Harnessing the Potential of Generative AI in Banking

Nancy Ozawa (00:06): Hello and welcome to Banking Out Loud, the podcast where we have candid conversations and we deliver insights on banking topics that our listeners want to know about. I'm your host, Nancy Ozawa, and today we're gonna talk about artificial intelligence, or we call it AI. Uh, it seems like nowadays in the news, in industry conferences, press releases everywhere you look, they're talking about AI. And you know, recently the president of Nvidia, uh, Jensen Huang, stated that the interest in generative AI is emerging globally. And that's boosting the demand for advanced AI chips. (00:41): So today, let's explore, artificial intelligence and how it's unfolding. Let's see if it can deliver on its promises or not, and how to get started since some of our listeners probably have not got started. So, to start that exploration of this topic, we've invited Kendra Ramirez, the CEO of KR Digital Agency, a digital and AI agency known for helping organizations to harness AI solutions and how to make it an integral part of your organization's growth and innovation. So, welcome, Kendra. Kendra Remirez (01:12): Thank you so much, Nancy. I was so looking forward to our conversation today. Nancy Ozawa (01:16): Yes, I am too. And you know, I, I've seen you all over the conferences. So, you've been out talking with everyone in, in a variety of different industries. not banking and financial services, but a wide range of industries and a wide range of groups. So, I wanna make sure we touch on things even outside of banking. So, if there's lessons to bring in, definitely, let's bring that in. Kendra Remirez (01:37): Absolutely. Nancy Ozawa (01:38): But before we jump into the topic, which we're all excited to do first, why don't you give our listeners a little bit more background, about yourself and how you got into this, um, and then we'll kind of segue into AI. Kendra Remirez (01:50): Perfect. Perfect. Yeah, I'm a technology nerd. I just love dabbling with technology and self-taught in many areas. It just connects for me. you can't learn technology by reading a book. You have to just do it. And so, growing up, in school learning, you know, in that manner didn't really work for me. But once I fell in love with, technology and found technology, then I was like, "Okay, this is it. This makes sense to my brain." And so, in my early days, I was doing a sales for data center network architecture in the first big move of, you know, the dot com era. (02:25): And then from that, I started my digital agency almost 19 years ago. And so 19 years ago, a lot of people weren't talking about digital. And, even to the point of I gave my business plan to three business coaches and they're like, "Ah, I think that's a fad. I don't think you should do it." I'm like, "I think you're wrong. I'm gonna do it anyway." (laughs). And so with that- Nancy Ozawa (02:43): Glad you didn't listen to them (laughs). Kendra Remirez (02:45): Yes. Yeah. you know, and with that, you know, know, we, extended into AI in our agency. Even prior to ChatGPT coming out, we were using tools to help brainstorm and strategize and create content and imagery. Uh, we were building bots in 2018. And so the things that we were learning, I just started sharing with our clients and our clients were telling other people. And the next thing I know, I'm on the road doing 54 AI talks across the country last year, (laughs), and working with hundreds of businesses on navigating AI. And it's just been so much fun in understanding the practicality of it versus just theory. 'Cause I still think a lot of people are in that theory bucket of this is, you know, what's possible and where I'm taking it down to what's actionable. What are things that you could literally take away from this podcast today and try or implement? Nancy Ozawa (03:38): That's fabulous. And 54 conferences in a year is an extremely, large number Kendra Remirez (03:42): (laughs). Nancy Ozawa (03:43): A lot of miles in the airplanes as well. When you're out talking to different folks, is there certain industries or certain groups that are getting it better than others? I'm trying to understand that, that range of skills and who's really adopting it. 'Cause I'm sure some of our listeners are saying, "What is AI?" And some are like- Kendra Remirez (04:03): Yes. Nancy Ozawa (04:04): ... "Yeah. Where do I put the AI?" So- Kendra Remirez (04:05): Yes. Nancy Ozawa (04:07): ... help us understand of where we fit within that continuum. Kendra Remirez (04:09): Oh, I love this question because every audience I'm with, it's the bell curve, right? There's people like me that are just, gung-ho and advocating for it to the other side of the bell curve that are really fearful of it. And they're like, "Nope, no way. I've shut down access, you know, to all AI in our business." And I'm like, "Okay, well great. Now you just caused shadow IT. And now what's happening are people are using AI from a work perspective on their personal laptops and their personal phones, and now you're opening up a whole other situation that could bring cause to the business." And so, understanding it impacts, really every industry. And I would say the commonality of those that are adopting it and, leaning in are ones that are just being