Track 4: Balancing AI Efficiencies with a Hands-on Approach

Transcription:

Ian Moloney (00:08):

I am Ian Moloney. I'm the Senior Vice President Head of Policy and Regulatory Affairs at the American FinTech Council, and I'm joined by a really, a leader in the bass space and a good friend of mine, Jesse Honigberg, who is a EVP of Product and Platforms at Customers Bank. And they are actually a member of a AFC. So it's great that we get to chat more often than just this.

(00:31):

So we're here to talk about balancing AI efficiencies with a hands-on approach, which I think AI is one of those topics that I think within Emerging Technologies has. It's a buzzword and we will try to stay away from as many buzzwords as we can. No synergies, no discussions around confluence, confluence. I mean, we're going to do our best, but I think to get into it, we really do want focus on sort of that hands-on approach, which is great that you're here. And so first off, I think it would be beneficial if you just kind gave an overview of some of the decision-making processes that you and that customers bank kind goes through when they're thinking about the Myriad of AI tools, AI ideas. What do you go through on that from?

Jesse Honigberg (01:22):

Yeah. No, it's a good question. I think that, first of all, I want to say that the lights that everyone's been complaining about is no joke, but there's a great line that I saw on the source of all great information, which would be Instagram, which is that what people don't realize is that AI is just a lot of, if statements jammed together, I probably wouldn't necessarily go as far as saying that's true, but it is some truth in it. I think that a lot of the things that vendors are calling AI today, they called machine learning yesterday. They call it conditional logic the day before that.

(02:02):

And so the tools like Chat GPT and Google Gemini and you go down the list are truly fascinating and they will be transformational. But I think that today what we're focused on at customers, and I think more broadly as bankers is how do we take these tools that seem both tremendously powerful but tremendously scary in some ways, and figure out how to apply them in ways that can be beneficial for organizations without taking on any undue risk and making sure that we can step with them slowly and thoughtfully rather than jump in with both feet and hope it goes okay. And so we've been really thinking about what are the tools and techniques that we can do that our CIO who couldn't make it here today. Nick Harris is a tremendous advocate and has really brought a level of expertise and thoughtfulness to the bank around AI. And we're exploring a bunch of different technologies that have been really impactful. And candidly, what I'd like to do is actually tell you guys how I save time in my life with it as a leader. And then we can go through all the real world examples.

Ian Moloney (03:10):

Absolutely. I mean, that's the best hands-on approach of them all. I mean, whether you're using a copilot or shoving things into chat GPT and trying to engage and summarize and make your life quicker, easier, more efficient as the title suggests, I mean, that would be really helpful.

Jesse Honigberg (03:29):

So, how many people here use Microsoft 365? Okay, that's pretty much 75%, right?

Ian Moloney (03:36):

Microsoft would be very happy about that.

Jesse Honigberg (03:37):

I'm sure they will Bill Gates and the rest of them and it would all be good. Okay. But I think that Microsoft copilot, which is an add-on for 365 is a fascinating tool. And so if you implement it correctly, and this is where you need to be thoughtful about it, but you can, it pulls together all of your emails and your chats from teams, all the stuff in your Word documents, your Excel documents, and you could be really interesting and thoughtful about it. And so one of the things that I struggle with personally is I have a pretty decent sized team and there's a lot of different things going on. I'm sure everybody else here, and how do I keep everybody aware of what everybody else is doing? And ideally, I wake up on Monday morning, I'm like, I should put together this email for my team of everything that everyone has going on the last week so that we could all be aware.

(04:31):

And inevitably we get on a team call once a week and we ramble about stuff, but there's not a structured thoughtful way that we go about it. And so a couple of weeks ago, our CIO was like, Hey, you should try doing that in copilot. And so I did and I created a bunch of different queries in copilot that consolidate the relevant material for each member of my team and then pull it all together in a single document. Now, something that I think if I was undistracted and focused probably would take me two hours. Being distracted and focused now takes me 30 minutes. And so that's a little thing that I never would've access to before and it's not transformational, but for my teams it is. And so for the people who look to us for information, the people who look to us to keep them informed, I really look at the efficiencies that AI brings are enabling us to be a better version of ourselves versus it's going to change the world.

Ian Moloney (05:34):

And I think oftentimes on AI, there is that sort of this is going to change the world and just let everything be with AI and it's a panacea. And so from a regulatory standpoint, since that's where I sit every day, I think about the different use cases and the different risks that are associated with them. And so I'm wondering from more of a banker's perspective, copilot obviously has a very different risk profile than leveraging AI within machine learning and the lending context. So I'm wondering if you can discuss how you and your team handles the varying risks associated with the different types of AI.

Jesse Honigberg (06:23):

I think that banks are nothing more than arbiters of risk, and we have to be really thoughtful about where those risks are, especially when they're emerging and how we deploy them or use them in thoughtful contained ways. And then as we get more comfortable with it, we can expand it. There's really interesting tools out there like Zest Finance, which I don't know if anyone knows, but is a really fascinating kind of recreate underwriting model. There's really interesting things happening on senses doing some interesting stuff. There's tons of different startups out there. But the way that I've thought about it with regards to the bank is how do I make my humans better? I don't want to replace my humans. I want to make, give them superpowers. And so how do I give my humans superpowers? And so I look at tools like Zest or Sincero or any of these other things to say, can I use AI to make my people feel more empowered?

(07:31):

Can I use it so that we don't need four eyes, but I can use two plus AI and probably get to the equivalent of eight eyes? And so what are the ways that I can look at compliance or I can look at existing processes and increase the reliability of them versus saying, I'm going to transform them one day. Maybe we'll transform them one day. Maybe we'll all be replaced with something that day wasn't today. And I think that when you start with the baseline of we're going to replace everything with AI, it makes it unfair to you and it makes it unfair for the people who work with you because you are undercutting the value that they bring. And so can we think about tools like Chat GPT? Can we think about tools like copilot or you look at a lot of BPM tools that are out there right now, and how do we get our people to feel empowered by them as opposed to threatened by them?

