Episode 71: Ben Udell

Introduction: I've had the privilege of being in and around banking for more than 50 years. Lots of changes during that time. We've gone from ledgers to laptops, typewriters to technology. One thing, however, remains the same. Banking is a people business. And I'll be talking with those people that make banking great here on, Jack Rants with Modern Bankers.

Welcome to Jack Rants with Modern Bankers. It's brought to you every week by RelPro and Vertical IQ. After six decades in banking, it's time for me to give back. And I hope this program does just that. Every week I feature top voices in financial services, bankers, consultants, best selling authors, and more. And there's a banker here today that you are going to want to listen to. 

Now, the goal here in this program is really simple. It's to provide insights, success practices, and to bring new ideas to the table that you can use to maximize your results in 2024. Innovation, a concept every community based financial institution is striving for. Generative AI, a tool that makes bank and credit union CEO's very nervous.  My guest today blends the innovative nature of ChatGPT with an approach that helps his CEO sleep better every single night. It's Ben Udell, senior vice president for digital innovation at Lake Ridge Bank.

Ben earned a bachelor's degree from UW Milwaukee and an MBA from UMass. After six years in sales and leadership at Wells Fargo, Ben joined Lake Ridge bank and in 2023 launched his career in the digital space. At that same organization, Ben speaks at numerous banking conferences, and he presents a fabulous breakout session at the 2024 Financial Brand Forum, one of the greatest marketing programs on the planet. Want some practical help in chatGPT? Check out this interview with Ben Udell on Jack Rants with Modern Bankers. Here we go!

So, as I mentioned in the introduction, it is incredibly refreshing to see a community banker really attack and get on top of some of our topics today. Ben Udell is going to talk about that. Ben, great to have you with us. Thanks for joining us.

Ben Udell: Yeah, thanks, Jack. I'm super excited to be here.

Jack Hubbard: Let's talk about, let's start out with the bank, because I think we need a context. Too often people look at AI and they say, well, that's a big bank. Talk about the bank, talk about yourself, and how this all started with AI, because for community bank, it's a little unique.

Ben Udell: Yeah, the short version is that I was just very naturally interested in AI, both with my role and personally, and it just started to build. And I think the difference when you talk about the difference of a community bank versus these larger banks is that I was using ChatGPT for the daily stuff, the writing, the creativity, the upskilling on certain topics. And once you start doing that and appreciate how AI can impact your role, it just comes natural to want to share that with the person next to you, with your team, with other individuals at the bank when you're solving problems. I think that's what we've started to do a little bit differently is I'm not getting caught up in the application of AI and ChatGPT and all these other functions for the super big technology lifts. I'm looking at it on how do I work smarter, better, how do I produce better work, because I have this really great tool, and that's something that can be applied across an entire organization, and that's easy and it's cheap and it's effective and you get immediate wins out of it trey.

Jack Hubbard: Well, you mentioned your organization. Talk about the bank a little bit. Lake Ridge bank, been around for a long time, different names. but, talk about the bank a little bit up there in Wisconsin.

Ben Udell: Yeah, we're based out of Madison, Wisconsin. We're $3 billion, approximately. We've got 20 plus branches. And I would say we're this classic midwestern community bank. And that we have a really strong, commercial arm and a mortgage arm, and a retail arm and an investment and trust. And we're really embedded in the community right now, we're either the fifth or 6th largest, based bank in Wisconsin, and we're really proud of that. but I do think as I talk to my peers, we're one of the few that has an active AI plan in this productivity 

Jack Hubbard: Great. And you are a leader in so many ways because you've decided, and the bank has helped you with this, to get out and do a little speaking. And you're going to talk, at one of my favorite conferences. I had a chance to speak at the very first financial brand program, and it's just grown exponentially. Jeffrey Pilcher and his team do a marvelous job. You're speaking there in May, and I believe you're doing a standalone seminar. Talk about the program you're doing there.

Ben Udell: Yeah, I'll be leading a breakout session, and my breakout session is all about transforming your marketing content, and that's really using ChatGPT, right. Which means it's a very cost effective approach to working much more productively, more creatively. So what I'm going to do is really work through it in three steps. And the whole point of this is that somebody can take this information and later that afternoon get to work and create something meaningful. And so we're going to start with basically taking your current marketing material, your writing, your content, running that through chatGPT, and having ChatGPT create a brand voice for you. And essentially you need that voice to then feedback into GPT so it can start to write in your, your brand. 

And then what we're going to do is we're going to basically drop that into one of chatGPT's and for those that don't know, within the paid $20 version of ChatGPT, you can create these little mini programs, and they're really easy to create because it's natural language coding. So you're going to drop in your brand voice and then you're going to give it some additional direction and support. And you're basically going to start tweaking that brand voice, because what comes out of ChatGPT is not going to be the final version, right? It never is. So you'll tweak that. And the benefit of that is now you have this consistent approach where they can, your team can use this GPT and then we'll actually demonstrate. 