curious, right? They're just open-minded. (04:54): And so, when people are asking me like, you know, "What companies are gonna be impacted?" And my answer is always, "Do you have a sales team, marketing team, operations team, IT team, training team, HR team, right? Any of the customer service team, if you have those in your business, you are going to be impacted." So, it doesn't matter from a, industry standpoint. And it's literally as big as when the internet rolled out, right? that's the last massive shift in how we work happened. So AI is literally to that same equivalent. Nancy Ozawa (05:27): I, I think that's a perfect analogy. in the 1990s, the internet came about and people didn't know what it was. This is before Amazon really became anything more than- Kendra Remirez (05:36): Mm-hmm. Nancy Ozawa (05:37): ... a little tiny website with a few books, basically. Kendra Remirez (05:39): Yes. Nancy Ozawa (05:39): And look at where we are less than what, 30 years later- Kendra Remirez (05:43): Mm-hmm. Nancy Ozawa (05:43): ... and it's the next revolution. before we get into it, let's do a level set of AI- Kendra Remirez (05:47): Mm-hmm. Nancy Ozawa (05:48): How do you define it, and what types are out there? before we go too deep on this. Kendra Remirez (05:52): Perfect. And it is, it's a great foundation. 'Cause I was with someone recently, she's like, "Kendra, you keep talking about AI, but where do I go to experience it?" Right? And I'm like, "Oh yes, thank you for asking that question." And so, um, you know, we're already experiencing A- AI, we may not be going to places and knowing we're experiencing it. If you're using Netflix, if you're using Amazon, if you're using Google, right, you're already experiencing AI. You're not having to go to it to experience it. And so, what we're gonna anchor off in our conversation today will really be around ChatGPT. (06:25): So to experience that, you would go to chat.openai.com. And so when I talk about laying a foundation of what is AI, at its most simplest form, it's the ability for computer to think and learn, it's the ability for computer to think and learn, that at its simplest form is what AI is. And then we hear the term generative AI. And generative AI is just the ability for AI to generate text, image, audio data, right? It's generating something. And then the other terminology we hear quite frequently is large language model. So ChatGPT is a large language model. And let's play a little game here. So if I say the word peanut butter, what word comes next? Nancy Ozawa (07:09): Jam. Kendra Remirez (07:11): Jam. Yeah. All right. Or jelly, right? Nancy Ozawa (07:14): Jelly. Yeah. Kendra Remirez (07:15): Jam, jelly. So most people, you know, 99.99% will say jam or jelly. So in that instance, that is what a large language model is doing. It has tokenized the words that are showing up and it's looking for patterns. So it's not going like a Google search and grabbing information and giving that like paragraph out of Google straight to you. On the fly, ChatGPT is coming up with every single word. So just like what we did and played the little game of peanut butter and jelly, it is deciding like, "Okay, this word always usually comes up with this word and it is patternizing the information that it's providing." And so that's something I hear quite frequently because everyone, when they first experienced ChatGPT, they see the little, you know, window that looks kind of similar to a Google Search window and they try to use it like Google Search and it is not even remotely close to Google Search. Nancy Ozawa (08:09): Right. Kendra Remirez (08:10): So being able to have a conversation with it, and that's where it implies ChatGPT, we are going to chat with the AI and give it task to do. Nancy Ozawa (08:21): And I almost think about when you're chatting with it, you're, you're teaching a child. Um- Kendra Remirez (08:25): Yes. Nancy Ozawa (08:26): ... present back to me an outline of a press release that has these three key messages. And obviously a child wouldn't do it, but you've gotta give them a lot of context of what it is you want, what information do you know, what kind of outline you want? And it's amazing how it'll come back with different versions of a press release. And then as a human- Kendra Remirez (08:43): Yes. Nancy Ozawa (08:44): ... I'm like, "Nah, I don't like this paragraph. Let me change it." Kendra Remirez (08:46): Correct. Correct. And so again, let's lay in that foundation and what are we talking about when we say prompting? Nancy Ozawa (08:52): Yeah. Kendra Remirez (08:53): So what I want you to think about when we're prompting AI, what we're talking about is you went, let's say you went into a restaurant, the restaurant did not have any menus, and the wait staff came up to you and said, "What would you like to eat?" And I look at them and say, "Oh, I would just love a piece of meat and a vegetable." And the wait staff goes back, the cooks make something and they bring a steak and asparagus. And I'm like, "No, I actually wanted a hamburger and a corn on a cob." Right? Well you didn't state that, right? So in prompting we have to be really, really specific about the thing that we're asking the AI to do. (09:27): So, I love your analogy I always say when we're working with AI, pretend like it