Podcast: White Clay CEO Mac Thompson talks 8 steps for AI implementation

Whitney McDonald 08:11:48
Matt, hello and welcome to The Buzz a fin AI news podcast. My name is Whitney McDonald, and I’m the editor of fin AI news. Fin AI news has rebranded for bank automation news, marking the next step in our mission to lead the conversation on innovation and Financial Services Technology. Joining me today, November 25 2025 is Mac Thompson, CEO and founder of white clay, Mac is here to discuss what financial institutions must consider when implementing emerging technologies from data strategy, basic business goals and talent. Thanks for joining us. Mac.

Mac Thompson 08:12:19
Awesome. Thanks, Whitney, hi. I’m Mac Thompson. I’m CEO and founder of white clay. We started about 20 years ago. My partner and I left bank of Bank of America and our last jobs at the banks before we left. And for about 10 years, we were a custom software consulting company, and we pivoted about nine years ago to a more SaaS model. Took us a couple years to do that, but we’ve been in business. Our clients range from about three 50 million in size to 200 billion, and we help our clients build deeper, more profitable relationships, and one of the ways we do that is embedding a lot of intelligence about the clients, utilizing very large data sets from transaction data and account types of all kinds. So very excited to have the conversation. Great.

Whitney McDonald 08:13:06
Well, we will get into all of that. Let’s kind of take a step back first. Let’s start with the state of AI adoption. We’re really at a place now where it’s not so much if financial institutions are going to be implementing AI, but rather when, obviously it comes down to size, capital, resources, priorities. But where does AI adoption really stand today?

Speaker 1 08:13:32
Well, I think probably one of the challenges in answering that is what? Because a lot of this is a definitional thing, because AI has been in banking for a long time, there’s about 20 different technologies that are kind of AI Artificial intelligence related technologies, and some of them I used, we were using back when I was at the Bank of America a long time ago. I think a lot of AI now is more large language models, generative AI, and that’s how a lot of folks are defining it. So I think adoption of some kind is pretty high. It may be as simple as someone on a personal level, signing up for it in terms of generative or large language model kind of usage. But most of the fraud models, most and a lot of the customer service models, chat bots, particularly, and things of that nature, being using some version of AI for quite a while, a fraud particularly. And so while most banks are using that and it’s more mature, a lot of the smaller ones are getting into it, I think the giant nationals are much further along in building their own internal, large language models, trained by themselves, built internally, utilizing their very large, comprehensive, statistically relevant data sets. And so there’s a large variation in that capability, but the nationals are definitely leading the way in terms of pure capability.

Whitney McDonald 08:14:57
Now when it comes to questions that your clients are asking about implementation, and maybe we can kind of lean more toward the emerging AI technology, what are they asking of you? What are kind of those questions that come across your desk that you kind of see a little bit over and over

Speaker 1 08:15:15
again? So I used to have a, I used to be a CFO when I was at the banks. Is one of my jobs. And I switched over the, you know, the dark revenue side, which is a lot more fun, the dark side, yes. So the, one of the questions I ask is, what’s going to be the ROI on this, the return on investment? And I think it’s a challenging question, especially the emerging AI technologies. I mean, the study from MIT, it came out in July, basically that 95% of projects don’t generate any discernible financial benefit. They may be benefits, but it’s not necessarily financial. Those 5% that do them have pretty outsized benefits from it. It’s one of the high levels from that. And I think when I answer the question about, How do you determine ROI, I said, I don’t think you should be worried about ROI at the moment. It’s like the internet in the early 90s. Mid 90s, it’s going to mature, it’s going to have much better use cases and return on investment cases, but as an organization, you’re going to have to think about what it means to be aI empowered, AI powered as an entity, and that’s a journey that’s not just technical, that’s a Cultural one, and just how you think about yourself, a paradigm almost. So the ROI question comes up a lot because they’re like, Well, should I go invest a lot of this? And one of the other things, if you’re a very small bank, going out and hiring three or four AI scientists is probably not going to be your best return on investment. You know, you’ve got 120 people in your. Company, you’ve got more people in your AI department than you do in your IT department. Probably not going to work out great, so partner up with some people would be my recommendation on that. The second thing that probably comes up biggest is data. And it’s, is my data ready? And the answer for most folks is, no, it’s a mess. It’s not it’s got all kinds of issues, and you’re gonna have to work on that. The other thing about data, though, is it’s not an end state. You need to start working on it. You need to start using it. By using it, you’re gonna figure out what you need to fix. You aren’t have perfect data before you start using AI,

Whitney McDonald 08:17:29
yeah, I think that those are two things that we have definitely covered on our side as well, that ROI that has to be a hard mindset shift, especially coming from a CFO background, that maybe the ROI doesn’t need to be the top priority at this exact moment. Now, let’s kind of talk through this eight step process that you have, that you share with clients. You have these, these eight steps that you share when approaching AI and implementing AI, that should be top of mind. Can you talk us through those?