How do you write marketing content? I need a blog. I need an email to support the blog. I need a copy for a direct mail piece. I need to create internal guides to support my bankers that can support the marketing that I have. In fact, I'm going to do another conference and we're going to talk about how you can take that information, put it into GPT and ask it for bias, for instance, because that's a big concern in the banking space right now on this topic. And so the beauty of all of this within ChatGPT is that how do you, as a marketer or anybody that needs to write and create content, how do you accelerate the volume and pace of connecting with clients? And what I tell people when they use this is that if you are going to take what you get out of GPT and assume it's perfect and right and factual, you are going to be wrong and you're going to have some challenges. So really what you're doing is the way that I described is that a, b, C and D is your professional skill and knowledge to determine what you want to create. What do you need to create? What does that look and feel like, e through s is really the writing process, and that's what takes the longest amount of time. 

Now, the t through z is that's where you're going to go and you're going to edit, you're going to review for bias, you're going to fact check, you're going to put in your brand even more. If people can do that, they can either, a, create more content, or b, they can be more hyper relevant to connect with clients better. And that's just a more effective way of using our team and improving that client experience. So that's a long winded version of what we're going to talk about at that session. But this is something that if you're a $20 million bank, if any of those exist anymore, or if you're a $20 billion bank, you can apply and get started on in an afternoon. And that's really exciting, right? That's really exciting for this client experience of being more relevant with clients and prospects.

Jack Hubbard: We were talking before we started, the recording about Guy Kawasaki and my program that I did back on March 6th. Go back and listen to it. It's on YouTube and in our public library. And we were talking about canva and Guy Kawasaki's whole approach and canvas whole approach is democratizing graphic design, allowing you as a community bank that doesn't have a lot of resources around to be able to do this yourself.

And this is a lot with ChatGPT. So it's November, I want to say, of 2021, and I might get the year wrong. It's around Thanksgiving, and ChatGPT hits the market, and all of a sudden people are just all up in arms, and some people are excited, and some people are fearful. what was the trigger for you? When did you get started and say, oh, my goodness, this is a great tool?

Ben Udell: Yeah. For me, it started kind of out of the work environment. And I remember being at Christmas parties, and we were playing around with ChatGPT, and we were basically creating new Christmas songs with different styles of lyrics. And, you know, the more you get into it, the crazier you start to make it, right. You say, well, what if I could do this, and what if I could do that? And it was just this fascinating, really experiment. And what I think people don't appreciate with ChatGPT is it can basically take, I want to do Christmas songs and the style of Al Green with a flavor of Miley Cyrus focused on Wisconsin. Right? And the reality is it starts producing content for you, and I'm not going to say, this is amazing content that all of a sudden I'm going to have a hit record with. 

But you see this ability to do it near instantaneously, and then when you start to say, geez, I wonder, how can I start to apply that at work? I have to do home equity promotions and I gotta do it differently from the baby boomers to brand new first time homeowners that are Gen Z. And I need to sell a high yield savings account. And how I sell high yield savings to people that have just retired, to people that are saving for a wedding like that takes a lot of time to write the content. And here I was just writing weird Christmas signs. So now you start applying that and all of a sudden it's off to the races with, again, working faster, more productively, more creatively smarter. One of the things that we talk through, and I think this is one of these takeaways for people, is I talked to my team and helped my team work with it. They'll get output and they'll say it's good, but I really would like it to do no more bullet points, talk in a friendlier voice. It should be more elaborate. How do I get it to do that? And my response is, just ask it to do it. It's crazy. 

And then it starts kicking all this stuff out. And again, I'd stress, you got to read everything and you're going to edit it. But that's still way faster than sitting down and putting pen to paper and really working it out. And that's, you know, I think that's the thing I would, I would tell a lot of people, Jack, if I get that your work environment may not be ready to turn this on, get a personal account and start playing around with it. Whether it's for personal or whether you try to write some work content or something else, just, just start testing boundaries with it. And it's pretty amazing. It really is.

Jack Hubbard: It is great and you mentioned if the bank may be a little reluctant. So I'm Paul Hoffman and I'm sitting in my office and I'm going, I'm a little nervous about this. And a lot of bank presidents are, Paul, being the president of your organization. But how did you get the bank? Because you're pretty open about using a lot of AI. How did you get risk and it, and Paul and the board and everybody to say grace over this so that you could actually get out and do all the great things you're doing.

Ben Udell: Yeah, that's a great question. And probably the one I lead a lot of my presentations with. So let's take a step back and think about the key things that people get worried about. Privacy is going to be the first big one, right? And if I'm going to be maybe a touch flippant about this, why was the, why would the bank entrust me with all of this client data that I can review, access, print, copy, do all kinds of stuff with? And then when it comes to ChatGPT, to say, you know, Ben, I don't trust you with that data, right? 