is a human. And pretend maybe you hired a really, really smart intern and you were giving them a task. And the more details we wanna tell that, intern that we hired, that, task-oriented person, you have to be really specific about what we want, what's the output? And then they can go to work. (09:48): But if I don't give them very much direction, they're gonna come back and not, really provide the thing I was looking for. So if we go in and say, "You know, provide me an email for a customer" and, and that's it. You know, it doesn't know who, it doesn't know what we were trying to convey. So the better we are about the input from a prompt standpoint, the better the output will be. Nancy Ozawa (10:09): Exactly. I think your analogy of the restaurant is a good one of trying to be as specific as possible, but I also realize that AI is not perfect- Kendra Remirez (10:17): Mm-hmm. Nancy Ozawa (10:18): Just because it gave you the answer doesn't mean it's actually correct all the time. You still need that human. Kendra Remirez (10:24): Yes. Nancy Ozawa (10:25): Um, so have you seen you know, you've asked your waiter if you wish, and instead of coming back with the corn dog, they come back with a piece of roast beef… Kendra Remirez (10:32): (laughs). Nancy Ozawa (10:33): ... and say it's, it's called a corn dog? Kendra Remirez (10:35): (laughs). Yes. Nancy Ozawa (10:35): Have you seen where AI is just, it's incorrect, it's gotta learn from us? Kendra Remirez (10:39): Mm-hmm. Yes, exactly. And so that is something that we have to understand. AI is not always correct. It does hallucinate. Hallucination is a technical term of inaccurate, right? So it's making things up. So, a really great, example that was used and was all over the press is where an attorney was using ChatGPT and preparing for a court case. And they were providing the court case, “Well, you know, this was the situation, you know, that happened." And the judge stopped everything and said, "No, actually those court cases don't exist.", and so in that situation, they should have fact-check the information. (11:16): So, it’s, sometimes not, accurate. It hallucinates, it is biased., when it was gobbling up all the data to learn and build, as any AI is having to learn from data, it, sucked in some of our bad behaviors that is floating around in data all over the world. And so, when I say biases, I asked ChatGPT to create an image of a female CEO of a large organization. And it took me two days of prompting and asking, 'cause it kept giving me a male, kept giving me a man, giving me a man, giving a man. I'm like, "No, I specified a female." And it finally gave it to me after two days of prompting it. (11:54): And so being mindful, of that, and then also being aware of the deep fakes. we've seen situations with Taylor Swift of it wasn't her, but it was, an AI mimicking her in very inappropriate, videos. So being really mindful of when we're looking at things, looking at images and video, could it be, a deep fake. you almost have a password or a safe code within your business or within your family. (12:22): 'Cause people are deep faking voices and calling and saying, "Hey, I have your daughter." and, talking about ransom scenarios and the daughter's fine. Or an executive that was traveling and someone spoofed his voice, called his assistant and said, "Hey, I'm traveling. I don't have my credit card on me. Can you read me the number?" Well, she didn't think twice about it, it was his voice, it was his phone number 'cause it had been spoofed. And then she provided the information. So, we're all having to be, very, very diligent. So there's duality in everything. There's duality in the good and bad in internet. There's duality in AI as well. So we just have to educate ourselves on the good parts of AI and also the not so good parts. Nancy Ozawa (13:04): Right. And that's probably got huge implications for, knowing your customer and spotting fraud within the banking world. Kendra Remirez (13:10): Yes. Nancy Ozawa (13:11): You know, I, I see AI being a very creative, person. So, if I don't have somebody to bounce ideas off of, AI can help and give me different ideas for images and press release ideas and things like that. But you're right. You need to fact check it Kendra Remirez (13:25): Yes. Nancy Ozawa (13:26): I actually heard at a conference that somebody had said that she put her blog into Chat and said, "Who published this? Who wrote it?" And Chat came back and said, "I did." Kendra Remirez (13:35): (laughs). Nancy Ozawa (13:36): She's like, "No, I did." (laughs). So, you know, there are times where it does hallucinate as you [inaudible 00:15:17]. Kendra Remirez (13:41): Yes. Nancy Ozawa (13:42): it's also true when an emerging technology happens, there's a lot of these little, uh, wonky things that come out until we understand what it's truly good for. Kendra Remirez (13:51): Mm-hmm. Definitely. Nancy Ozawa (13:52): We, did something similar to a picture. We were trying to, ask one of our, generative tools to create an image. And we asked for a very culturally diverse group of business professionals around a table. And it came back with all white people. We tried it again and it came back with all African American people. Kendra Remirez (14:12): Mm-hmm. Nancy Ozawa (14:13): It didn't understand the