Building an AI-Powered Community Bank with Greg Solomon

speaker-0 (00:21.144)
This is the podcast by Bankers for Bank, Caleb Stevens with South State Bank’s Capital Markets Division. Today on the show Nichols sits down with Greg Solomon. He’s the IT at Bank Independent in Al. And like we’ve done many times now on this show, we return to the topic of AI. There’s a lot that’s going on in the banking world, and this episode talks about what it actually looks like when a community bank starts putting it into action.

speaker-0 (00:50.424)
Chris talks with Greg about building AI agents, about how to automate workflows, getting your employees to adapt the technology, and also put the question of ROI, you know, can you really measure your return on an investment like this? And they also talk about why the next phase of AI change how banks operate and serve their customers. So before we get there, I want to tell you quickly about the ARC program. You know, with a few

speaker-0 (01:21.236)
gets payment certainty, but you, the banker, can expose yourself to interest rate with a far was taking on that payment risk. With a head with the ARC program, you get the best of both worlds. You get a floating or gets a fixed rate, and unlike most programs and interest rate swap programs, there is zero derivative or running on books to

speaker-1 (01:31.47)
Yeah.

speaker-1 (01:39.224)
Hedge programs hedge account on your book manage, plus you can earn up

speaker-0 (01:47.502)
Upfront fee income as well. It really is the worlds where your borrower gets fixed rate certainty, but you get a floating rate that will help stabilize your net interest market. To learn more, click the link in the show notes of this episode. Now Chris Nichols and Greg Solomon.

speaker-1 (01:50.722)
Best of both worlds.

speaker-1 (02:18.318)
Greg, appreciate you being on. Welcome to the show. You and I talked a couple weeks ago in in Dallas. Yeah, we were in Dallas. And as we were talking, I was really impressed by kind of what you’re doing on the AI side and the agentic AI side. So wanted to kind of showcase that for some of our listeners as I think they’d be interested in your direction. So Greg, welcome and thank you for being on.

speaker-0 (02:40.3)
Yeah, Chris, appreciate it. And it’s always a pleasure to talk with you and and talk about technology and what’s going on.

speaker-1 (02:47.48)
So give us the background of your position in the bank right now.

speaker-0 (02:50.254)
Sure. currently Director of Information Technology, Senior Vice President at Bank Independent. Been here about nine years total, about thirty years in the industry altogether. And you know we try to strive to not be bleeding edge but a little bit of cutting edge with our initiatives.

speaker-1 (03:08.92)
And I would say you’re one of the more forward thinking backs. How big is the bank?

speaker-0 (03:12.392)
we’re over approaching over three billion in assets. All right.

speaker-1 (03:17.122)
All right. and tell us about what you have going on in terms of AI, gen gen AI, and then genic AI.

speaker-0 (03:26.776)
Sure. You know, a little bit deeper we started an RPA program several years back to help improve automation and efficiency across the bank and the organization for departments and then AI as it progressed became a part of that process and what we could adopt in artificial intelligence technology and one of the key initiatives we had early on was knowledge and the knowledge that our people had and documents and day to day operations and so we we built an

Basically, a governed data structure around document retention and document management. And then we built AI generative agents on top of that. So basically, we built a support structure through an AI agent that our team members can now ask it instead of picking up the phone and calling and asking someone. And so it is working really well. The adoption rate has been really high. We’ve also adopted some of the Microsoft framework.

obviously since we’re a Microsoft Shop, about what tools they present to us and Microsoft Fabric and Foundry has been a big part of that. And we’re now ingesting a lot of our core data into that environment and we are starting to build more AI agents on top of it to be more productive and analytical with that data and then some product, you know, analytics and predictive analytics around around that data.

speaker-1 (04:51.51)
So let’s delve into that a little more. for starters, how many on the AI, Gen AI, RPA side?

speaker-0 (04:59.938)
Team members? So we started out with a couple, had an intern that came in from one of our Division One schools and helped, and then but right now we currently have about four dedicated team members in that environment. One of them acts as a Microsoft engineer and the other ones are they have two AI analysts on on staff currently. but that team is around four people.

speaker-1 (05:00.97)
Yeah.