The reality is, we have our IT policy, and the IT policy is very explicit about not sharing private data, not uploading private data, only using private data with sources that it and RISC has fully assessed. So the reality is we don't really have to create necessarily a new IT policy on AI. We need to embed AI just to be more explicit. And so, as we've used AI for the most part up to this point, it's never been with private data. And that's okay, because we're going to get a lot of wins out of that. Now we are starting to test pilot, co pilot Microsoft's version of AI, which is effectively in the cloud and secure. From a privacy standpoint, that's a whole other topic for down the road. But you can get so much done with ChatGPT without using private info.

I think the other thing that comes up is bias. And obviously, that's so important to us in the banking space. The reality that I think is a practical piece is when you start writing with AI, AI is really pulling from the Internet as a whole. And when you think about all the content that exists out there in financial products, it's coming through financial institutions. So by default, a lot of that, if not nearly all of it, is really good with removing bias. But my key point here is when I create content, either pre or now, with AI, and it's going out for public consumption, I'm never just sending that out. Right. I have a marketing team that reviews it. I have a compliance team that reviews it. 

So this kind of hits that same piece with privacy, and that is, I'm not doing anything different. I am using my internal resources and tools and processes and policies to have my material reviewed. If I had bias pre AI, they're going to help catch it for me. And if I have bias with AI, they can help catch it for me. Right. I think the other thing that starts to come out, is also the factual side of the world. So from a factual standpoint, what's funny is people think about all these issues where AI hallucinates and it makes stuff up. Well, the funny thing is that if you ask for, I want to know the largest crops grown in Wisconsin for a crop agricultural blog I'm doing, I can ask AI, cite your sources. And if AI takes me off to sources that are credible and I can link to, and I can credit appropriately, that's a win, and I'm done. If AI takes me off to sources that seem to be made up, and I really like that type of content, I can go into AI, or if I need to, I can go into Google and I can say, find me the crop yields for 2021 for Wisconsin, and now I can find good content. And so I'm fact checking myself on that. The last thing I would share is, from a hallucination standpoint, if I ask for the thousand best restaurants in Madison, it's either going to stop giving me restaurants or it's going to want to fulfill that and start hallucinating. So I think what we find from a hallucination standpoint is the more specific you are in what you want to accomplish, what that looks like. Maybe where you get your sources from, you lose this level of hallucination and it starts to write in a theme that you want. 

So I share all of that because I think as you have conversations internally with individuals on how to get started with it, a lot of these pieces are just simply either me being a professional, and I would always fact check and review and run through compliance and be smart about it. That really doesn't change. It's just now, different. And I think as people start understanding and appreciating that, that's where you start to say, okay, now let's start to pilot this, and how do we start working with this, and how do we start getting others involved? And it's kind of off to the races at that point in time.

Jack Hubbard: Well, you just, if I'm in the audience now, if I'm a bank marketing professional, I'm thinking, okay, you got to ask them this, and you got to ask them this. One thing that's come up in my mind is, so you in marketing, and you came up through sales and such, and you're in digital marketing, et cetera. But what about if I'm a commercial lender or I'm in treasury management, or I'm a branch manager at your bank? Can I access ChatGPT now as well?

Ben Udell: Right now, we have about 10% of our associates that have access to ChatGPT and what that looks like for us is I've run them all through a two hour training, and the training covers certainly the topic we just talked about, right, with bias and privacy, et cetera, to make sure we're good. And then we start working through the whole prompt piece. And so we have a cross section of the bank that continues to kind of work through this, develop, pull out some best practices that we're seeing. But I think what's interesting then is when you start applying this across other areas of the bank, what ChatGPT and a lot of these AI programs are great with is they're great with communication, they're great with writing, they're great with providing different perspectives, point of view. 

So if I'm a manager, say, in a branch, and I have to have a difficult conversation with somebody, ChatGPT becomes this really great private way to say, I have to talk to somebody about their work performance. This is their personality style, this is what they're struggling with. Help me understand how I can have a good conversation with this person. What could be pitfalls? What are questions I could ask? What's future opportunities for accountability? And so now, as we move away from this marketing realm, we get caught up in a little bit with ChatGPT. It's a really great tool to be able to have really, the sounding board, this tutor, this person who can upskill you in addition to your normal resources. That's really going to help you with that. Imagine going out to a business client who is, you know, you have these business clients. 

Some of them are extremely detailed. Others are, you know, thinking about the pie in the sky. Others need to know the true roi, others need to dream a little bit. And as you start matching up personalities, you can run that through AI. And none of this is private data we're putting in. Right? That's the other thing, right? This is all anonymous, helping me work through this stuff. I have a difficult client I need to talk to. I have a, sale that I need to close, and all of a sudden you have this opportunity to engage. One of the things I tell people, and I think this is what happens with AI, is they put in one prompt and they don't get a great result. They put in a couple prompts, and it doesn't really get them anywhere. The real magic happens five or ten prompts deep. Or when you start stacking a bunch of prompts together, there's prompts that literally might be a word document page long that you can use repeatedly. And that's really where this magic happens, because you can be repeatable and consistent and apply it not just to marketing and content. And I'm happy to give you examples on that, but I'm rambling a little bit here, but that's how powerful this can be. 

