Share
Greg Kanevski is the Global Head of Banking for ServiceNow. He joins Senior Editor Rachel Koning Beals on the BAI Banking Strategies podcast to talk about how AI could first boost the importance of data-management roles and M&A scouts within your organization. Plus, Greg suggests AI might be the addition to your compliance team everyone welcomes, and much more.
Key Takeaways:
Don’t miss the other insights, methods and tools that can help your financial institution shape and evolve its AI strategy in the BAI Deep Dive: A practical approach to AI.
Greg Kanevski, Global Head of Banking for ServiceNow, welcome to the BAI Banking Strategies podcast.
Well, hello and thank you for having me here. It’s a pleasure to be here with everyone today and appreciate those that are listening in.
In many ways, traditional AI has been changing banking for years, but it’s generative AI that will now transform our industry. How so?
AI has been a part, most folks would think of it going back to the days of RPA [Robotic Process Automation], where you bring the bots in and the bots would automate normal business processes, but it was really the unintelligent AI function. So that’s been around for banking almost 10 years now, and it was the first wave of really taking the most unvalued human manual processes and trying to find a way to throughput those in the operational functions. But of course, bots became on top of bots and then it took people to manage those bots and that became unsustainable. So the AI rave has really been here for a while and slowly the environment has grown.
But what I’ve heard recently through some of the education that we’ve done is all of the data collected in the history of the planet up through 1990 is collected now every two years. So every two years, all of that data comes back into the environment. That ability to mine the data and to take that data and then propagate it into the processor is really what’s changed the game. So it’s no longer a unintelligent bot. It’s now the ability for machines to process huge amounts of data in order to truly provide AI type capabilities.
Generative AI, the next level of intelligence, which has really emerged over the past 24 months, is the ability for us to take that data and then intelligently say, “Before we had to complete a step, which was A through D. Now we don’t really have to do A through D. We can take a look at all the data we have in that process flow, on that customer, for that customer base, whatever the scenario may be. And we can analyze it and then take out on an ad hoc basis, maybe steps B and C. It’s going to go right from A to D, but we only need to do B and C. And we can do that on an individual basis for all of the scenarios in seconds over a large class of issues, over a large class of customers.”
That is truly why generative AI is changing the game because of the dynamic nature of how it applies itself. So that’s really the progression of how it entered into banking and really where it’s grown to in the past 24 months, accelerating rapidly.
The speed and the scope that you’re talking about here, it’s connecting systems, it’s connecting processes across really an entire institution, front and back office. I think this is really important. It’s not just an improved customer interface. It’s not just speeding up call time. It’s really foundational to a bank’s operations, I think. I’d love to hear more.
Completely agree. It’s actually bringing the entire flow together for the first time. The banks have really operated regardless of size and their functions. Front office, where they are customer facing, middle office, processing it, and then back office, the support structure that allows those three to operate.
But they’ve operated in their own silos historically. And some things like RPA, we discussed earlier, have tried to bring those together, but nothing’s been able to bring it together at scale for multiple different scenarios.
And then dynamically, which allows, meaning the process for your issue versus my issue can be handled even if they’re the same category. They can be handled differently because of the dynamic nature of generative AI allowing all of us to be processed at the same time for the same type of issue, but differently based on our needs and the needs of the institution.
Gen AI can sound like a sizable investment as soon as our listeners hear it. How do you shape that conversation, the big budget questions, with your clients?
It’s a great question. We were at a conference recently, a banking and financial services conference, where multiple clients were together and there were breakout sessions. And during that breakout session, there were three options that they could go to. One of them being a generative AI discussion that myself and a colleague were leading. Out of everybody, there were less than a handful at the other two, everybody else was at the gen AI table. And at the gen AI table, what we heard from the narrative is we’re getting pressure from the board, we’re getting questions from the regulator. Our senior management wants to execute against those pressures because there’s a strong belief that we can save not only process steps, but we can actually turn this around and reinvest those dollars into the business.
But we don’t know where to start. We don’t know which use case to start at. We’re still struggling with what’s the difference between AI and gen AI? How do we qualify these opportunities? Who’s the expert in here that can help me make sense of the noise? That is the conversation we hear most.
Now, there are some on the leading edge that already have their strategy. There are some banks out there, I won’t mention them all, but one of them is like JPMorgan. They’re very well-versed. They’re down the line. They have a very, very strong capable unit and they’ve spent considerable energy and dollars towards getting the right people informing. But they’re one of the four main in the U.S. and they’re one of the four money center banks. They have the ability to do that. There are 5,000 registered banks, roughly 5,000 registered in the United States. That’s one of four money centers. What about all the rest of them? And a lot of them are struggling to frame the discussion.