Speaker 1 08:18:01
Yeah, I’ll sort of walk probably with you all eight steps. But that really starts with, what do you try to do business wise? So a lot of technical reason why the ROI on a lot of technologies, including AI, don’t work, is that the technology investment wasn’t grounded in a business something you were trying to do. And so, you know, I think the MIT article actually came back to the number one challenge a lot of folks are having is integrating these new AI technologies into their workflows and the work processes and all that. So the first thing is figure out what you want to do business wise. And then second part of that is determine if there’s any of those things you want to do where AI would be very helpful to you. And you kind of just start with those basic business questions, because if there’s not really anything that AI can help you with, and what you want to do business wise, you got to think about what you’re doing. Probably the next two pieces we just talked about the data piece. Start working on your data governance. Start working on a data strategy. Start down that data path. It’s going to be a path. Don’t try to bowl the ocean. Don’t go out and hire 12 vendors to work on your data. Starts more slowly but deliberately, working on evolving your your data capability. But with a data capability also comes to people, is that you’re going to have to you’ve got large teams that have worked with you, that love to work, take care of your customers and all of these things, you have to come up with a way to help develop that talent. So as you’re developing strategy in parallel to that, you’ve got to develop your organization’s human capital capability to be able to start thinking about these technologies. Not that they have to be experts, but they have awareness and they can what they need to do. You know, probably the I’ll actually stop, I’ll stop one. There’s one thing I probably do to start on all this, though, write a one page definition of what AI means to your financial institution, bank or credit union. So when you’re talking about AI, are you talking about generative AI? Are you talking about large language, whatever that is. Just to find the terms, because I’ve been in rooms with 20 people in there, there’s five different definitions of AI being used, and they don’t know what it is. So that that common language around what this is, it gives you a basis to start working on education. But the first is that when you’re using terms, that everyone knows what those terms mean. And if you have a vendor or someone coming in, it’s also helpful, because you can define the same terms. So when they tell you something, you may think they’re saying X when they’re really saying y. So it’s probably one of the first things. Don’t overthink target, like where you’re going to be in three years on this? Because we don’t know, there’s a lot of people coming up with AI strategic plans for five years, and I don’t know how in the world they’re ever because if you had talked gone back three years, what would we have been talking about, right? It wouldn’t have been this. So don’t overthink that. Don’t overthink long term tech strategies. Unless I’m not talking about the giant nationals. I’m not talking about even some of the Super Regionals that are making very large investments, talking about most of the banks out there. And probably one of the more important pieces around all of this is how you start thinking about governance, around your data, around models you may use this to help empower decisions with does it have any regulatory impact? Are you creating unintentional bias and things that you’re doing? And you know, all these sound kind of complicated, and they are, well, what helps is don’t try to solve everything initially and just start the journey, because it’s going to be a journey we’re going to be on for a while, and it’s going to take a bunch of different turns. And. It’s all right, but just start. I would say, start the journey is probably the first thing. I would say,

Whitney McDonald 08:21:57
Yeah, I like what you mentioned there about, you know, you don’t necessarily have to have that three to five year strategy in black and white. Just start. And I also like what you mentioned too, about, you know, defining what you’re really trying to solve for? Create that one page plan for your institution specifically. Don’t just invest in AI for the sake of saying that you’re doing it. We’ve seen that, you know, backfire a little bit too. But making sure that you have a definition, what are you trying to implement? What are you solving for? Is it, you know, not just using AI as a broader term, but do you want an agent? Do you want a chat bot? Do you want X, Y or Z? And I think that having a really simplified, a simplified document that says exactly what you’re solving for is a great place to start. Do you

Speaker 1 08:22:47
because one of the things I’ve seen a lot of boards are just we have to be an AI. They have no idea what that means, but they’re demanding that their bank, whatever or credit union, whatever institution, be involved in AI, even though they don’t know what that means.