It's amazing when you start using it for those reasons, I'll tell you, Jack, and not to spin out too much here, but if individuals have not yet spent time with the ChatGPT version where you can talk to it, it's amazing. And if you think about how I think of my grandmother who's 88 years old and she's an assisted living and she's not around people, imagine taking ChatGPT and in a very human voice, allowing her to have a conversation with AI, to talk about anything. If she wants to revisit the 1960s and have a conversation in a very empathetic tone, like that's really wild to think about, but it works really well. I've had conversations just driving between meetings with AI on trying to learn a new topic, and it's wild how human like it is and how impactful it is. So that's a long winded answer to your question, Jack, but it kind of highlights thinking creatively about all these ways to apply it in our professional and personal life in a very meaningful way.

Jack Hubbard: And I got to believe that the conversation piece is more advanced and it costs you money. So at what point do you suggest that a marketing professional go from the free version to the $20 a month, which sounds like a no brainer because you get, I think you get Dolly, which is the graphics piece. When does somebody tip over and say, okay, I'm going to spend some money here?

Ben Udell: Yeah, I think for $20 a month it should be a no brainer for anybody that is certainly in marketing. Anybody that works with systems like the Microsoft systems, I pretty much learned how to build Microsoft bi, graphs and charts and data with ChatGPT. anything you need to tutor on for $20 a month, it's a steal. And here's a little practical example: if you think about how long it might take you to write an email to send to somebody that you want to persuade them on, it's a difficult conversation. You need to show support, you need to take all these creative thoughts in your head. You're going to save $20 worth of productivity on day one just by using ChatGPT. And, a personal example of how I like to use it is trying to be creative throughout the day. 

My creativity will hit me in different bursts. I will use the dictation function in Microsoft Word, and as I have ideas, I just basically dictate that out into words and I can come back and forth to it. I'll take those ideas in Word at the end of the day, drop them in the ChatGPT and I'll basically say, this is a dictation of XYZ topic. I would like you to rewrite this into an executive summary, into bullet points, into a full document, into a blog, into a communication. And so, you know, dictation write is literally just a blurb of words with no punctuation, no anything else, and it'll do it for me. And if I don't like that first version, I'll let the magic happen by iterating on it to say, no, I need it more of a summary. I need elaborate details. And I'll give you a practical example of this. There's probably a lot of listeners right now, at a lot of community banks, where they have individuals in their departments who have all these policies and procedures in their head. It's this ingrained collective knowledge they've had through being in the job for 510, 20 years. And at the end of the day, we don't hire a lot of people for their writing skills and for their policy procedure skills, right? Take that person, have them dictate it step by step into word, and then once they're done with that, move it over into ChatGPT and say, this is a procedure that I'm writing. The audience will be bankers. I need you to write a detailed process on this with steps to help somebody who doesn't understand this process. And then in seconds, all of a sudden you have this process. 

And by the way, that person technically doesn't need ChatGPT right? Everybody listening has Microsoft Word with probably the dictation function, you're done. And by the way, I'm not suggesting like, you're done, right? But that writing process, when I go back to like e through s, imagine somebody in your deposit ops area where they're just like getting interrupted with questions and they're like having to type it out. And how many people are not, again, the best typers? And then you have a format for your bank that you want to use. The reality is let chat CPT do it. Oh, by the way, when you have like this is what we're doing now. When you have a process and a template for that, you can upload the template and say, take this process, rewrite it with this template and it starts to help you become more consistent with the material that you have that's going out. And then oh by the way, write me faqs, write me an email that's going to go out to the bankers to educate them on this. write me a quiz to test for knowledge. Write me a training guide for the managers to test for knowledge and have a discussion. And you can either do that that way or we're just starting to create a lot of templates for all that through GPT. So okay, Jack, slow me down when you need me to, but that's it.

Jack Hubbard: Want to ask you about prompts, because you mentioned them a couple of times and I want you to kind of talk through exactly what they are and some, some best practices. But let me ask you a question. So we're doing this recording. What if I, is it possible that I could take the link or the recording and put it into ChatGPT and say okay, watch this recording and write a blog based on it? Are ah, we at that point yet?

Ben Udell: Good ah, question. Not quite, but it's not hard to make that leap. So let me give you two examples. One which we've done and then what that might mean for this type of podcast. So we basically are rolling out an upgraded, expanded mobile and online banking tool through Fiserv. And we basically downloaded 5 hours of video from Fiserv. Because that's their overview. That's their what? they're sharing their rollout, right? It's everything in 5 hours. We downloaded that. And most of us here probably have Microsoft, which probably means you have a stream. Stream will basically convert videos into transcripts. 