So what we help them do is really focus on what’s important to you first, what are your strategic goals and initiatives? Based on those strategic and goals and initiatives, then let’s apply the AI lens to it. Let’s bring order to the discussion. And then once you have that order, then you can start sizing out opportunities, pick the first apple off that tree, get your first win, and then build your momentum. But you’re never going to have a shortage of use cases. It’s which ones you qualify as really best in class.
But starting that discussion of bringing order to it is first and then getting your data in line in order to support that AI because it’s one of the core fundamentals.
Let’s dig into that data piece. It inherently kind of makes its own case because it’s data-intuitive in a lot of ways, right?
AI is not generative AI without data. It needs the data in order to generate that need, that service or that solution. So take an example here of a commercial banking unit wants to improve the experience for an existing customer who wants to open a new account. Let’s say they have a deposit account, they want to open a line of credit. Do they really need to go through a brand new application and ask for all of the data like they’re not a customer today? And the drop rate, the abandonment rate is high in commercial banking or higher than commercial bankers would like because these clients feel as though they’re being treated like a brand new client when they don’t need to be.
If they’re connected to the data — they being the bank is connected — between the front office, the middle and back. When we understand where the data is, the system can go out and generate an application specific to the need of that client, the need of the bank based on regulatory and underwriting concerns that’s specific for that client and that client’s existing relationship rather than having them start over, adding addresses and articles of incorporation, things that just aren’t necessary.
That’s the power of generative AI, but it needs the data behind the scenes in order to collect what’s necessary for that client at that time and didn’t do the same thing for the next client doing it at scale and speed.
And it seems like that would help make the case as you are selling it from a budget perspective or a buy-in perspective because we are then talking also about back, middle and front office seamlessness out of the silos. It really does become a unified platform…
Absolutely. And for those below the top 50 to 60 banks, you’re also looking at unstructured data. Data that’s sitting in spreadsheets. Data that’s sitting in Microsoft Access. Data that’s not part of a chief data officer’s repository. And that’s a large number of banks and credit unions in this country that either don’t have a CDO, just due to affordability or the ability to fund all of the technology projects. And then secondly, is how do I get this unstructured data into a repository in order for me to then apply generative AI on top of it?
It sounds like we can make a case that it doesn’t necessarily matter the size of your banking operation in determining if gen AI is a good fit. Given the JP Morgan example earlier, any good examples at either end of the market?
Absolutely. And for the medium to large size institutions, let’s say the top 100, 125, their issue is data is all over the place, right? It’s bringing it together. It’s providing order of preference and enforcing a standard. And one particular client we’ve worked with, they’ve had a chief data officer for a few years now. They have a structure, they have a system of record or repository of this data, and they’ve worked long and hard in order to provide order and sensibility to how their data is structured.
So what we worked with them on is to then apply… Now that they have their terminology set, they have their governance set up and it’s working well is to now be opportunistic to the biggest areas of pain points that they have, whether it may be regulatory compliance-related, whether it could be customer service-related. And we worked with them on a few of them that actually were the nexus of both of these together where they needed to provide efficacy of their process flow, but at the same time wanted to respond to the customers more timely.
And what I’m talking about here is customer complaints were portable to the CFPB and how they can respond to these issues much more quickly and in a manner that’s much more efficient. And to do so, there’s a marriage there between regulatory compliance, customer service and truly employee satisfaction. And this firm was looking at it to, “I want to spend my dollars wisely and I want to provide the maximum value that I can.” And they were looking at those as the ugliest issues that they wanted to solve first.
And if there was a residual underlying core maintenance cost, because I can reduce the number of systems or if there was a operational cost because it doesn’t need to touched by many hands, that’s fantastic. But I also want to solve the regulatory compliance side as well. I want to help with my assurance. I want to help with my regulatory enforcement reviews and I want try to, if you will, seeing it as a triangle, I want to solve all three. And that’s the way they looked at it, which frankly is a little bit market leading in my opinion. And it really has provided them a dynamic nature.
Greg, walk us through where gen AI could help when it comes to regulatory risk management, especially potentially helping banks meet those requirements at a lower cost.
Of the top five scenarios that we hear from our customers, risk-related activities fall in the top three. And what I mean by that is sure, there are scenarios for a second line of defense, meaning the enterprise risk management function of risk modeling testing. But the first line of defense, the folks that actually own the risk, the folks that are doing the work every day really strike the best benefit. And what I mean by that is organizations, let’s take a technology department. They hire people to provide those technology services, coding, asset management, service management, operations management, whatever it might be.
