We're sure you've heard a lot about AI recently. But you may still have some lingering questions around whether it'll actually help your bank.
In this episode, Maria Abbe chats with George Neal, Chief Analytics Engineer at PrecisionLender, about the different types of AI and which one will bring the most value to your bank. The conversation is based on the article, Unlock the Remarkable Power of Pragmatic AI .
Helpful Information
Unlock the Remarkable Power of Pragmatic AI - Chatbots Magazine
Podcast Transcript
Maria Abbe: Hi and welcome to the Purposeful Banker, the podcast brought to you by PrecisionLender, where we discuss the big topics on the minds of today's best bankers. I'm your host today, Maria Abbe, content manager here at PrecisionLender, and I'm joined today by my cohost, George Neal. He is our EVP of analytics here at PrecisionLender. The title of today's podcast is The Type of AI That Truly Matters At Your Bank. We'll be basing much of our conversation today on an interesting article called The Remarkable Power of Pragmatic AI, and the link to that article will be in the show notes.
I imagine some of you have already groaned out load when you heard AI mentioned, and chances are you've been hearing a lot of it lately, but chances are also high that you've yet to be shown why it's so important to your bank. That's why we brought on George to talk us through this article that really sheds light on why AI is important within your bank and how it can be applied. Welcome to the podcast, George.
George Neal: Well, thanks a lot, Maria. It's good to be back.
Maria Abbe: Yay. We'll jump right in here. The article sites Forrester research that indicates that 58% of enterprises are researching AI, while only about 12% are actually using a model. We'll start here. What does the disparity between those two numbers say to you?
George Neal: Well, it says a bunch of things, but I want to point something out. This particular research says 58% of enterprises are researching AI, 12% are using a model. Model and AI aren't always interchangeable terms. The thing that these numbers really call out to me are a few points. The first of which is, there's clearly interest at the enterprise level. I mean, when 58% of enterprises are actively working, spending time, money, effort on finding out about AI, it seems to say that the businesses know this is coming. They know it's important. They know it's going to impact delivery channels, their customers, their industries, et cetera.
It also shows, though, that a much smaller percentages of businesses are actively using the technology now. While that might not be surprising, I mean typically you expect to see a small group of early adopters and then a bulge bracket of adopters, what I think is kind of shocking is how big the falloff is. If you consider 58% of businesses are researching, 12% are using, that means you've got a 30% laggard group. I think that's a direct reflection on the speed in which this technology is coming into various industries, the speed of evolution of AI and machine learning, which I believe this article is treating almost interchangeably. I think that that says a lot about the potential leave-behind capability of this technology. There's a lot of potential for industries across the board and for businesses across the board to be left behind if they don't put some attention here.
Maria Abbe: Interesting. Then the article, it makes a key point that not all AI is the same, and it talks about two distinct forms. Open or pure, as they call it, AI, and then pragmatic AI. Can you give us a brief explanation of each of those forms?
George Neal: Sure. In this context they're saying pure AI is unconstrained by the specific task that's trying to be done. It's capable of learning in a broader context and it's much more of the science fiction movie kind of AI that everyone loves to imagine, where you sit there and you say, "Hey AI, how does this outfit look?" It gives you an answer. Then you say, "Well great, book us a table, and by the way, make sure that we have that meeting scheduled tomorrow." A very dynamic, diverse, and open AI that could be applied to any given context.
By contrast, the thing that they're labeling as a pragmatic AI is a very context-specific, results-oriented, goal-driven set of technologies. It's purpose built to perform a function and make a specific type of insight. When we think about that, you start thinking about things like, even as simple as your map application uses what is effectively an AI to route you. Now, that same AI is not the one that you're going to use to book your next flight. That's a very pragmatic piece of technology.
Maria Abbe: Got you. If I'm remembering correctly from the article, the open type of AI would be like what Google is building.
George Neal: Right, so Google, Amazon, IBM, Microsoft, they're all very, very active in this space. It's interesting to note, though, they're also active in the pragmatic space. Again, that mapping algorithm is a Google manufacture for a specific service offering. I think what this points you to is there are a set of industry leaders and technology companies that are trying to solve the bigger open case, but even they have acknowledged and have manifested these pragmatic AIs as very practical tools. I think that's really where enterprise is going to find the value in the shortened middle term of this wave of AI technology.
Maria Abbe: Interesting. I thought this part was also really important within the article. "AI as a solution on its own is weak," and that's a quote from the article. What do they mean by that?
George Neal: Well, there's a few meanings to it. In the broadest context, there are few of any businesses that are actively trying to sell you an interaction with an AI. I mean, at least I for one, and I'm in this space and fascinated by it, but if you said, "I'm going to sell you for $10, George, the opportunity to interact with an AI," I'd be like, "Eh, pass. I don't think it's quite ready for that yet." On the other hand, once you combine AI with a customer-focused interaction and with the ability to change what the customer experiences or how you deliver on the customer's expectations, then you've got power, then you've got value. Again, the ability to determine the route that someone's going to take isn't really useful unless you can deliver it on that phone.
Maria Abbe: Right. Then the article, it goes on to say that, "AI needs to be integrated with your current systems." I think that that's kind of what you were pointing to in the last answer there, but using banks as our basis here, can you give an example of what that would look like?
