Transcription:
Penny Crosman (00:03):
Welcome to the American Banker Podcast. I'm Penny Crosman. AI leaders come from many different backgrounds and walks of life. At Huntington Bancshares, it's Chief Financial Officer Zach Wasserman who is leading AI strategy for the bank. He's here with us today to share what he's thinking about as the bank implements AI. Welcome, Zach.
Zachery Wasserman (00:24):
Thanks, Penny. It's great to be with you.
Penny Crosman (00:25):
Thanks for coming. So to what extent is AI a priority at Huntington?
Zachery Wasserman (00:33):
It's a tremendous priority. I personally believe, and I think we generally believe that AI will truly revolutionize corporate America. And the applications that we're seeing evolve now are really just the tip of the iceberg, truly of where this will ultimately go. And so our view is we want to stay well ahead of that curve, be very actively engaged not only in these early stages, but out into the future as well. So we're not only creating a significant organization behind it, but lots of inertia around building it deeply within our strategy and go to market approach.
Penny Crosman (01:13):
And how did you, the CFO, come to be the overseer of AI at Huntington?
Zachery Wasserman (01:19):
It is a little bit unusual. My responsibility is to lead and oversee all of the finance functions of the company, but also all of the strategy functions including strategic planning, our mergers and acquisition function, our corporate ventures function that does early stage investing and a lot of innovation work and partnership development with typically fintech companies and the enterprise data and analytics capability. And one of the thought processes we had is that data analytics and where we're going with Gen AI is so critical that we want to deeply embed it with all the other strategic planning and strategy functions. There's also a lot of organizational synergy, I will tell you, and the team finds quite a bit of benefit being together as one, both from a career planning and just learning perspective. So that was the thought process around it and we're seeing really nice outcome from that.
(02:14):
I will tell you that as you think about AI and data analytics generally the reality is back to what I said before, this has application in almost every corner of the business, both the internal processes, the capabilities that sort of support on a mid-office perspective, customer facing activities and the customer facing activities themselves. And so the question that we often wrestle with is are we putting enough resources against this? And so there's a real benefit in having the finance team be heavily engaged in this because we want to continue to be fueling significant growth in the resources. But then also the key question is what aren't we doing right now? Where should we be putting more analytical and more kind of ideation against potential use cases that we aren't doing yet? And so that's where the integration with a strategy functions incredibly valuable because that team really knows where we're going as a business and therefore can get ahead of it from a data analytics perspective.
Penny Crosman (03:25):
So did I hear you say you're increasing your spending on AI and can you quantify that in any way?
Zachery Wasserman (03:32):
Yeah, we certainly are. Generally speaking, and this is admittedly off of a relatively low base, we're effectively in a doubling period almost every year that's going by here. To give you a sense, I think about this as a CFO, as I liken myself to a, someone working on an 18 hundreds railroad and we're trying to shovel coal into a furnace as fast as we possibly can to really fuel acceleration Huntington. Over the last five years, we've almost quintupled our technology development and AI is on an even faster ramp than that. So it, it's pretty dramatic in terms of the growth we're seeing and I think that'll continue for some time.
Penny Crosman (04:18):
And with the CFO mindset, are you looking super closely at those expenses and expecting to achieve a return on that within a certain amount of time?
Zachery Wasserman (04:31):
Yeah, so the answer is yes. We look very clarify at it and an analysis of the business case and the return on investment is an integral part of every piece of that part, every piece of that growth in terms of where we're going with data analytics, and I'll tell you we're already seeing very significant benefits from it and we can elaborate on this in more detail, but just at a high level, the impact on marketing, the impact on pricing on what we as a bank are doing in terms of deposit gathering and also internal efficiency in terms of what our colleagues and all of our internal processes are doing is quite meaningful and we're already seeing true and good profit outcomes in the p and l today.
Penny Crosman (05:25):
So what are some of those specific use cases? Can you share a little more detail about some specific ways you've been doing this so far?
Zachery Wasserman (05:33):
Yeah, absolutely. In terms of traditional machine learning, not so much large language models, but just in terms of traditional machine learning ai, there's been a tremendous investment in what we call MarTech, marketing technology, which really enables multiple different use cases within the customer acquisition and customer interaction process. One is around optimization of which customers we reach out to with what marketing offers specifically. Another is around optimizing the engagement within our, what we call our digital storefront. This is the platform that we have within our digital channels to engage customers with a deepening of their relationship with additional products and services. With us machine learning is powering all of our pricing optimization and major business lines that have a lot of high frequency pricing interactions, like indirect auto financing and financing for our RV marine business. Another one that we do a lot of work on is around fraud modeling and reporting.