So whether I'm having an internal meeting through teams, I record or I can download a video now I have this giant transcript. So then I could take that transcript and I can drop that into ChatGPT and I can basically write a manual for me. And so instead of somebody writing through five or listening through 5 hours and figuring all that out, we let ChatGPT do it. Now admittedly that's a little bit more advanced because part of the deal is you've got I don't know how many tens of thousands of words, but the point of it is when I start manipulating that into a manual, a guide, and spinning that off into educational materials, and maybe I want to actually roll that into my marketing stuff, it's going to be faster. 

And it was funny when we did this with that info I was talking about, we did 15 minutes of working and manipulating it in AI and then we got to this point of, oh, you know what, that output is not great. What I really need to do is create a template for what I want it to do. And then I basically had it read all the data to give me an outline of all the topics. And then I'm going to put that template in to write it. And so it's funny, we're like, oh geez, I just wasted 15 minutes of my time. And then we chuckled to ourselves, like, wow, that was faster than watching 5 hours of video. So we're okay. So here's what you probably do. You probably have access to your chat transcripts, right? Through zoom and teams, you can take that transcript out and then what you could easily do is say, I'd like to write a LinkedIn post. 

To promote this, I would like to pull out key topics. I would like to do x, y and z, and it'll start to do that for you. A hot tip that I found is you basically would tell it, this is a transcript of our podcast. I need to rewrite it in a way to communicate it by blog or for my website or whatever. Sometimes what happens, and this is probably the real hot tip, is it helps sometimes when you give AI an example. So what we've done with transcripts to keep it from getting confused, is we'll give it a snippet of the transcript and we'll say, this is the chat transcript. 

Each speaker has a timestamp and their name after it. When you write meeting notes, you will note that that person was in the meeting. If they reference somebody in the notes, that does not mean that they're in the meeting, because it was just seeing a name and it assumed that it was in the notes. And so that's not hard to do. And so now we have this chat, we're doing this big CRM, review and meeting with different departments. We basically wrote a, we basically recorded the video from our 45 minutes conversation with each department, dropped it into our GPT and had it rewrite it as a summary. And that piece there helped us really dial in the output that we had. And then now we just, we don't take notes, we just record it on teams and then run it through there. And then we send it out and fact check it, review it. 

So a couple practical examples and I think I'll wrap that section up by saying a lot of what ChatGPT is actually honestly really great at, from a productivity standpoint, is I would bet you're similar to me where you have this type of session, right? And you're like, I wish I could do this and I had time for that, and I wish I could build this content out, and I wish I could send it out to two different people because they have different needs, and that just takes so much effort and time to write. How do you just let Jackie T. Do that? And so your time and effort is the editing piece for all of that, right. So, wild.

Jack Hubbard: That's unbelievable.But this all does start with prompts. And, you know, when I started, and I'm a rookie, on this, I'll just put a little bit of a prompt in. It'll help me. I don't have a Persona that I want to talk about, et cetera, but talk about prompts and what they are. And what are some best practices you've seen in terms of people using them?

Ben Udell: Yeah, great question. Let's start from the basics for everybody, and we'll accelerate from there. A prompt is really the direction that you want to give to ChatGPT, and I think the mistake that people make is they assume it's going to be like doing a Google search. Right. Tell me about home equity, or, tell me about baby boomers. Well, that's fine. And you're going to get searches from that. What starts to make chatGPT and some of these other services really great is your prompt starts to sound like, write me a blog about home equity focused on baby boomers. It should take about five minutes to read. You are the marketing director for a community bank, right? At a 10th grade reading level. Include at least three key points that are bullet points. Include a call to action when you're done.

And so what happens is when people start using ChatGPT, they'll see these ultimate prompt guides and they'll do one prompt at a time. That's not necessarily bad, but if you wanted to write this block I just mentioned, it's not going to get you anywhere close to that. The more specific you are about what you want to accomplish, the better direction you give to AI. One example that I like to give in my training is if you say, tell me about home equity for baby boomers. From the Persona of a PhD professor doing research on this topic, that starts to give you a whole different type of output than the one that I just provided. 

So as you think about using prompts and layering prompts and stacking them together and then iterating on those prompts to get five or ten sessions deep, that's where this magic starts to happen. And when you think about that, what I find, I mentioned this earlier, is people will say, that's great, but it's written too corporately. Two, we're using ten cent words and we use five cent words. And I'll ask my team, I'm like, okay, so what do you want it to sound like? I just want it to sound more casual. All right. Write it more casually. Write it at a 10th grade reading level. write it to an audience that does not understand home equity. They've never had one. And it's this natural language kind of discussion back and forth that starts to prompt them. I would say at times when people get, maybe you get some interesting results out of it or you don't get what you want, is probably a better way, is sometimes ChatGPT is going to take you very literally.

So, for instance, that transcript conversation we just had, I was, early on, I was saying, here's a transcript of a video. Summarize this for me. It was giving me an executive summary of two paragraphs, and I was really frustrated. And it kind of dawned on me, literally giving me a summary. What I need to do is I need you to rewrite this in full with all of the details so that I can communicate effectively to internal associates so they understand the content in the video. That was significantly better than some of this other stuff. It's really just this working through and manipulating it as people get started with it. One of the guidelines that I would give is I would look to push it to the extremes. 