George Neal: Sure. Imagine if you had a customer churn prediction model and you had the ultimate of all customer churn prediction models. It was 100% accurate, but its output was a data file that no one ever read. Well, it's a great model with no value. If it doesn't generate change, it's not generating value. On the other hand, you integrate that same model with good CRM, you make it available to a call center, and then you have the opportunity to prevent customer churn every time a customer calls in.
You combine that even with, say, a next best offer AI, and you put that in your CRM and you put it in the hands of your bankers on the line, and then they have the ability to say, "Hey Maria, we see from these patterns you might be interested in our brand new product." Not only do you feel like you've gotten something that's been customized for you as a customer, because you have, but we've also have the ability there to generate a stickier relationship through that churn model and the combination of the churn model, the CRM, and the next best product offering. That's exactly where these kind of specialized pragmatic AIs are capable of adding tremendous value, but they must be integrated to where they can make a difference.
Maria Abbe: Got you. Now, would a chat bot that a bank has implemented within their own retail banking platform, would that be part of a pragmatic AI or a piece of that?
George Neal: Well, it certainly could be. If your chat bot is intended to change behavior, if you're offering that chat bot to get people to use that channel rather than call into your call center, if that chat bot is able to service your customers in a different way that your customers find more delightful, absolutely. That doesn't mean that your chat bot has to be back-ended by the latest and greatest machine learning. It means it has to successfully achieve those goals and leave your customer in a better place.
Maria Abbe: Okay. Then, Andi would also be an example of pragmatic AI at work as well. Right?
George Neal: Yes. Andy's specifically designed to be a context specific expert. She's focused on putting the smallest necessary bit of information on exactly the right set of eyes at exactly the right time to impact the opportunity that the banker and relationship manager is discussing with the client. To the article's point, delivering that information at that time to achieve impact, it requires integration, and Andy's integrated throughout our entire application as a result.
Maria Abbe: Wow, that's really cool. Then, next within the article they delve into long tail and fat tail solutions. They argue that open AI, it's made for long tail solutions while pragmatic Ai is made for fat tail. What are fat tail solutions, and then how do pragmatic AIs help solve them?
George Neal: Fat tail solutions in this context are frequently occurring tasks and solutions where businesses spend a lot of time. Typically, they have a finite number of outcomes or solutions. You're talking about areas where traditionally businesses spend a lot of time training people to give a fixed set of answers, or to determine which of a set of answers is going to be the right solutions. A good example, again, is that next best offer. Traditionally, we would train or banks would train their line bankers to look at a relationship screen, figure out what the customer currently has, do some deductive reasoning to figure out what they think they might take, and then offer that.
Well, if you think about it, that's an interaction that's fraught with execution risk, it's inefficient. It requires a whole lot of faith in the customer to be patient to go through that process, the banker to successfully execute it, for them to know what the most likely product is, and all of those are things that can be overcome by a next best product AI that does all that analysis for you, is integrated into a CRM, and puts it right there in front of the banker the minute the customer walks up. It's a whole lot easier, it's a whole lot more efficient, requires less time and effort on everyone's behalf. It's better for everybody.
Maria Abbe: Now, what about the deployment of pragmatic AI? How easy or difficult is it?
George Neal: I would start by saying it's a whole lot easier than trying to develop and deploy a pure or open AI, which is why you see Google, Microsoft, IBM, et cetera tackling those open AI questions, and then a whole lot of companies driving value out of pragmatic AIs. Now, just saying that it's easier than tackling a 2001 Hal kind of AI problem doesn't put it in a whole lot of context. I still would say that where we are with the technology right now, it's not easy. Integration is almost always a difficult task, and depending on your organization, it could be more or less difficult. Different organizations have different level of skill in achieving that. What I've seen with banks most recently is that there's been a whole lot more success in the buy versus build front, but there's a lot of factors that go into that, and maybe we should cover that on a separate conversation.
Maria Abbe: Next week.
George Neal: Sure.
Maria Abbe: Great. Well George, thank you so much. That was really, really insightful and really helped clarify a lot. Before we wrap up, is there anything that we may have missed that you'd like to cover?
George Neal: I wouldn't say we've missed it so much as I would like to say that banks, businesses, enterprises shouldn't let the fear of difficulty keep them from realizing the value of this. Clearly, 12% of businesses have already tackled this problem. Means it can be done. 58% are actively looking at it. I would encourage everyone that's listening, whatever group you're in, if you're in the 30% that isn't looking at it yet, if you're in the 58% that's looking that hasn't done it, whichever of those two groups you're in, try to move into the next group. If you're in the 12% that's already done it, do more, because it's coming. The technology's moving fast. It's a challenge to keep up with it, even if you're fully in the space. If you don't try to tackle the problem now, unfortunately you might get tackled by it.
Maria Abbe: Yeah, that's great. Well George, thank you so much for coming on again and sharing all that with us. That will do it for us today. Thank you all for listening. If you'd like to learn more, visit our resource page at precisionlender.com and again, we will have the link to the article that we just discussed today on the show notes page. If you like what you've been hearing, make sure to subscribe to the feed in iTunes, in SoundCloud, Google Play, or Stitcher. Of course as always, we would love to get some ratings and feedback on any of those platforms. Thanks again for listening. Until next time, this has been Maria Abbe with George Neal, and you've been listening to the Purposeful Banker.
About the Author
Follow on Linkedin More Content by Maria Abbe