word culturally diverse. Kendra Remirez (14:14): Mm-hmm. Nancy Ozawa (14:16): So it was, it's, you know, it, it does still need to learn if you wish. Kendra Remirez (14:18): Yes. Yes. Kendra Remirez (14:21): And tell it, you know, tell it when it's correct and tell it when it's wrong. Right. Because it is learning. So just like you would have that, you know, human assistant with you, you know, when, when they do something wrong, you tell them. When they do something great, you praise them. Nancy Ozawa (14:33): Yeah. Kendra Remirez (14:34): And so it's always learning. So, make sure if it does produce something wrong, tell it that, it produced something wrong. And when it does well, you know, tell it that it's done well. Nancy Ozawa (14:42): Exactly. We, we gotta realize it's not like an Excel where it's always perfect. You do need to, to train it- Kendra Remirez (14:46): Mm-hmm. Nancy Ozawa (14:47): ... and work with it quite a bit. Kendra Remirez (14:49): Yes. Nancy Ozawa (14:50): Well, I've kind of dipped into a little bit of how marketing is using it. So I wanna circle back to something you said earlier about certain departments that are kind of the first places that are touching on it. You mentioned marketing, HR sales. Do you wanna say a little bit more of what you're seeing in some of those different departments? Kendra Remirez (15:07): Yeah, it's really fun to, you know, sitting down with teams and getting them to understand, you know, one, as an organization you need to have AI policy. And with that AI policy you also need to train your employees on that AI policy. So just like you do from a cybersecurity standpoint, you have the, you know, policy and then training on phishing attacks, making sure employees really understand the good and the bad of it. You know, it's not that an employee has ill will or they're not trying to do anything, they just don't know how these systems work. And then with each department in your organization coming up with use cases, what are some low hanging fruit internal AI opportunities? (15:44): So, I always consider internal AI opportunities first before going external, and showing any AI to a client, you know, externally or a customer. because we gotta, you know, navigate that world internally and adopted internally before we go external. And so within each of the departments looking at what's some low hanging fruit? Look for a repeatable task that feels very manual. You're like, "Gosh, you know, there's gotta be a better way to do this." Nancy Ozawa (16:08): Yeah. Kendra Remirez (16:09): And so most people are happy to, uncover those opportunities 'cause it's gonna give back more time in their day. So, in marketing and sales, any kind of content, any content you can imagine, emails, subject lines, proposals, sales enablement things, presentations, blogs, SEO, websites, landing pages, press releases, focus groups, you can do simulation with it. So, like I love showing sales teams that you can go in, there’s a GPT inside of ChatGPT that is called Negotiator. And you can role play a scenario with AI and, help you better your negotiation strategy. Or you can build your own little mini simulation of a focus group and say, "Hey, here are five personas, and I wanna be able to chat with each of them and get information back in a potential focus group." (17:03): So, I love brainstorming. I brainstorm with it every single day. And, I'll provide it something and then ask what else should I be considering? Or what am I not including that I should? and I'll just, ask it to think through something with me. And so to the point of it told me the other day, "I enjoy collaborating with you." I'm like, "Well, thank you. I enjoy collaborating with you." Like it was, it totally made my day (laughs). Nancy Ozawa (17:30): That is cute. And that was a, that's inside ChatGPT, I haven't seen that before. Negotiator. Okay. Kendra Remirez (17:35): Yes. Nancy Ozawa (17:36): That's awesome. Kendra Remirez (17:36): Yep. It's one of the GPTs, if you go to, on the, left panel, it'll say Explore GPTs and there are thousands of them. So the GPTs are like little mini apps, right. Like your app on your phone, and they have a specific task that they're going to do. so, there's all kinds of different ones that you can play with in different scenarios. And there's a search capability. So you can search all of the GPTs that are out there on the different tasks that you wanna play with. Nancy Ozawa (18:03): And I would imagine as a salesperson, maybe a junior salesperson or somebody who's got new territory, then you could say, my audience is, this kind of a banker, a lender- Kendra Remirez (18:14): Mm-hmm. Nancy Ozawa (18:14): ... who lends these kind of CRE type loans and give it some training and then- Kendra Remirez (18:19): Mm-hmm. Nancy Ozawa (18:19): ... say, "Okay, now let's just do a role play. Let me-" Kendra Remirez (18:21): Yes. Nancy Ozawa (18:21): "... talk to you. And then you respond as like the potential prospect." Kendra Remirez (18:26): Yes, exactly. Nancy Ozawa (18:28): That role play would be amazing for salespeople to, think through different scenarios of how to- Kendra Remirez (18:31): Mm-hmm. Nancy Ozawa (18:32): ... help the customer. And would it also evaluate them to say, "Hey, this is what you missed, this is how I interpreted that question"? Kendra Remirez (18:39): Yes. So, one of the things that I love telling everyone, when you are giving a prompt, you know, giving it some context, my little secret sauce is the very last sentence, say, "Ask me questions for clarity." And then what's gonna happen is it will prompt you. 