speaker-1 (05:23.438)
Got it. And talk a little bit about your tech stack. You said you’re a Microsoft shop, so I’d love for you to explain fabric versus foundry and I imagine you copilot and

speaker-0 (05:33.666)
Yes, a lot of copilot. I’m sorry, yeah. All of our agents are currently built inside of a copilot and copilot studio. So they’re custom agents that we’ve built that are shareable and they you can present them across the Microsoft three sixty five platform into Copilot. pretty easy to deploy those.

speaker-1 (05:51.64)
Got it. And you sh those the copilot your gen AI and a gen AI driven off open AI, Chat GPT basically, or do you also have Claude?

speaker-0 (06:01.256)
so the other big initiative we have is using artificial intelligence assisted code development. So we do have several developers in house and we do have a lot of custom code that has been created over the years, some one off tools. And recently we have just learned that it is that’s one of the most efficient things we’ve we’ve experienced with this is converting old code to new code and building applications faster and

Honestly just quicker to UAT testing than what used to take a year takes a week, you know, sometimes with the initial launch. but it’s we’re doing a lot of that also. and it’s I think that’s gonna pay dividends.

speaker-1 (06:47.436)
And and that’s most banks’ number one use case is what we’re finding. So where they’re getting the most immediate lift. What in your opinion, how big a bank do you have to be to take advantage of that? I mean, you’re I get I get all the time we’re s you know approaching 70 billion. So we get, you’re South State, you know, you can afford all this. I’m like, no, like it’s even more efficient for a smaller bank that can move quicker. and you’re an example of a three billion dollar bank, but w do you think a it would make sense for a billion dollar bank?

speaker-0 (07:14.338)
I think it would make sense for any bank at this point. If you have the right technology stack, which most corporations and banks are some sort of Microsoft shop. Microsoft provides the basic tools. It just takes someone to take the initiative to learn how to use those tools and and and work within their environment. but and even just general day to day, we we tell people here that, you know

you’re responsible also for your own AI initi not initiatives but just learning it and making yourself more efficient. But you know, some banks I think have been cautious because of the data loss prevention and you know, where’s my data going? Is there is there data governance there? What are the guardrails in place? But I don’t, you know, if you’re a five hundred million dollar bank to a fifty billion dollar bank, if you’ve got IT folks who can dig in and learn this, it can

it can help you in your day to day no matter what.

speaker-1 (08:12.702)
For sure. For sure. before we move on, maybe just give an executive level explanation of fabric versus foundry versus you know, copilot and copilot studio.

speaker-0 (08:24.046)
Sure. So I would say Foundry’s the basic foundation. Microsoft has Fabri f Foundry IQ services, which is where some of their machine learning and their AI tools lives within the Foundry environment. Fabric’s part of the overall environment. And basically what Fabric is is the repository. so if you think of Google BigQuery and their data hub and Looker, that’s Fabric is the same thing in Microsoft. It’s just a place to

You can build your pipelines, ingest your data, use Copilot on top of that data. You can build your semantic models to do Power BI reporting, which is our business intelligence team is starting to do that. And it basically streamlines that function. But Fabric is basically the data warehouse or data lake repository that Microsoft presents to you. And then they they put the tools on top of that like Copilot and Foundry IQ and machine learning for you to manipulate the data and present it to you to whoever.

speaker-1 (09:22.458)
Got it. And how how far down does the or in the organization does Copilot Studio go? So are you giving your business line managers access to this or is it just really the realm of IT right now?

speaker-0 (09:37.452)
The studio custom piece is in the realm of IT only. So we build the it’s built off governed data because data accuracy obviously plays an important role and we knew that early on that a lot of our documentation that lived inside of SharePoint and things were old and outdated or there was multiple copies of it. So once you get the data governance in place, so we control that in Copilot Studio ourselves. But if we have several

Copilot premium license in the store, they can just use the general copilot, obviously, agent in their desktop and they can build their own agents internally or adopt already built agents to help them with their email or write a document or build a PowerPoint and so on.

speaker-1 (10:24.096)
And do you have published public published agents for the organization yet? Is are you are you approaching that?