And so that home equity example is. I think we'll use that for a second. Right. Ask it to write to baby boomers. And then ask it to write to Gen Z. Ask it to reference, relatable, concepts to baby boomers. Relatable concepts to Gen Z. Ask it to write in, you know, with 2024 references. Ask it to write it like a 1980 or an 1860s gold miner. Ask it to write like Tom Hanks. Ask it to write like Anthony Bourdain. And, and not that people are going to do that, but that starts showing you these kinds of boundaries and how you start dialing in what you want to happen. And all of that stems from not doing one prompt. It stems from really thinking through the direction you want to give it and then iterating on that to get to a point that you want to get to just, You got to play in the sandbox. That's the exciting part. That's the fun part, too, honestly, it is.

Jack Hubbard: And you mentioned that you've got about 10% of your bankers there. So let's say it's 30 to 50 bankers. And you talked about the training program. This is near and dear to me. Talk about the training program a little bit and how you teach your bankers how to use this.

Ben Udell: Yes. So really it's, what I've been doing is a two hour session. The first 30 minutes is really back to the bias, the privacy, making sure they understand our policy about using AI safely, the guardrails, for lack of a better word. And then the next hour and a half, it's really just running through examples. And so whether people bring examples or I create examples, we really try to say, like that example I had, just a second ago on write it for Gen Z, write it for boomers, write home equity, write high yield, write a difficult email, write a topic to a, manager, a leader. You need to persuade, have a difficult conversation. it's really just running through that so they can start to see all these little pieces and how it applies. And so it really, I think, needs to be this kind of hands-on laboratory where they can kind of play around and figure that out. And I think, honestly, part of it is also having some fun. 

So, like you mentioned, Dali is, People don't appreciate it right now, in that paid version, you can upload documents, but you can also generate images. And so then it's kind of fun to start generating some images off of it as well, to keep let the 2 hours fly by. I don't think Dali is at a point where you're probably using any of these images externally, but if you think about trying to communicate something internally, or having some clip art, for lack of a better word, it's definitely pretty good with that. 

So once they're done with that session, it's really like, go forward. We have a team channel that I run that we try to share and communicate things that worked or didn't work. and I think for a lot of people, though, the message when we're done is start everything in ChatGPT, right? If you have to write something, communicate something, you need a difference of opinion. You need to have three different versions of something. you need alternative perspectives. Start with ChatGPT and then as a follow up, when people send stuff out and it's maybe not as polished as it could be, or as concise, you know, you could kind of follow up and say, I don't think that went through chat. JPT did it. And they're like, no, I didn't quite get there. I got to do that, Ben. 

And just, you know, and I think this is one of these things where from an accountability standpoint, what I think through is, you know, imagine, imagine bankers back in the day who said, I'm not going to use excel, I'm going to do everything by hand, or I'm not going to use the Internet, or I'm not going to use my phone. And at some point in time you got to get on board. I mentioned we're starting to roll out Microsoft Copilot, we've got ten testers rolling that out, which is a whole other animal, same thing. This is going to be embedded for literally every worker. And that's a difference maker. It really is a difference maker, yeah.

Jack Hubbard: And community banks and big banks pick a size. We don't have a lot of bodies laying around that we can give work to. And so this is going to really make us a lot more productive. I got and we're running up on your time and I want to be sensitive to that. But in the program you're going to do for the financial brand, I got to believe you're going to talk about Personas. So if I'm a community banker and I don't have a lot of staff, how do I build my marketing Persona.

Ben Udell: With ChatGPT yeah, the short version is what I did is I basically took 2030 pages of our content that we've already put out. That is our brand. I put it into one PDF to make it easy, and I uploaded it into ChatGPT and I said, I amon our marketing team. I need you to review this material and I need you to help me define what my marketing Persona, brand voice is, however you want to phrase that. And I want you to give me a detailed report of what that is. And so it gave me, I don't know, a half page Persona, right about we're community involved, we're client focused. I think we tend to be a little more folksy or colloquial given where, you know, south central Wisconsin base. Right. We can do that. and it gave me some other key pieces. 

So now I've got that and then I can kind of go in and I can tweak that a little bit more. I can add to it. It's never an end document, but now that I have that, when I take writing, I can basically drop it in and say I need to write again a high yield savings promo to 30 year olds that are saving for a home or big purchases. And I can run that through and have a first draft. The other part of that Persona then is you can start to also say, you'll notice AI likes to use certain words, like, for whatever reason, it loves to use the word realm. And I'm like this, like, for some reason it loves Lord of the Rings. And I'm like, no, we're not using real, so do not use the word realm. I like writing at a 10th grade reading level. I think what we find, I think the stats say that people generally like to read and write at that level. And I think that you think about how from a financial, product standpoint of really making sure you connect to clients and they understand products, that seems really fair for a mass audience. 