'Cause it will say, "Okay, based off the information you gave, I need this, this, and this before I can actually provide the output you're looking for." So everyone and their brother is selling prompt packs, right? Like, "Buy my thousand prompts," you know, all this thing. And, and you don't have to do that. Right. It's a very, very intelligent tool. And so, you can just simply ask it, "ask me questions for clarity." You know, do that after your prompting. And then it'll say, "Well, I need these couple of things before I can actually give the output." Nancy Ozawa (19:24): Yeah. Kendra Remirez (19:25): so that will 10 X your, your prompt output. And, in that same vein, I really wanna make sure it's clear between the free version a ChatGPT and the paid version. So, these GPTs that I'm talking about, so like Negotiator and the other ones, those are in the paid version and that's $20 per month per, per user. The free version is very, very limited. It is only context and content. You can't do images, you can't do the GPTs. And even the output is, is way better. So I like to describe it as you can have a little tricycle with the free version or you can have a Ferrari with the paid version. and it more than pays for itself in regards to efficiencies in getting things done. But I just wanted to clarify that and where those things are 'cause that is, some of the things we talked about are not in the free, version. Nancy Ozawa (20:14): Okay, good. And when you mentioned free and, paid version, when you're in the free version, that is where you don't list your company name- Kendra Remirez (20:21): Mm-hmm. Mm-hmm. Nancy Ozawa (20:22): ... manager's name, your customer's name, any product name, anything that's directly can come back to you. Kendra Remirez (20:29): Mm-hmm. Nancy Ozawa (20:29): But in a paid version, I believe that's more private and kept within your instance. Is that kind of accurate as far as the logic goes? Kendra Remirez (20:36): Yes, that's correct. So, I always say, if it's free, you're the product. Right. Nancy Ozawa (20:40): Okay. Kendra Remirez (20:40): So you have to be really, really careful what you put in, in any AI tool. 'Cause you know, there's Claude and Perplexity and Gemini and all these other fantastic tools that I love using. So, any AI tools, you do not wanna provide any intellectual property, strategic property, names of individuals, you know, anything that's confidential. Right? So just being mindful of no matter a free version, paid version, anything, just the extra layer of security. So yes, in the free version it is training off of whatever you provided. And in the, paid version, you can actually toggle, there's a little toggle switch that you can go into your settings and turn off and say, "I don't want you to train on this, information." Nancy Ozawa (21:17): Right. And sometimes what we've done is take the company name out, we just put in the word company- Kendra Remirez (21:21): Yep. Yes. Nancy Ozawa (21:23): ... or product all in caps, so once it comes back, I can change it, but that way I'm not sharing anything that I don't want shared somewhere else. Kendra Remirez (21:31): Exactly. Nancy Ozawa (21:31): Now, you mentioned other departments, and I'm curious 'cause a lot of us have operations, the back- Kendra Remirez (21:36): Mm-hmm. Nancy Ozawa (21:37): ... operations, and that's an area where a lot of people are looking for ways to streamline. They can't get enough workers in that area. Kendra Remirez (21:43): Yes. Nancy Ozawa (21:44): Are there other kinds of AI tools that you've seen that are really kind of targeted to the operations group? Kendra Remirez (21:49): Yeah, so, one of our clients in operations just using two tools, saved them 1400 hours per year. Nancy Ozawa (21:56): Wow. Kendra Remirez (21:57): And so one was just simply transcribing their meetings because of being in operations and project management, every single meeting they had, someone had to sit there and write up the notes and write up the recap and the next steps and who owns what and email it out. And so, having, Zoom, Teams, Otter, there's some fantastic transcription tools out there that you invite to your meetings. So they were averaging 20 meetings a week. So taking all of that with their team. And then the other tool is Tango. So it's Tango.us. And so, anytime you have to produce some kind of how to document of something that sets on your computer. So, either, from a, a Chrome extension or, or just on top of your desktop where you have to do step by step, you know, click here, click the dropdown, click the gear, click this word. Right. (22:44): normally what we have to do is use a snipping tool or a snag-it tool, grab it, write what we did, grab the next step, write what we did. And that's very time-consuming. And so by using Tango, it literally just watches what you do. You don't even have to do it slow. It grabs a photo, an image of what you did, and then describes what you did and put it into a PDF. which would be fantastic to use in