speaker-0 (10:30.53)
Yeah, we have one published, went live this year. it is got about a seventy percent adoption rate. Microsoft has some benchmarking tools inside of Copilot Studio where it kind of gives you some basic analytics on adoption, usage, time saved, and things like that. And yeah, we went live with that. That’s our knowledge based system. it’s called Billy and it is it’s doing well. So it’s used in the front line for customer.

service agents or tellers or people in the back office, it’s got a lot of our internal documentation about a lot of things in there and those documents are now governed like I said and then it’s got a high adoption rate right now.

speaker-1 (11:13.102)
Yeah, that’s right. so that and that’s that includes policy?

speaker-0 (11:15.598)
Policy, procedures, procedures, how-tos, general knowledge, sales material, limits, HR information, they don’t have to call HR anymore. and and we’re building up on that, and it it’s it’s providing a lot of value right now.

speaker-1 (11:21.074)
Sales material.

HR information.

speaker-1 (11:37.038)
So as as we talk to, you know, 4,000 banks in the in the industry, that is by far the number one use case. I mean, that’s how South State started with our Tate. sounds like that’s kind of where you started. And so that’s been a popular starting point. Like just make your employees more efficient. and one of the offshoots to that, I think you and I discussed is that it made our employees more confident. their ability to talk to customers.

was increased and I think like their general satisfaction was increased. So in addition to saving time where we used to have to search the inter intranet and get a couple of links like Google and then figure out which links appropriate, now we just ask Tate and it tells us. So I imagine you’re you’re seeing the same savings and seventy percent adoptions. Fantastic.

speaker-0 (12:20.534)
Yeah, it’s same same concept, same results. It’s just an anything that we can now validate and put in there, we’re we’re just gonna use that and and we’re hoping to use Billy as the, I would say, master agent. And then if we build other agents that is coming down the pipeline, we would just, you know, bring those agents in under Billy so they have theoretically one interface but have access to multiple data sources.

speaker-1 (12:47.214)
So let’s talk about that structure for some of the bank executives that are listening that may not be familiar with with an agent team or an agent swarm. You’re gonna use Billy as also your orchestration agent. So if you need to find or do some document intelligence gathering, you may have a document intelligence agent or research. And so Billy will do the orchestration. Is that kind of how you guys are gonna think about it?

speaker-0 (13:10.606)
That’s in theory the architecture we want to pursue is is use him as the orchestrator and then as the agents we start building like we have two or three agents right now that are probably in beta that are probably getting ready to come online, I hope, in the next thirty to forty five days. And we want to just use Billy as the or orchestrator to bring that in. And so it’s basically you’ve got you know, it’s kinda like a organizational structure. You start building that out and

You know, you keep your interface consistent with with your frontline users.

speaker-1 (13:44.93)
Got it. So if I’m in your organization and say I I develop an agent for you know, vendor management, one of my favorite topics, or you know, expense management, or what have you, how long would it take that to get approved in your organization and published?

speaker-0 (14:00.992)
so we would theoretically run that through our robotic process automation team funnel and they have a team that meets who and then we have an intake form where people submit general ideas for process improvement on our internet system. So we we take all this and we vet it out into priorities and what you know it’s not necessarily what always has the biggest bang for your buck, but it’s just what makes the most sense at the right time and we kind of prioritize those and put them in a pipeline and then and then once that hits, if it’s a PMO

type project. We obviously get PMO involved, our project management office and then and then we run with it.

speaker-1 (14:38.295)
And you so you think six months? Is that

speaker-0 (14:41.89)
Yeah, I think, you know, as AI progresses, I think it it gets shorter, maybe. And especially as you’ve you know, like you’re riding a bike, as you’ve built one, you’ve built two, and you built three and you and you get your structure and your architecture and your administration down and the support, I think it goes quicker. I think a lot of it depends on where your data lives too. As we’ve learned with trying to corral core data and ancillary data sources into into a system that can use

AI, I think that’s part of it too. But I think six months is reasonable, honestly.