But a good 30 to 35% of their time gets diverted towards risk related activity, assurance testing, RCSA process, regulatory exams. These folks have to leave their day-to-day job in order to support those activities. Generative AI can help not only the risk functions, say the core risk, the second and third lining, enterprise risk and the auditors. But the first line of defense risk teams that are there to do the assurance testing and issues management to complete the RCSAs can use generative AI to actually lower the burden upon the core folks that are hired to do in this instance, in this scenario, the IT operations and IT service management.
And what I mean by that is, let’s take an example of an assurance testing, let’s even take RCSA process. Instead of going back and meeting with all of the folks of the department to then review what issues are out there, download all the incidents that have occurred, download any of the changes to the process, generative AI can go out to the various systems and source information directly that relates to what happened from the last RCSA to this one, how many incidents are out there? How many issues are out there? Have they changed at all? What process changes? From the assurance testing, have there been any failures?
Have any tests not been conducted? And provide that summary assessment right up front that can allow them to have a better starting point. So instead of starting in the first inning, you’re starting in the fifth or sixth inning. You’re saving those three or four innings worth of work that allows them to get to that point faster and even if it’s provided in the right manner, and actually allow the manager to see that activity before the RCSA even starts. So now they’re preparing for it before the risk teams even come to them to provide the general information that’s needed to complete that RCSA.
That savings, sure, they still work, right? They still have to play the last four innings of that game. But the savings of three to four innings lowers the burden upon the first line of defense to allow them to do the activities they were hired for. That allows them to solve the issues and the incidents that are out there so they can get ahead of that curve, which in my experience over the 30 years has always been the problem. I need my people to do the work that they were hired to do in order to get ahead of this so I can start to pare back. That savings and that reinvestment is really what’s needed.
And generative AI, which is one of the reasons it’s a top three, is providing folks with tremendous amount of value of approaching it from this sense is allowing them to get a leg up on regulatory management and efficacy of their processes. On the other side, when you look at much more of the smaller institutions that are aggressively trying to grow and position themselves for growth, they’re looking to modernize quickly by saying, “Okay, I realize I don’t have the complexity of large regional, super-regional and money centers, but I don’t want to repeat the sins of the past.” And as I grow and acquire, I don’t want to get that complexity.
I don’t want to gain that complexity that the traditional banks have, but I want to start with a greenfield and I want that greenfield to be how my data is structured so that as we acquire and grow, we can bring that data in, we can bring those systems in. We can shed the underlying technology structure so that we don’t increase our spend tremendously, but also keep those costs under wraps because I want to be able to now apply AI on top of it because I have that structure. And as I grow larger and larger, the foundation is set.
And there’s been a few institutions here that we’ve worked with that are on a very aggressive, whether it’s credit unions or what I would call the interstate size banks, they cross a couple of states and they’re looking to grow and to become a regional. There’s a handful of those that have really looked at this is saying, “If I apply a few dollars today and I look at this from a greenfield, I will gain many more efficiencies and benefits back and it actually will help me fund not only my technology improvements, but help me reduce those operational costs so I can fund the acquisition time and the goals that they have for their growth over the next few several years.”
There are some fears that gen AI replaces human talent. You certainly have some thoughts on this.
Absolutely. I’ve been hearing that 30 years now in the industry, and I’ve been hearing that my entire career. At the beginning, I was fearful of people talking about the advent of computers was going to destroy the talent pool automation, that was RPA, you hear it over and over again. I am not in the camp. There was even an article yesterday that was shown on CNBC about the invention and generative AI is going to limit the number of roles and really hurt the human talent. I actually think it’s just the opposite, much like it’s been historically, some of the mundane tasks aren’t going to be there anymore.
Some of the repetitive tasks aren’t going to be there anymore, but that allows those associates to do higher level work, reinvest those dollars as long as we retrain those associates. I’ll give you an example. Prior to my current role, I was managing a first line of defense risk function and we were predominantly in that first line of defense. We were predominantly staffed for assurance testing. If I can automate those tasks and take those people and retrain them to help fix and solve the errors that are found and/or get into predictive analytics based on the AI and actually avoid some of these issues from coming, it doesn’t mean you need less individuals.
It just takes those individuals and applies them to higher level efforts. And that’s really where I feel as though the AI and generative AI is going to allow us as an organization and us as an industry to really push forward, not only service the clients better and provide a better experience for the associates, but actually provide more benefit for the associates because they’re going to net new skills and they’re going to be solving more complex issues and allowing those institutions to provide better service more quickly that’ll help them retain or attain a higher client base organically.
So I’m not one that buys into that: AI is at the cost of human talent and/or jobs…
What can we stress about AI and a decreased cost of service?