training and development or learning management systems. And you can go in and edit anything. Maybe it grabbed the wrong thing or, it didn't quite get the full description of what you wanted to describe. You can totally, edit that, as well. So just those two tools, was 1400 hours. Nancy Ozawa (23:25): That's amazing. And I mean, all of us need notes and next steps from all of our meetings, even outside of operations. So essentially what you're using is if you're in Zoom, you're using Zoom companion, AI companion or something else like that. Kendra Remirez (23:39): Yes, exactly. Exactly. Nancy Ozawa (23:41): I'm gonna have to try the Tango one. I was just last night train to do some how-tos and, uh, that would've been very useful as well. Kendra Remirez (23:47): Yes. It's one of my favorites and we've been using that a long time. Nancy Ozawa (23:50): Okay. Now you touched on AI policy, so let's kind of head there because I'm sure the folks that are listening right now, our risk folks are like, "Okay, (laughs), you've really got my attention now." Kendra Remirez (24:02): Mm-hmm. Nancy Ozawa (24:03): And people, as you said, might be doing shadow AI where they're- Kendra Remirez (24:05): Mm-hmm. Nancy Ozawa (23:05): ... doing this on personal devices, which is not exactly the direction we wanna go. Kendra Remirez (23:10): Right. Nancy Ozawa (24:11): If you want an AI policy, and I know a number of people have been, kind of talking about this lately. What are certain elements that they should be thinking of to put into an AI policy and do you have any tips on, you know, how to get that embedded across the company to make sure that they are protected or at least thinking about it, especially if they don't quite know how to put their arms around AI? How did they come up with a policy of something they're not quite sure how to put their arms around? Kendra Remirez (24:36): Yeah. So use ChatGPT to write the policy, right? (laughs). And so it's not your final, right. You want your, your legal counsel to do a final review. And, and I've worked with, many law firms and they're like, "We're happy when someone comes to like..." because they didn't wanna start from a blank piece of paper anyway. And plus it, you know, helps the billing time, right? So- Nancy Ozawa (24:55): Yeah. Kendra Remirez (24:56): ... starting there. But it's really, a really acceptable use, right? So, making sure that, you know, confidentiality thing, information's redacted, making sure that employees, you know, understand what tools are approved. "Here are a set of tools that you can use. Here are some tools, you know, you can't use." different scenarios. There might be, if there's any compliance involved, they have to, export that conversation and show proof where they got the information. So, it really depends on the different layers and different departments of what they, need. And I have a couple of templates that, I've built as well. (25:30): But, start in, ChatGPT to get a nice foundation and then have a human that, you know, has that legal counsel, background to go in and, and elevate that. But then it's making sure that you have the policy and then educating on the policy. So, making sure that employees understand all the things that we're talking about, today about, the good uses and how you can use it. And then also the pitfalls, right? Things that, for bias or hallucination, or deep fakes, making sure even how they can, just like with phishing attacks, you know, how can they identify a fake photo or a fake conversation? (26:08): because that's something else, especially being in banking, the red alert has to be really, really high and making sure that they're identifying scenarios or situations that may not even be, be real. Nancy Ozawa (26:19): Yeah, true. And you mentioned AI policy is for the employees, but something you tipped off, uh, a little bit earlier is it's not just the employees using it, but if they put it out on their website and it's a chatbot and it's talking to customers- Kendra Remirez (26:35): Hmm. Nancy Ozawa (26:36): ... there is a potential risk that- Kendra Remirez (26:37): Yes. Nancy Ozawa (26:37): ... that chatbot might do hallucination. And- Kendra Remirez (26:40): Yes. Nancy Ozawa (26:40): ... there's a risk level there too. Kendra Remirez (26:42): Exactly. And that happened with Air Canada. So, Air Canada had an AI chatbot on their website, and it was that external AI an individual wanted a bereavement refund on a flight. And, the policy, came back on the website, "Yes, we'll refund that." And, it hallucinated in that moment, that's not actually true. unfortunately, they couldn't figure out, you know, the, the scenario and take care of it themselves and just refund 'em and call it a day. Air Canada fought it and they're like, "No, no, no, it's not our fault. it was the AI's fault." Well guess what? You are the one that implemented it and they lost the, the lawsuit. (27:20): So now you have a whole bunch of not good press on something that could have been just handled, have they, said, "Hey, it was on our website, it was our scenario." And so just being mindful of that, if it does, come up, that sets the precedent now of- Nancy Ozawa (27:34): Mm-hmm. Kendra Remirez (27:34): ... you are on the hook for that because the website is yours... So, they didn't