speaker-1 (15:14.584)
Yeah. And and hope hopefully shorter over time, right? Yeah. Yeah. I know all banks are struggling with that and all you know, all business line people like myself are f you know, always frustrated. So, you know, you gotta crawl, walk, run, but you know, I’m I’m looking to run faster, but it you know, six months is is killing me. but like you said, like I I have hope and I have reason to believe and there is evidence to say that we are getting shorter as a industry.

speaker-0 (15:37.39)
You it’s funny, you know, they always say, I want it tomorrow. Well that that might happen soon.

speaker-1 (15:42.38)
Yeah, exact exactly. Exactly. and I do give banks a lot of credit for, you know, I always joke like, that fintech is, you seven billion dollars and they just started up on like, well, they don’t have a lot of legacy systems. They don’t, you know, all their data’s in one place. You know, they they have a lot of advantages. And when you have a legacy bank that’s been around, you know, a hundred years, it makes a big difference in terms of what you can do. And a fintech without the regulations can move even faster. So I’m like, you know, give banks some credit. Like it’s hard, it’s hard to move a battleship.

speaker-0 (16:10.786)
Yes. Yeah.

speaker-1 (16:11.998)
t tell us about the future of your w where do you see banks? You know, you you’ve had experience. So what advice would you give someone in your position that’s just starting the engine journey to get going? What what would you say is critical for them to get done in their organization?

speaker-0 (16:31.286)
You know, our first when when when when this first came about and the technology really hit the last couple of years, we we really focused in on the guardrails first. how are we gonna guardrail and protect obviously our data integrity and our customers’ information and our team members’ information and we don’t know how large language models are using this data in their training models and how do we know it’s not getting leaked out? So I think guardrails

Guardrails and governance to me is one thing you have to have a concept of in place before you begin. Because if you don’t understand that, it might lead you down the wrong path or have some data loss prevention issues. I think that’s key to that. And then having the resources to learn it, adopt it, figure out, you know, what is the best path for my organization. We we took some time, we looked at Google, we looked at Microsoft, we looked

We’ve looked at dozens of architectures and you know, we always came back to Microsoft because we’re a Microsoft shop. and then, you know, figuring that out and then putting a plan and process in place to have intake for, you know, what are you gonna use AI for? Is it knowledge, is it data, is it a gentic to take the place of a process and, you know, free up people’s time and and et cetera. So I think those are the key steps in getting into it. And once you get into it, just maintain a

a team and a structure of continuously learning what’s coming next because this is evolving fast. Kind of keep your line in the sand, but be ready to move it occasionally, so to speak. But I think those are the key steps, you know, if you’re starting now from scratch, I think that’s key is to get governance and guardrails, pick an architecture, pick a framework, and then find the right tools that are, you know, that your team can use and utilize to help.

provide benefit and value to your organization.

speaker-1 (18:28.51)
Tal talk a little bit about the quality of Microsoft. Are you happy with them and then what you know, we’ve leaned on Microsoft quite a bit for additional resources. What’s your what’s your big experience with Microsoft and how should banks think about, you know, asking for more help from that company?

speaker-0 (18:41.962)
I think, you know, overall satisfaction is okay. Copilot has struggled off and on. It has obviously changed dramatically, even the interfaces and things since we started using it a couple of years ago. I think they’re getting there. They’re adopting other models into copilots, so you have more a little more choices on models and usage. and I you know, copilot itself, you know

is in the top three or four, I guess, of overall models if you look at Chat GPT and Claude and some perplexity and some of the other ones that are out there. But I think they’re steady. They’re obviously a big company. It’s a fast moving target in the adoption and I’m sure they get a lot of feedback from millions of people daily on how to use it or what to do. But I I think, you know, we’re about eighty, eighty five percent with them.

speaker-1 (19:30.594)
Yeah. Yeah, I’d I’d say my experience has been you know somewhat positive. I give lot of credit. They moved that that battleship pretty quickly. And while they didn’t get it right to start, they quickly got it right. And and while Copilot’s not my favorite LLM, I will say it’s becoming more and more useful and it’s to the point where I think it’s actually dang good the fact that it’s integrated in my Excel spreadsheets and now integrated my Outlook, you know, at the PowerPoint. so

I I think it’s fantastic, first effort for banks that they can only choose one thing. I would say, you know, get co pilot premium for everybody UK.