So right at a 10th grade reading level, and you just dial it in. Like, I think we're probably 80% of the way there from dialing ours in because at the end of the day, your brand is more complex than run it in and run it out. But as we see the output that comes out of it, we can start to see things we like and we don't like, and that allows us to continue. If I can get our brand Persona and GPT to like 90%, I'm going to be really happy with that because that means that output's going to be pretty darn close and high quality. I think it's high quality now, but I think it just makes people's lives easier.

Jack Hubbard: Well, the humans can take it. The other 10%.and I want to ask you about the future. Ask a quick question about the president. One of the things I'm listening to here as a community banker is, okay, he's got a $3 billion bank. He does all this digital stuff. What kind of staff do you have behind you? And if you are going to advise people, who do you hire going forward to help with chatGPT.

Ben Udell: Yeah, yeah. Good questions. So from our standpoint, I've got a unique role where I lead digital innovation, which means I can get involved in all of this stuff. So on the one hand, my team is small, but I work with tons of people across the organization. So from my standpoint, our marketing team is basically 1234 ish people for this. And that's all marketing. my team itself, I've got a marketing analyst and a, ah, digital product specialist. And then I work closely with all these project managers and people from across the organization. And so my guidance in that type of situation is I think you've got to find somebody like myself, who's ready to champion it, who's comfortable getting embedded with all of these other departments, somebody who really wants to push the boundaries without pushing too far. and so, and I think part of that excitement, like, I'll give you an example. 

On Friday, I basically had some of our deposit, or had one of the ladies from our deposit services team basically sit with me throughout the day. And we didn't talk about high level ChatGPT. We talked practical, we talked about writing policy. We talked, we basically talked, I'll give you a great example. We took some procedures that they have, that they've been scanning and using through time and have handwritten notes on it. We basically scanned that, and rolled it into ChatGPT. And for a lot of that stuff, ChatGPT will read the document and the handwritten notes, write us a new version in our brand, and we can greatly accelerate the quality that we have in bringing everything up to, up to par. And so I think that's the key piece, writing a GPT for the marketing team to use, you know, being comfortable with doing something similar for, again, deposit services. 

Now you can start seeing how other departments can start to benefit from doing that as well. So I just think you got to find the people who are excited for it, they want to apply it, they buy in, they're willing to share, and if you can get that, you know, that 5% of your bank starting to do that, you know, people next to them are going to be like, wow, how do I do that? And how can I get involved there? And, it's just going to start to grow.

Jack Hubbard: So look in your crystal ball, Ben. Where is this headed? What's the future hold for ChatGPT, AI, all of that?

Ben Udell: Yeah, my niche is very clear. I think on the practical side, you put it well, that democratization across an organization and across the industry, what I think is probably going to continue to happen is that clearly ChatGPT is going to grow in usage. I think the other piece that's going to be interesting will be Microsoft Copilot. Having Copilot embedded into all of your Microsoft products for all of your associates is going to be a huge game changer. And plus that's going to contain your data into a private space. So you're going to be obviously comfortable using private data. ChatGPT offers that as well. We just haven't got into that because we have a copilot coming down and there's so much other stuff to do. 

So what I'm getting at is I think there's a couple pieces here. The first is I don't think with Copilot coming down the pipeline, people should wait on that and skip ChatGPT. I think there's going to be a need for both of those. Copilot will be very appropriate for the masses within an organization because you can just simply rewrite an email, you can change your word document around quickly. ChatGPT is probably going to be like a lot of other software programs where 1510 percent of your organization is going to have access to it or a paid access to the others. You got copilot, you don't really need that. I think the real challenge here is twofold for organizations. 

The first is making sure, making sure that you have somebody that's driving forward to find the productivity gains, the creativity gains, and the upskilling games with skill sets. Think if you can do that and how it applies and get departments using it and elevating it, you're going to win. The other part is there absolutely needs to be training for all of this. And so you really do need a strong training program. And I think we've done a nice job so far, but we're still not there yet. We're still probably in spring training, for all intents and purposes, with AI. But who is really your AI trainer? Because an AI two plus two actually equals eight. And so you've got to really help people understand how to do a lot of this. And I would wrap that up by saying, I think certainly in the long run, I don't know if there's going to be a lot of need for these prompt engineers, people that are building these gigantic prompts out. 

But I think in the short run, and certainly in the long run, this is like a mini embedded person, it's like back to who understands excel really well and who is your IT person that's building programs? There's got to be people within your organization that's going to say, here's a two page prompt that's going to, that has been built and tested to help you be more productive. I think we're going to have to find people that understand that and can build that. And so, that's going to be one of those game changers as well. 

So hopefully that's a little coherent. But maybe it's a crystal ball, too. So it's always a little hazy, but, I would step away from all that and summarize it by saying, get people testing ChatGPT, get people testing co-pilot, get people looking for practical examples that are meaningful for the organization. So people buy in and build it, and then start building out, whether it be a training plan or some sort of method to help people understand how this is used. And I would wrap that up by saying we were joking. Some people remember back in the day where you could go to the library and you could sit down with a librarian and they would help you with how to use the Internet. I think that's a great analogy for where we're at right now with ChatGPT. It's just, how do you get those people internally to be able to do that? So anyways, wild times. Wild times. And that's the tip of the iceberg.