handle that very well, Nancy Ozawa (27:40): Yeah. Kendra Remirez (27:42): ... one making sure that you understand, the tools that you're implementing and because there are good chatbot tools out there that don't hallucinate, but still you've gotta test and test and test and make sure before you publicly start sharing anything with your, customers and clients. Nancy Ozawa (27:58): So you mentioned a few that don't hallucinate. What are a few of your favorites that don't hallucinate? Maybe we won't talk- Kendra Remirez (28:04): Yeah. Nancy Ozawa (28:04): ... about the ones that do hallucinate. Kendra Remirez (28:06): Right (laughs). Well, you know, it's back to, really being able, testing it. So, Microsoft has their whole enterprise suite, it doesn't, you know, it doesn't share any of your information. It's, it's what I call moat, right? It's got a moat around the information. And so we know that that is a enterprise level AI. So, there's over 52,000 AI tools that exist today. And a lot of them are these, like they're wrappers, right? Meaning they took ChatGPT and layered some stuff on it, and then now they've called it a tool that they have. (28:37): So, making sure you're vetting those vendors and making sure they are, real enterprise level, have SOC compliance, really dig in and understand, where these tools are, coming out of. So, like the, the Microsofts of the world, using Copilot, 'cause Copilot is their specific AI tool in the enterprise. And Microsoft invested $10 billion into OpenAI. So really and truly under the hood it's ChatGPT inside of, Microsoft Copilot. And so making sure to use, platforms, like that, that are more what I consider enterprise level, tools (29:12): And then there's other tools, customgpt.ai is another, fantastic tool that, you can build off, documentation or website and it's, being able to train off of that without bringing in the external information. From an AI standpoint, it's training off your information, so the information it's gonna produce is going to be more correct. Nancy Ozawa (29:35): So, if, I was trying to vet my AI tool to make sure it's appropriate for my institution, what I'm hearing is kind of go for the larger leaders who've invested- Kendra Remirez (29:43): Mm-hmm. Nancy Ozawa (29:43): ... a lot into it. Kendra Remirez (29:44): Mm-hmm. Nancy Ozawa (29:45): But I would suspect that there's a lot that are showing up that are startups and they've gotta start- Kendra Remirez (29:49): Yes. Nancy Ozawa (29:49): ... somewhere and they're a little bit more focused on this. Is there any advice you have of weeding out the good startups that are a little bit more safer than others? Or is it- Kendra Remirez (29:58): Mm-hmm. Nancy Ozawa (29:58): ... just give them time to develop and be a bit larger company before you take on their tools? Kendra Remirez (30:03): Yeah, and especially in banking, Nancy Ozawa (30:05): Yeah. Kendra Remirez (30:05): I, I would be uber, uber safe and uber crazy, about the level of detail and vetting. And so, I always say, look up the company, look at how many employees, look at when they started. Go look at some of the leaders, right? because when, when you start looking and digging a little bit, pulling that onion layer back and you can't find leaders and you can't find when they started, that's a big no. Right? or they, you know, started six months ago, that's a big no, right? You, you do not wanna be their Guinea pig. Nancy Ozawa (30:34): Right. Kendra Remirez (30:34): and they, there's, time and place for, those tools and they'll, they'll grow. But specifically, in the banking, you have to be really, really lockstep from a compliance, you know, standpoint and making sure you're really evaluating more these enterprise tools that have been around for a while and have made big investments. So just being really, careful of these, like I said, I call 'em, you know, wrappers, right? They're, they're taking some enterprise things and they're just layering a little bit on top of it. (30:59): So also look at just your current vendors, look at your current vendors or already approved vendors, or already people that you've worked with. A lot of them will have naturally built in AI tools. So, I encourage you to go back to your current technology vendors and ask them like, "Hey, come update us on what's new in this particular tool, from a AI perspective." Because you already know you've gotta a relationship with them, you already know them well. Nancy Ozawa (31:24): That's an excellent idea because we're experimenting in the marketing team around AI a lot and we're finding that a plethora of the current tools that we have are being powered by AI. I mean, Photoshop is bringing AI into it. Kendra Remirez (31:26): Yes. Nancy Ozawa (31:36): So many other... You mentioned Zoom. Zoom's bringing- Kendra Remirez (31:38): Yes. Nancy Ozawa (31:38): ... AI in. So, a lot of our common tools already have AI in it, and it's already passed our due diligence as well. Kendra Remirez (31:45): Yes. Yeah. And just one caveat I would say is don't lead with the tool first, lead with what's the goal and objective? What's the thing that you want, to accomplish? So we really need to start from that strategy first of what are we trying to achieve? What are we trying to accomplish? And then determine what tool would make