speaker-0 (20:07.81)
Yeah. I think the integration of them into the office products in their own environments has been it was kind of slow like you said at first, but I think they’re getting there now to where it’s integrated into every you know, everybody uses Office, right, in a corporate environment. So I think they’ve got that integrated well and they’ve added Copilot and the SharePoint integration has really come along nicely. So if you use SharePoint as for document repository or other things, it it it integrates with it really well.

speaker-1 (20:32.908)
Right. And you and your one drive is your primary file storage, right? For your your employees. Yeah. Yeah. Yeah. talk a little bit about the connections. The other big advantages I see to all LLMs is this extent that they’re connected to other applications. I’ve been somewhat impressed in the last year.

For Copilot getting connected to more and more applications for so banks on workday, for example, you now have that integration. That’s, you know, we’re a workday shop. So I’m very optimistic that that’s gonna really save me time in the future. Any any any comments on integrations and what you guys are doing there?

speaker-0 (21:06.382)
Yeah, we’ve been a little slow on that front, maybe. We do have it connected to our PMO tool, which is Monday.com. that’s just started out. We’ve just got that going the last, I don’t know, last 30 days or so. you know, our HRS system, I think we are looking to possibly see how that can connect in to ADP or, you know, any other HR system that we have also, but we’ve not we’ve been slow on the connectivity side.

basically because of the guardrails, we do have some restri we are restricted somewhat at our egress levels on connecting to other other models and and systems. But I think as we slowly and methodically get comfortable with that and figure out the connectivity more, I think we’ll have more adoption.

speaker-1 (21:54.536)
Excellent. do you think you’ll go down to the citizen developer or business line side on Copilot or you think you’re gonna hold steady? Is that the so that’s the plan?

speaker-0 (22:03.702)
I think we’ll hold steady until we get everything in place. but I think getting it deeper into the organization only helps us become more efficient and helps that team member become more efficient in their job. and I think if the citizen developer gets deeper down, I think there will be I’m not sure how far we’ll take that, but I think as some of these department managers and they have some leads and things like that, I think the adoption for that is is gonna be high for us.

speaker-1 (22:34.004)
w what do you go w how would you characterize your F T account, your staff at the bank in terms of access to Copilot Basic or Copilot Premium? Does everyone have it or how do you ever

speaker-0 (22:47.022)
Has access to Copilot Basic. that’s how they access Billy, the knowledge base system. And then we have I think we’re closing in on about 50 to 60 premium licenses. We do have a Chat GPT Enterprise account that we have about a dozen so folks that in IT may basically and and some executives that use it. And then we’re about to adopt a clawed enterprise model for just IT only.

speaker-1 (23:15.32)
Got it. Interesting. And how many employees there?

speaker-0 (23:18.21)
Probably. I think we have about f I think we’re about five seventy.

speaker-1 (23:23.586)
Got it. Got it. And as you look ahead, I imagine you’re starting the budget process or the middle of the budget process.

speaker-0 (23:30.664)
Yeah, so you know, budgeting for AI is a hot topic.

speaker-1 (23:34.466)
Yeah, so w what are your thoughts there? What wh what are you gonna do next year?

speaker-0 (23:38.232)
We actually just saw a company that they it’s the first one I think and then maybe it’s coming out, but they built a tool that kinda looks at all the models and the cost models for all the different organizations and kinda built a quasi cost cost tool that you can ingest how many licenses you have per per tool like ChatGPT or Claw or Copilot Premium and then how it uses and what you’re using for and it would kind of spit out a number but

You know, we’re building budgets now. I think we’re building budgets to see how if we have third parties come in and help us build out more agentic things, in our loan processing stuff. But I think the budgeting piece is is is kinda tough. It’s grown. A little bit of sprawl and so we’re trying to get that kind of nailed down for next year.

