or not they're being effective, and how do you take the employee feedback to say, well, if a consumer asks this question, this is how I would answer it. So then that's a way to program it. So you make them involved in designing a process that works well for everybody. And that's something that we've been doing.

Uday Akkaraju (19:34):

Yeah, I mean to our banks, we say this, right? I mean in Silicon Valley they say only one person of startups succeed. 99% of them fail. Even in AI, if you see the implementations in banking, like I said, 85% of the projects don't see the result. So we actually tell them, just treat it as a startup. An AI project is like a startup. So when you kind of start a company and try to raise funding, you see two important things. One is your team, and the second thing is the product market fit. I mean, it's the same correlation. You treat this AI project at your bank as a startup and make sure you have the good team, you have the right team. It's so important because with a new subject like AI, knowing what is needed is not that important, but knowing what can go wrong is much more important.

(20:20)

So you have to have the right team, and then you have to absolutely see what is needed by your consumers. Because we see a lot of banks, one of our clients spend about 10 billion. The AI project did not see the light of the day because they were addressing the wrong need. So product market fit becomes so, so important, and there's so many modules. You have RVA, you have back office, you have lending, I mean you have retail. I mean there's so many AI modules. Identifying the right one to start with is so again, for that, you have to have the data in place and they have the data strategy in place.

Venezuela Carr (20:52):

I love it. Exciting. As we are closing up today, any final points? We've got about four minutes left. Any final points that you want to convey to the audience?

Akhila Rao (21:04):

My only thought would be that as you're investigating or thinking about whether AI is the right solution or not, whether you want to start or not, I think I believe that start small. Think about areas that really need help immediately and not think about this is the one size fits all thing that's going to come and fix everything. So that's my mindset on it. Treat it as a startup because when you're a startup, you really focus on things that you really need to work on and not 150 things that you need to fix. So I would start with that, and then I couldn't stress more that the data has to be in one place so that it's not all fragmented. And your user journeys have to be very, very clear on what problem you're trying to solve and what your user is already going through and doing a simple analysis that will show you how many clicks, how many documents, how many things you're printing, and that's a really easy way to think about, okay, here's the place I want to automate and

Uday Akkaraju (22:04):

Go from there. Absolutely bootstrap it. The second most important thing is if you're a bank, implement a chatbot. Everybody, the chatbots are everywhere. I mean, chat GPT, you have so many. The reason why stress upon that is you can have a conversation with a consumer. Even in this age, if a consumer wants to talk to a bank, it's just so difficult. So having a chatbot is like default. You have to have a chatbot. And third one, my team is here. They'll actually scream at me if I don't talk about our benefit. But I think they said the data piece, it's so important. I think we offer that service for free. I mean, it's not that we make any, but we want the bank to be really, really successful. Focus on the data strategy, get the data in place, get the right scripts in place, because running a script on a core, it's difficult. It seems easy, but it's not easy. So those three things I think are super important.

Venezuela Carr (22:54):

Is anybody in the room exploring AI already? Yeah, yeah. It's an exciting time. Are there any questions? We've got two minutes. Any questions for Uday or Akhila myself?

Audience Member (23:08):

I think I got one.

Venezuela Carr (23:10):

Yes.

Audience Member (23:11):

I think a lot of conversations that around prompting chat, chatbots, prompting AI, any thoughts on going the other way, using that's forward outreach to actually engage customers or ask questions rather than keeping it a search engine?

Venezuela Carr (23:31):

That's a great question. He said versus instead of using it as a prompt, are we investigating using it to ask the questions of the consumer to gain information?

Uday Akkaraju (23:44):

Absolutely. Right. It's very advantageous in banking, right? So in terms of chat, GPT, you only actually give prompts to it and it gives you back the answer. But in banking, you actually can be proactive. You know the data of the consumer, the bank can actually give that prompt, be proactive and let the consumer react to it. So banking has an advantage because it has access to consumer data to be both proactive and reactive. So that's the advantage. That's why I said implement a chatbot and use this.

Venezuela Carr (24:12):

Wonderful. Well, that is all we have for today. We have barely scratched the surface of artificial intelligence and chat, GPT, and these holistic uses of Bond and what you're doing and Akhila what you're doing. Thank you both so much for sharing and for bringing light to the side of artificial intelligence that we're not really seeing how it can actually enhance what we do every day. Thank you so much. I'm the V. Carr and we have had a great session with you. We will be around if you all have questions. Thank you all so much for coming out.

Uday Akkaraju (24:43):

Thanks.