Whitney McDonald 08:23:01
Now looking ahead to 2026, we’re seeing more real applications. We’re seeing more efficiency gains, we’re seeing more manual processes being replaced. What are you watching for, for 2026 what are some of those tangible use cases of AI that you think are gonna pop up? What are you excited about? What are you hearing from, from white clay clients?

Speaker 1 08:23:27
Some of the ones that are more tangible are the operational automations of workflows where we’re pushing paper around, right? I mean, it sounds funny, but we banks, we push a lot of paper out. Even where we have automated systems, there still seems to be a lot of paper going around. So I think that this isn’t really large language models doing this. This is more text paper to text to and then how you embed it all more workflow oriented. Lots, lots of folks are doing that on a practical level, and they can get some efficiencies, because they’re essentially digitizing processes. One of the things I think is a challenge is they’re digitizing the legacy processes, not thinking about, if I had this technology, how would I, how would I not even use this process? I would just do something completely different. And this is banking, and we’ve been doing this for a while. And when we basically automate cow paths, you know, where cows walk from one destination to the other, they build these paths. And a lot of roads are actually built on these old legacy wilderness paths that animals, Buffalo and whatnot, would create. And a lot of what we’re building is automating those, digitizing those paths. And I think the really big step this goes back to your business. What are you trying to do? If you really thought about do I even need to do half the stuff that I do is where there’s tremendous opportunity and efficiency and impact, because we, right now are doing a lot of digitizing of legacy things. So we’re seeing that that’s on the more operational, trying to get some efficiencies right now. The other thing that is out there is this movement from and this has been gone up a little bit, but we ran into this headlong you originally think about a spectrum where you go from offers to insights to recommendations to solutions. A lot of folks are using AI and other technologies to create all these offers, next product, logical product and things, product pushing on a way the other thing we got into is we’re generating insights. And here’s all these insights that we can now generate, and our ability to generate insights has massively outpaced the ability of the people in the field, they’re interacting with clients, to do anything with these insights. We did this ourselves. We’re guilty of this. We created, you know, we had couple 100 insights per client, and that, you know, in a branch may have 2000 clients. And so what do you do with all so what I’m seeing is coming up is, how do you take all of this, simplify it, and turn it into something that can be really constructive for both the client and the bank. And that is, I think, the next evolution of all this, and that’s getting into agentic, is one word. But agentic, of course, means 25 different things to all kinds of different people, right? I mean, they Gartner’s symposium down in Orlando. You know, agentic was, you know, agentic AI and a Genty web was buzzwords that are out there, but what it means really depends on the problem the person. But that concept that we’re going to be able to take all of this intelligence and put it in motion, put it into action, is, I think, the next evolution, and I see some people trying to get into that. There’s vendors trying to do it. There’s things trying. Union is trying to do it, but I think that agentic evolution is coming, and it probably will be talking more about agentic in 26 than we were talking about generative, because it’s basically, how do you get a personal assistant? That’s this agentic agent doing things for you instead of but once again, we’re probably back to automating Cal pass, because we’re trying to get them to automate things that we currently do. I think the next generation, which probably a 27 thing, is when we’re starting to get into the agentic web, where the web is more like a resource we interact with that we have people go do things with. What happens to the web was a more proactive agent for you, instead of a resource, it more empowers how you were thinking. It’s just a very different way of interacting with these massive data sets that are out there, kind of scary in some ways. I mean, people run into that, but I think, you know that’s kind of longer term where we’re going. But in general, it’s how you start taking all these capabilities that we’re building, that we have created and beginning to integrate them in a way that makes people’s lives actually simpler. Because right now, we’re actually making life a lot harder for a lot of our bankers. We’re trying to help them, but we’re just give overloading them with so much stuff they can’t use it. And how do you how do you get that value out? I think will depend upon us simplifying it, making it more actionable, more simple, and I think that’s where we’re going.

Whitney McDonald 08:28:15
You’ve been listening to the buzz a fin AI news podcast. Please follow us on x and LinkedIn, and as a reminder, you can rate this podcast on your platform of choice. Please be sure to visit us at finaI news.com for more fin AI News, thanks for listening. You.

Transcribed by https://otter.ai