speaker-1 (24:24.886)
Yeah. And so on the budgeting piece for token usage, I imagine that’s starting to be an issue like many banks on how to control that and how to manage that. And you know, you you want to start burning tokens and have all your employees play with the tools, but i you also have a budget to worry about. So it’s kind of a balancing act is what we’re finding across the industry. So

speaker-0 (24:44.96)
It’s kind of like, you know, it’s kind of that they get you in there and they get you to using it and you’re like, this is great. I can do more, I can do more, and then all of a sudden it, you know, doing more costs more also. And as those tokens get, you know, as get higher and higher, we yeah, it just I think the more we use it, the more we’ll pay. But I think I still think there is ROI in the process of using those.

speaker-1 (25:12.248)
And then like you said, having some governance tools to help you with your different models and you know token management and usage and make that visible, I think, to employees. I think that’s kind of the next push for many banks to get bet a more handle on that. Get a better handle on that next year. on budgets, do you think your IT budget is up three percent, five percent? Where does that if you had to guess, because every bank that’s listening to this is trying to figure out where to set the budget next year.

speaker-0 (25:37.302)
Yeah, we were looking at budgeting numbers for just AI last week, early last week, and because I approve those every month and it’s it’s I would say we’re in the two percent, maybe one and a half to two percent range of our overall budget that might be hitting and that includes our Microsoft fabric environment. You have to obviously get into an F sixty four license with Microsoft or one of those upper tiered frameworks and and that has a cost. But we were already a Microsoft shop using Power BI tools, which is included in some of that.

but I think it’s probably about one and a half, maybe two percent tops.

speaker-1 (26:13.026)
Got it. So fairly steady compared to this year. but I imagine AI will take a larger piece of that budget overall budget.

speaker-0 (26:20.716)
we start we’re looking at more tools too that are out there for day-to-day for workflow mapping like we talked about the other day in Dallas, you know, describe and who loop and some of these other tools that are out there can it help actually help us on identifying but you know workflows that we can automate or building documentation faster. So I think as we bring more tools on the budget will will increase once we can get those identified and approved.

speaker-1 (26:45.176)
Quick, just talk about that for a second, because I think that’s lost on many laymen. if you’re in the business line listening to this, you know, if I’m dealing with a loan process, maybe it’s a maintenance issue for a loan or maybe it’s onboarding, what have you. I have a certain workflow and talk a little bit about how these tools help. if you want to automate that process.

speaker-0 (27:06.706)
Sure, today without the tool, you obviously have to send a resource down to sit with them, try to map it out or try to have them map it out while they’re trying to do it. now they have these tools that you can load in the background and it kinda sits there and records and documents what the user does on a daily basis. So if you let it run for a week, they’re smart enough now like Scribe to to kind of sit there and map out, okay, this is a consistent process over a five day period, let’s build out the workflow for this and

Those that all happens in the background with no human interaction. And then obviously there’s a human in the loop to take that workflow out and review it and say, okay, this is this is correct. And so for like, you know, a loan underwriting or mortgage underwriting, which is a consistent basis, obviously it doesn’t change from day to day or some deposit ops process that’s done day to day. Those those things can really help now, you know, go in the background.

record what the user’s doing, build out the documentation, build out the workflow. Then you can take the workflow and say, okay, how can I automate this? It’s basically done in a couple of days versus now, you know, a month or so of people’s time and then and then if you take the workflow, you just ask Chat GPT or ask Claude to, hey, build me an automation process for this workflow. They’re gonna go b it’s gonna go build it in a few seconds and then you’ve got a model to play with. I mean, so that adoption is is

Is I think that’s key going forward too to be competitive and and, you know, I would say do more with less, but those tools have really progressed.

speaker-1 (28:40.236)
Yeah, so I want to emphasize that just for any banks that are thinking about that, that that I think is almost critical for a couple of different reasons. Like you said, the old process was to get, you know, we used to have 20, 40 people in a room. It would take at least a month, if not, you know, three or six months to try to figure out what the process is. We would use, you know, sticky notes or or what have you. We would then, you know, document it, diagram it out, re meet, you know, relook at it, and then figure out like how to automate the process. Now