Jack Hubbard: It is, Jack. It is. Absolutely is. Two more questions. I know you have to go. I'm a banker, and I get called on by vendors who now have some expertise in this space. What advice do you have when vendors come? A call?

Ben Udell: Yeah, I think it's probably a couple quick things. The first is, of course, everybody likes to say that they are AI. I think we have to take a step back. Technically, AI is really this machine learning umbrella. So you might have under this umbrella the generative AI that we've been talking about for our time together. That's really the kind of the writing, the comprehension, eventually some of the video and image creation. Right? You also have machine learning. You also have RPA, the robotic process, automation. All of that's under basically this machine learning. So I think you really have to understand what that is. And one great example that starts to come up is using AI for chatbots. 

From what I'm seeing from vendors, they like to call it AI. And the reality is, it's machine learning that you still need to program and monitor and work through. And so when we as consumers, that frustration with using chatbots, and I can't understand what I'm asking for, I still think that's generally the case. So I would pay attention to what that actually looks like and what you want to accomplish. because technically, is it AI? I would say, yeah, absolutely. But is it the AI that we've been talking about for 45 minutes? No, I don't think it is. I think the other thing is that you could do so much with a $20 a month ChatGPT setup. I don't know that for a lot of community banks, you have to make large investments. Right now, if you think about how you manage fraud, we've got Visa, Mastercard. 

Our core providers are really good with helping manage that. What more am I going to do? We've got vendors who do a lot of this, this AI stuff, and how do we just feed off of that versus building or buying it on our own? the other thing I was going to say is, Oh, Microsoft Copilot is kind of fascinating right now. if you're a sharepoint user and you have all your documents in a cloud, there are providers that basically want to come out to you and say, we can build this program that will basically read all of your documents. So if you have a client support person who needs to quickly reference documents, we can sell that to you. 

The reality is it's in copilot, and I'm testing that right now. And I think Copilot's like $30 a month for users. So not only can I go into copilot, ask for our wire, I'm sorry, ask for our overdraft policy to see if I should waive something, but then I can say, this client is not going to accept that. How do I best share with the client? I'm not going to waive the overdraft. What are analogies I can provide? What are good questions I can ask? And so you start thinking through this client experience of blending all of this together, and that's the exciting part. And I'm not sure you really need to spend big money to make that happen. You need to invest in your people to help them understand how these systems work and the benefit that actually comes out of it.

Jack Hubbard: This hour has gone by like nobody's business. And one of the things I love about this, Ben, is you're so practical in what you do. I can imagine that out at the, financial brand, you're going to be a huge hit. And I got to believe that the trade associations see this and others, they're going to say, I got to get Ben, I got to get Ben to be speaking. You're going to be a very busy man. How can people reach you, Ben, if they want more information on you, and certainly more information on how you and your bank have used chat?

Ben Udell:  The easiest way is, you can certainly find me on LinkedIn. I'm easy to contact on there, and I'm sure we'll kind of post that along with this as well. You'll also see some of the commentary and insights that I provided. I also run a website called robots. Robots are people.net. And so that's kind of my way away from the bank to generate generally probably a weekly blog on AI content, practical applications, and some of those other pieces as well. And that's found on my LinkedIn as well. and I think you're right. My niche is definitely on the practical side, the democratization of it. And, how do you spend $20 a month but get some amazing gains, and how you're, basically a better professional when it gets down to it?

Jack Hubbard: Well, you spent $20 a month on ChatGPT, but this interview is priceless.

Ben Udell: Ben, you down?

Jack Hubbard: Thank you so much for your time. All the best, and really, appreciate it.

Ben Udell: You're welcome, Jack. Happy to be on, and, I, hope everybody, enjoys the conversation.

Jack Hubbard: Thanks for listening to this episode of Jack Rants with Modern Bankers with my great guest, Ben Udell. You look at ChatGPT a little bit differently, but you do. After this program, when I want information on banking, I have to go to a person. It's Chris Nichols. Chris is my guest next week, and we hold a wide ranging discussion on the economy, community banking and much more. It's all here with Chris Nichols next week on Jack Rants with Modern Bankers. Now, this and every program is brought to you by our friends at Vertical IQ and RelPro, this LinkedIn live show. Well, it's a podcast, too. Subscribe and get the latest episodes of Jack rants with Modern Bankers. And please leave us a review, if you would. We're on Apple Podcast, Spotify, Google Play, iHeartRadio, and others. Visit our website, too. It's themodernbanker.com, and you can get more information there. And don't forget, sign up for our free public library, 26 brand new ebooks all about LinkedIn and sales. You can get them for free at themodernbanker.com./publiclibrary and also remember, make today and every day a great client day.