(06:45):
Clearly fraud in the industry continues to rise and grow faster than overall transactions are. So there's quite a bit of focus on understanding that, parsing that, really making sure that we can see both positive and negative negative signals there. In terms of gen AI leveraging large language models, this is earlier stage, but it's really building momentum quickly. A couple of recent use cases that are getting a lot of traction, one is around software engineering and code generation. So think you're a software engineer and you're using a copilot to write the next page of code. That can drive a lot of efficiency that for you other applications we're using it is in knowledge management. The application there would be looking at all of the policies and procedural documents of the company and quickly being able to synthesize the most important elements of that for someone who needs to understand how to interact with a customer or manage a process. The biggest opportunity I see in the near term to continue to drive this forward is around staff efficiency. So just huge opportunities to leverage large language models to accelerate the productivity of our teams who are summarizing documents, pulling together materials to interact with customers, creating client reports and things of that nature. So there's really no end to the kind of efficiency improvement you can see within core internal processes.
Penny Crosman (08:36):
Well, I know you're not in the compliance or governance area, but just generally speaking, there are several foundational generative AI models out there that banks are gravitating for towards. How do you decide which ones you're going to try and work with? What are some of the things that you look for when you're vetting these companies and models?
Zachery Wasserman (09:05):
Yeah. Well, I think we obviously want to use the models that are the most tried and true and tested. So back to what I was saying at the beginning of our discussion, this is still very much the early days here and I think we're still as organizations learning as much as applying and so ensuring that the models that we engage with and ultimately rely on are the most advanced and most tested and validated as possible is the most important consideration. Another one comes down to security. And obviously this has been a theme across the whole industry, want to be very, very cognizant of where is our data and how can we ensure that our data does not leave our enterprise and stays within our walls only is not used more broadly within these models. And so the ability to ensure that and validate the security of the data is another major element. What's interesting, and I think this is kind of evolving in the AI space generally, is the largest players within the industry are really building the foundational models, but there are a number of other smaller, more innovative companies that are kind of leveraging that underlying technology to then customize the application of that model in various ways. And so that's sort of an emerging area that represents a lot of opportunity as we continue to move forward.
Penny Crosman (10:45):
And I think you said you have some kind of operating model that you use when you are deploying AI projects. Can you tell us a little bit about that?
Zachery Wasserman (10:54):
Sure. The question of what's the operating model is honestly almost just as important as what is the specific application or idea for leveraging analytics. Because as running a scaled enterprise, you don't want to just create one great business analytic use case. You want to create hundreds or thousands at pace and with the best alignment to your strategy as possible. And so how you organize against executing in the data analytics space through an operating model is honestly the real secret sauce. And there's a couple of dimensions to it that we've really thought about in Huntington, a lot to optimize for us. I will tell you that there's no one answer here. And really the operating model that works best for any given enterprise is very much a function of where that enterprise is and its maturity and the different elements of this business. But I'll just share you with you for us four or five key attributes that we thought about.
(12:00):
One was to what degree are the analytic resources centralized versus federated throughout individual business units and really distributed throughout the company. And where we've landed is we want, the ideal scenario is our analytic resources to be as close to the business as possible and therefore as federated as possible. With that being said, a lot of our true, what I'll call hands-on keyboard analytic resources are centralized, but they're servicing groups that have been created in each of the business units that are developing the analytical use cases, ideating on the next best area to deploy analytics, and then ensuring that that gets driven into business processes in each of our businesses. So it's a federated ideation with a centralized analytic capability. Another element, and I think I've touched on this a little bit earlier, is you always want to be thinking about what's the strategy and the roadmap answering the figurative question, what's next?
(13:07):
And so another element of the operating model is we have now a very explicit focus on developing the next five or six likely use cases for machine learning and gen AI in each of our major strategic pillars and ensuring that there's always a roadmap of continued acceleration. One area that the whole industry continues to evolve and really think a lot about is governance. This is a new technology, there's lots of promise, but there's also potential risks. And so really thinking about how can we make the governance process as effective as it can be, making sure that the governance process is very, and approval processes for new models is very clear to stakeholders so that we can go through that process in a expeditious and efficient way, making it risk-based. So that's really an important element of it. How complicated is this application and what would be the implication of error to ensure that you're putting the risk management governance with a focus on where the biggest risks are with more efficiency, where there's less, and importantly always keeping a human in the loop. Another one, I think you touched on this a little bit earlier in terms of returns, but we're putting a fair amount of resources against tracking and metrics to deliver value.