It is definitive. No one’s BSing or no one’s going from memory. You now have a definitive workflow of your process. You can ask as you said, you can ask AI, how can you improve this process? So it uncovers, you know, issues of where you emailed and you know, it asked for some approval that you might have forgotten about. So it captures that type of workflow and then suggests a better workflow. So you’re not just automating a bad process, you’re now automating and using AI for a refined process with hopefully less steps and then

you can now code off of that. So it’s a fantastic tool that I think our banks are just starting to loot use. I know when you talked about it, I’m like, that’s impressive that you’re already down this path because I think most banks some banks are just thinking about it. Most banks are not, but I want to make sure that’s highlighted.

speaker-0 (29:54.314)
Yeah, I think one of the key factors to that is is don’t automate a bad process. It it may identify a process and you look at that process and go, we can do this a different way and be better at it. And so I think that’s another key key point you made there is is just you know, don’t automate a bad process, review the process and make sure it’s the right one before you go down the automation path.

speaker-1 (30:18.242)
And we all talk about that, but it is amazing how many times I’ve seen us sit in a room and just apply traditional, you know, what I call banker think to a process because we’ve always done it this way. We don’t know anything better. We assume that our policy or regulation. And the reality is that there’s actually flexibility in the policy. There’s flexibility in regulation. We don’t have to do it that way. There’s other, you know, third-party data we can buy to augment this process and speed things up or whatever it is and come up with a new process. And well, you know, we all

We all talk about not not automating a bad process. It just a it’s a knee-jerk reaction that you constantly gotta watch for. So great point. Brent, you just alluded to this, but on on the ROI side, you know, hot debate about what is our ROI for AI. I you know, I make the argument that we might be looking at this a little wrong, at least to start in that co-pilot, for example, has made me more productive, but I’m just producing more emails, more reports.

You know, more vibe coding, more graphics, et cetera, using Copilot, better s better looking spreadsheets. But my company’s not really getting an ROI out of me per se. I’m doing more work, sure. So so somewhere it factors in. But there’s not an outright ROI direct that’s directly tied. Are we in your opinion, are we is that question are we getting that question wrong, or is it just yet to come that we have to get to the enterprise, you know, side of things before we really see a true ROI? What do you think?

speaker-0 (31:42.88)
I I believe it’s the latter, honestly. we’re still kind of in the infancy of of AI and the adoption of and I think, you know, we did establish some benchmarks early on and time to measure, you know, KPIs around AI. I think everything deserves a cost benefit analysis, whether it’s automation related or not, you know, and to say you’re getting ROI off of saving my time.

Because my time is obviously going to go towards something else. And is that something else producing an ROI also? And so I think there needs you know, Microsoft is providing us some basic tools around copilot adoption and copilot usage in ROI, but that’s not the total picture. I think. I think as a company, if you want to run a cost benefit analysis of providing automation processes across the board, I think there’s gonna have to be some kind of enterprise.

adoption of what that model looks like. And I think that’s probably subjective at times, honestly, because it’s our opinions of I saved you X amount of time, it saved me X amount of money, but the tools and the cost associated with that, I think there’s a lot to factor into that. And then AI has ramped that process up a little bit and, you know, made that term a little more if emphasized.

speaker-1 (33:00.846)
Yeah, so it’s that messy middle that I think bankers have to get comfortable with as we transition from you know, just strict productivity to real true cost savings.

speaker-0 (33:10.678)
You know, and I think also I think the organization I think a lot of times we get we see the the big organizations adopting AI quickly and are we doing that fast enough? And we do talk about ROI, but is it, you know, gonna make is it do we need to adopt AI quicker to be more competitive in in our marketplace is sometimes the bigger question around

speaker-1 (33:34.69)
Yeah, right. And I think, you know, at at South State it’s become, you know, like a merger for us that we report on it. I’m responsible for it. It’s in my performance review, you know, it’s in everyone’s business line objectives. So, you know, I think that’s the level that it needs to be, at least they give us credit for trying to figure that out. But it’s a big push around here. And whether we overindex on it and use it, you know, too much and not officially enough, you know, remains to be seen. But it’s at least a decent way to deal with it, I think.

speaker-0 (34:02.85)
No yeah.

speaker-1 (34:04.326)
Greg, you are one of the more experienced CIOs in the industry. Your work here is impressive. That’s why we ha we have you on. What advice given your experience that you learned in the past year, for example, that you might give someone that’s just starting off in your position?