If done properly, once the process begins of distributing and deploying AI to existing processes, if done right, it should be self-funding. It should allow the institution to post that initial investment to then provide or have to produce less and less funds in order to… Because that should be able to be reinvested. Once you start with process A, process B to apply AI to that should be minus 20-30%. That 20-30% should be coming from the funding of the savings from process A. Again, if it’s done correctly, self-funding should become its own effort…
And that order of magnitude continues to grow and grow and grow so that the expense ratio… There’s folks saying this is going to be the largest investment in the history of technology. I don’t disagree, but that’s on a scale of the entire marketplace. That doesn’t mean each institution has to spend the most that it’s ever spent before. It’s going to spend a bit upfront, but that savings should be realized if deployed correctly to fix the biggest pain points, fix the biggest operational costs. The reinvestment should far exceed the benefits of the initial capital expenditure.
For sure, gen AI has excited our industry with its speed. Nowhere is that needed more than in keeping pace with fraud. Where do you see the best use cases in this area?
The cat and mouse game for fraud has existed as long as banking has been around, and that is not going to go away. The sophistication game, you have a very well-funded dark web. The CAGR for fraud is 23% to 25% over the next five years. Fraud is not going away. What’s happening to the institutions that are trying to combat fraud is data producing a tremendous number of alerts that’s hard for them to keep up with. Separating the noise, focusing on what’s important. So let’s take credit card fraud as an example. You have institutions that have systems.
You’ve got a Hadoop cluster with algorithms and you’re trying to look for the anomalies. You have customers calling in saying that they’re recognizing issues. You have cyber defense groups that are monitoring transactions online for some sort of potential piggybacking or break into an account or particular card. These three areas are, there’s water coming over the bottle and sometimes all three of those are getting alerts at the same time. Customers calling in, cyber defense teams are realizing that there’s been a compromise on an account and you’re getting an alerting via the proactive system.
How are the teams going to be able to bring that together? How are the teams going to understand that, “Hey, wait a minute. The projective analytics tool just alerted on Greg Kanevski’s account. And now cyber also has an alert on Greg Kanevski’s account. Oh, by the way, Greg Kanevski is calling in because he is noticing charges pop up. He’s getting alerts via his texts, his SMS messaging. How do we bring that together and how do we bring that together at scale?” AI in general, the generative AI on top of it is able to go across the streams horizontally because these are different organizational functions.
These are different operational teams. They’re on different deployments of different systems. How do I bring these together into a single event? And then what platform can I manage those on to bring these functions together at scale to combat the fraud? And there’s been an increase recently in check fraud. That’s because the teams have been doing a fairly well job of keeping up on the digital fraud. So the advent of check fraud has come back. How do I handle all of these across the different streams and then manage the next event coming through quickly because they find an ability to exercise a loophole or some sort of system issue.
So fraud with its growth is one of the top five issues we hear from every client. How do I manage my fraud, my claims and my complaints, my disputes? How do I bring those three functions together and deploy AI to help me not only minimize my losses, respond to them as number two and manage those events? And then three, assure regulatory compliance because the regulators are very, very tightly focused on us. So again, that triangle of opportunities for generative AI in this space is one of the top five we’re hearing from our customer base.
What do you think remains the biggest myth around gen AI that’s maybe holding back our industry today?
It’s too expensive or we don’t have that kind of money to start. A lot of institutions are saying, “We’re not big enough to deploy generative AI.” And I’ve said to them, “At what point do you think you’re big enough to deploy it?” “Well, we don’t have the kind of money for that. That’s just for the big folks.” And I’ve said to them, “Large or small, order of magnitude of what you tackle and how you tackle it. Small institution A is not going to match the deployment of a money center bank, but that doesn’t mean you can’t deploy it in your institution in order to help you save as you grow.”
So that myth of only the big folks can do it, and we’re not big enough and we don’t have the capital. AI investment should be commensurate with the size of your organization and that size, it does not have to be cost prohibitive for you to enter that marketplace. You just need a plan. Your C-suite has to be aligned to what that plan is. Your CIO, I mean, it takes human resources to be engaged. It takes your head of retail and commercial banking. The C-suite needs to be aligned to what the goals are, and then biting off those areas in which you want to choose and move forward.
It shouldn’t be a size thing. It should be an order of magnitude in which you deploy it. And that is the biggest myth. It’s not whether or not most folks believe blindly that AI and generative AI will help. It’s the fallacy that only the big institutions can do this, and that really is stunting some people from realizing benefits.
Greg Kanevski, Global Head of Banking for ServiceNow, many thanks for joining us on the BAI Banking Strategies podcast.
Thank you so much for having me here today. I’ve appreciated the time.
Join our community to unlock exclusive content, connect with industry experts, and gain access to valuable resources that will help you stay ahead.