(14:40):
Just as an aside on this, I mentioned earlier that one of the biggest areas of effectiveness or value creation is in the area of efficiency within the staff. I will tell you that there's plenty of applications that we can develop that would save 10 to 15% of someone's time. And in fact, if you look at industry studies, they would say that if a full application of gen AI could typically save somewhere between 20 and 30% of the average staff members' time, however, actually extracting that value is very difficult. How do you get a savings out of 10% or 20% of somebody's time? And so thinking very planfully then about how we would restructure organizations so that the technology driven work is extracted and we restructure roles throughout the organization to actually create those savings is another major element. And then just lastly is around resource allocation and growth. And I think the orientation is really with a forward planning look saying we want to significantly increase resources over time to this area. We don't exactly know what we would do in 2026 here, but planning very much for growth, scaling the organization, and ultimately allocating the specific resources on a time by time basis. As we go throughout the course of the year
Penny Crosman (16:13):
With this emphasis on efficiency, what would you say to employees who are worried about their jobs?
Zachery Wasserman (16:21):
Look, I think the only constant in life is change. And to some degree the phenomenon that AI represents in terms of efficiency opportunity is the same fundamental opportunity that a personal computing and computing offered in the 1990s and early two thousands. And the same is true throughout the course of the industrial history. There have always been technological advancements that have the potential to increase productivity of people, but that would mean potentially less people working on any given unit of work. And so what's critical is that we train ourselves to focus on where work is adding value, where people can have a unique element of work and to be flexible in taking on new roles. I will tell you, if you think about the major themes that are emerging in terms of the implications of ai, one of the big ones is that roles are changing.
(17:24):
The role of people is changing and the role of banks is changing. And I think I'm thinking about this in my own right, what's the role of a CFO in the future and how much of that will be done by AI? And the reality is some of it will. And so I need to do just as what everyone else would do, which is, well, where can I add the most value and how can I leverage this technology to create value for my company? And if I'm doing that well, then undoubtedly, ultimately there'll be a great role for me here.
Penny Crosman (18:00):
Is there anything on your wish list of things that you would like AI to be able to do that maybe it's not doing today?
Zachery Wasserman (18:09):
It's a good question. I think sort of as I said at the outset of this call, ultimately AI is going to touch almost every element of life and commercial activity. We are starting now with the areas that offer the best return and with the least probability of risk because it still is early days. And so I think as we continue to go forward, there's not really areas that I wish were able to be deployed that aren't. But I certainly believe that over time we will get more and more into customer interactions and customer facing capabilities. And that's really the next bastion of where this will go. Obviously that not present a lot of opportunities to create technologies that are actually more responsive to what customers want and really help to build trust and ease of use and customer value, but if deployed incorrectly could represent risks. And so I think that'll be the area that we'll continue to see the fastest development and the most future activity.
Penny Crosman (19:25):
Alright. Can you tell me a little bit about Huntington's venture investing work, and have you invested in any AI companies?
Zachery Wasserman (19:33):
Sure. The Huntington Corporate Ventures team is a really exciting area that we spend a lot of time on at its heart. What the objective of the organization is, is to accelerate innovation for our business units. And the approach is to be very, very close to what's happening in the innovation economy, to create partnerships that would bring capabilities, new functionalities to our business as well as potentially new customer bases. Practically what we do is the team is invested in a number of venture capital funds and we making direct private equity investments in a number of early stage and more technology driven companies that really help us to stay at the forefront of what's going on. And the proof in the pudding that I always look for as a CFO is what are we doing to drive customer value and to drive revenue back to the company?
(20:39):
And I'm really pleased with approaching a dozen scaled partnerships that are really in fact doing that. One of them that's really notable in this area that is linked to the conversation we're having now is an investment we've made in a company called View, which is headquartered in Florida that is innovating AI based services for the healthcare industry. If you take a step back, I mentioned a minute ago, the role of banks continues to evolve and AI will accelerate that change. And one of the things that we are trying to do is to stay at the forefront of advice and guidance and new value added services that we can put around the banking relationship. And so in our healthcare business, one of the issues that healthcare companies often wrestle with is the process of billing between healthcare providers and insurance payers to create efficiency in that process, let less risk and potential fraud in that process and faster speed. And so view is innovating a use case that rapidly accelerates the ability to do billing, reduces human error in it, and creates both cost efficiency and working capital benefits in terms of speed. So great application and as we think about the other major industry vertical focus areas that we've got within our banking business, there are untold number of other AI applications that we could generate. So that's a really fruitful area for innovation.
Penny Crosman (22:20):
Alright, well thank you for sharing that. Thank you all for listening to the American Banker Podcast. I produced this episode with audio production by Wenwyst Jeanmary. Special thanks this week to Zach Wasserman at Huntington Bancshares. Rate us, review us and subscribe to our content at www.americanbanker.com/subscribe. For American Banker, I'm Penny Crosman and thanks for listening.