speaker-0 (34:21.598)
You know, one of the things is adoption. you know, sometimes our technology experience in the industries we get locked in sometimes on certain things and how things should work, infrastructure, cloud based applications, software as a service. those things are evolving quicker and quicker. And I think now we have to pay more attention and you know, constantly be researching and educating yourselves because

That technology environment is changing very quickly. And if you look back a year ago or even two years ago where we were, you know, that has evolved so fast and these tools have put the power of of technology deeper into the organization with I would say reduction of outside resources or outside help. And I think that has really opened us up and I think, you know, trying to

Trying to adopt your team and ch and transform your team to think about things differently because of AI has really evolved in the last twelve to twenty four months.

speaker-1 (35:30.006)
Great advice, great advice. I wanted to end it on one more question. being that it’s so topical, Muse, everybody, you know, the fastest app, the biggest selling app out there right now, you know, I think ten days is number one spot. you know, Meta’s Muse, where you have a personal agent now looking at things connected with plaid the banks, it’s Muse, it’s Instinct, it’s Dots, it’s a number of these agents. What do you think the stance should be? Any opinions on

How banks should think about it is, you know, we saw Amazon and JP Morgan on their earnings call the other day said, we’re gonna block it. and now JP Morgan says that and then often reverses their position. But you know, one blocking it is definitely one idea. The other is to lean into it, you know, somebody like a Walmart that has said, we’re gonna lean into this and we’re, you know, if you’re a bank, we’re gonna get more business from it, not less. Any thoughts on that on on what what a bank strategy might be?

speaker-0 (36:27.074)
You know, we had this conversation here I think about two weeks ago about how our digital presence and the digital the digital experience that we give our customers and I think our customers are gonna want more integration with these platforms and they want more ease of use and exposure to these things. And I think as that progresses, obviously it’s a little scary with the connectivity and

things into your platform so as as your core providers provide these secure tools like Plaid, PLAD integration, you know, Plaid has come a long way and we’re it’s basically being integrated in a lot of different areas inside the digital banking world. But as ChatGPT gets plugged in, you know, we talked about this is like, you know, I think it just it gives you the AI analytics of an advisor or an agent or a you know financial planning, the things like that. I think that the people are going to

speaker-1 (37:21.688)
Yeah.

speaker-0 (37:22.732)
Yeah, a private banker. It’s going to integrate more and more to the fact to where y you’re going to be the account holder, but you may have these tools that are going to to be expand the digital experience for your customers. I mean, so the fact that you can go in there and say, Hey, run an analysis on my monthly spending and tell me where I can save money. I mean, right. Obviously that can’t be done in some of the digital experiences right now with some of the platforms, but I think

That is what our consumers are going to want to expect. So I think as that adoption rolls out, I think it’s just gonna be part of our research and experience of how we provide those tools to our customers in a safe manner.

speaker-1 (38:05.314)
Yeah, and if for any banker that hasn’t played with it, if you feel comfortable, it’s worth playing around with both ChatGBT and and Muse. it does connect to your bank accounts, it does make recommendations on which banks you should be using. It does look at how easy the aging can pull information off of banks. so to the extent that it can pull rates and pull account structures and things like that. It often suggests that. no surprise that the Allied Bank, Capital Ones, Wells Fargo.

are some of the banks at CIT are more open to leveraging AI information and it may as a result it makes recommendations off the bank. And then it’s just a matter of time before, you know, what we call the PAP the agent protocols get adopted where they’re actually transacting on your behalf. and now banks have to make that decision. But I think that’s somewhat interesting that should these protocols get adopted, and you do have banks, agents moving money from banks or opening accounts at banks.

It’s a whole new world to think that th these agents are our new customer base and how are we gonna deal with that?

speaker-0 (39:08.268)
Yeah.

speaker-1 (39:09.88)
Greg, fantastic. Thank you for being on and sharing your wisdom with the industry. We appreciate it. We look forward to hearing more of your journey. And I can’t thank you enough for spending your time with us.

speaker-0 (39:19.496)
You’re welcome, Chris. Always good to talk to you.

speaker-1 (39:22.67)
Take care.

speaker-0 (39:23.31)
All right, see you, man.