sense. (32:05): 'Cause we're leading with the tool that's gonna set us down the wrong path and we have to make sure that it's leading back to our overarching strategy. Nancy Ozawa (32:11): Agreed. Agreed. It's, when we got started, we kind of did our use cases. What are the big things that are taking us a lot of time- Kendra Remirez (32:18): Yes. Nancy Ozawa (32:18): ... or difficult to do, let's look at that as the low hanging fruit and figure out a tool that's really gonna help on that one. Kendra Remirez (32:24): Absolutely. Nancy Ozawa (32:26): So, we've definitely been talking about banking. So let me ask you, you've talked to a lot of different industries. Is there anything specific about banking other than we need to be more cautious, which I think inherently we are. But is there any other tips or things that you're seeing that are different about this industry's adoption of AI besides being our normal, cautious conservative selves? Kendra Remirez (32:48): Yeah, you know, just looking at the internally operational things and having an AI first mindset of when you're running through a scenario just stopping and saying, "Could AI help me with this task? And if it could, what would it look like?" And so, what I encourage organizations and, even, is starting like a, an AI committee, an AI task force where you are really intentional about talking about AI in your team meetings, in your leadership meetings, and start building a AI super user in your organization, and someone that's gonna help lead the charge. this is something that I started when we do it, you know, doing digital transformations 19 years ago with businesses. So it's the same methodology. (33:30): think through how you went to market of figuring out your digital transformation and apply those same principles in doing an AI transformation and putting, a task force together and are departments thinking about it? And then sharing back success stories or learning situations where there was a pitfall situation, making sure to, full circle that information. That's something that will really help move the needle. And I think it's just making sure that we're embracing and we're being curious and just leading with that curious mindset. I know we're, very risk averse as we should, and I want you to be that way, as an end user, I need you to be that way. Nancy Ozawa (34:07): (laughs). Exactly. Kendra Remirez (34:08): but at the same time, you know, we don't want, you know, the organization to be left behind because they were unwilling to even consider it, I met with the CEO, he is like, he was very proud of themselves. He is like, "We've shut down access to everything." I'm like, "Great, you've caused a bigger problem now (laughs)." So that's not the answer. making sure that you are, you know, leading the charge and rallying teams around and, the impact of the knowledge worker. (34:34): you know, women tend to play more of those knowledge worker roles, and AI is going to impact eight in 10 women versus six in 10 men, from the AI impact of the workforce. So being mindful of that, it's our job as leaders to upskill and re-skill and prepare our teams for now and into the future. Nancy Ozawa (34:55): Right. Well, uh, you have wrapped this up with a bow, I think you've given the listeners and myself, a few steps if you're not getting started with AI already, I think what I'm hearing you say is start experimenting, start talking about what your big challenges are, which are things that take a lot of time, that are very repetitive and start experimenting with those. But experiment... only internally first. Kendra Remirez (35:18): Yes. Nancy Ozawa (35:18): And that's kind of one of the key lessons I'm hearing you say. Kendra Remirez (35:19): Definitely. Nancy Ozawa (35:19): Is there any other, key points that you wanna summarize as we box this up? Kendra Remirez (35:23): Oh, thank you. I, I just appreciate this opportunity and, and those are listening, just let us know, right. your takeaways or what you're learning, because I believe in collective wisdom, we're all learning together. Nancy Ozawa (35:35): Yeah. Well, I, think this was a, a very useful and informative conversation and I loved all the tips and tools, you gave me a few new tools to go try out too. So thank you so much, Kendra. if any of our listeners wanna get in touch with you either to talk more about AI tools or strategies to adopt and experiment with AI, how do they reach out to you? Kendra Remirez (35:56): Oh, thank you so much. Yeah, it's kendraramirez.com and I'm Kendra Ramirez on LinkedIn as well. And happy to connect. Nancy Ozawa (36:04): Perfect. Well, thank you so much. Kendra Remirez (36:06): Thank you for having me. Nancy Ozawa (36:07): we’ll make sure to share that with our listeners, your contact details, so they can reach out if they wish. And thank you so much listeners, for tuning in again for another provocative discussion on banking topics. If you haven’t already make sure to check out our other episodes and also subscribe that way you’re the first to know when a new episode has dropped. We are always looking for feedback on what to discuss next, so if you’ve got something in mind or maybe you want to be a guest on this podcast, please let us know by emailing us at bankingoutloud@pcbb.com. Thanks again. See ya next time.