(08:28):

And the way you do that is by saying, you are good today, but what happens if I can make you great tomorrow by leveraging this tool? And I think that's where AI for banks is really the great opportunity. I think you'd have a hard time going to your regulator and saying, I'm going to replace every single person in my BSA and L group with this bot and it's 95% accurate. And I think that they'd say, that's great, but what about that other 5%? And so if you said, Hey, I'm going to take this bot or this LLM and layer it on top of a process that is 99.9% to get it to 99.99999, they'd be like, that's great. Let's do that and figure out how we scale and take the learnings from it.

Ian Moloney (09:09):

I think that really reminds me of Hannah Fry's book Hello World. And for those that have read it, it's a good book. I'm sure you'll agree and for those that you might want to check it out, but one of the.

Jesse Honigberg (09:23):

You say AFC is sending everybody a copy.

Ian Moloney (09:25):

I didn't say that.

Jesse Honigberg (09:26):

You did.

Ian Moloney (09:27):

So, don't hold me to that. But if you want one, come join AFC. And we might be able to have a conversation about it. But I think the important piece from that book was that really AI is a tool that is not replacing people like you're saying. It is helping them to be better at their jobs, to really ensure that we're terrible at recognizing patterns. AI is great at recognizing patterns. So being able to do that and being able to leverage that I think is really beneficial. And from a regulatory standpoint, yeah, they're going to be mad if you come to them and I won't have you be in the room, I'll be in the room. But if you come to them and say, Hey, we're going to replace everything with AI, they're not going to let it happen. So I guess.

Jesse Honigberg (10:13):

Or Should you? To be very candid, I mean our responsibility, if begs lose trust, we have nothing to be very candid with you, right? No one's going to put their money with someone they don't trust. And so our ability to treat that with absolute privacy is critical. And so AI is a tool to increase trust, not decrease, and that's something we have to be really thoughtful about.

Ian Moloney (10:41):

So I think the trust piece and then the really figuring out how to leverage people and ensure that they're doing what they need to be doing, those are two important pieces. How do you handle some folks, and I don't know if you've run into them, I know I've run into them that have said AI is just the solution and will get rid of people. And how do you manage those sort of challenges internally to the extent that you've seen them?

Jesse Honigberg (11:10):

I think that, does anyone here remember the first time they saw Google Earth? I don't know.

Ian Moloney (11:18):

I do. I remember.

Jesse Honigberg (11:19):

So it was this thing that blew my mind. Someone was like, you're going to put your address in and there's going to be this globe that spins and it's going to drop right down onto you. And I was like, how the hell are they doing that? Right? This is amazing. And it still is, right? And the first time I ever used A GPS, which maybe I'm dating myself, but it was something that seemed revolutionary. And those are all things that we take for granted today and we've integrated into our life. And I think that that will ultimately be what AI is maybe to an exponent, but it's not something that we say, now that we have Google Earth or now that we have GPS, we don't need to know how to go anywhere. We don't need to not visit the Grand Canyon. We could see what it looks like from an image on our computer.

(12:13):

It's something that helps us experience things in a more effective way. It helps make our lives hopefully a little bit better. And it's augmenting versus replacing. And I think anybody who tells you it's going to replace X, Y, and Z, I think there will be jobs that certainly replaces, but it's going to take some time and it's also going to be something that opens up even more opportunities that opens up more doors than it closes, I think, in the long run. And I think that as someone who's sometimes proud, sometimes not proud owner of a Tesla with buggy full great until it isn't. And so I think that while I would never let it drive itself, it makes it a lot easier when I'm tired and I'm driving on a highway, but it's not a panacea. And I think that it's the same thing for people who say it's going to be X, Y, or Z, it will be transformational. It'll make our lives better. I'm sure in some ways it'll make our lives worse, but the idea is that it augments where we are. It doesn't replace it.

Ian Moloney (13:23):

I think that that makes total sense in the analogy that you gave around GPS and Google Earth, I think those are really quite relevant because I know without GPS, I would've gotten lost from my hotel room. I'm terrible with direction. So it augmented my life and it made it easier, it made it better. And I think with AI, from what we've seen.

Jesse Honigberg (13:46):

We would've eventually made it here.

Ian Moloney (13:49):

Yeah, I schedule in wandering time whenever I go to somewhere new and it ranges between 15 and 45 minutes depending on the complexity. Perfect. Yes. These are real example. That's my hands-on approach. But I think, thinking about that aspect, there's also a democratization associated with you think of the folks that maybe couldn't get to the Grand Canyon, they can now see they can experience it in some capacity and there's that opportunity to do so by leveraging AI to do that in the lending context, in the payments context, and in many other ways in compliance as well. And I think whether it's fraud detection or other aspects, just making more efficiencies, it really will kind of on the backend from what we've seen across the industry, help to allow bankers to go back to the products that they're really working on and not spend as much money on compliance, but still have that 99.9% accuracy. So with that long-winded way of getting to this, how are regulators seeing this? What's been your experience with regulators? Because I think they're a very important piece to the puzzle.

Jesse Honigberg (15:02):

We're not doing anything today that's kind of changing models or replacing models along that line. And customers is a pretty robust model risk management approach that regardless of what we're doing, things go through. The way that we've really framed this in conversations with both our compliance folks as well as oversight is these are tools that augment process, don't replace process. These are tools that allow us to say yes when maybe we would've otherwise said no, but they're not going to supersede the box. So if we have our credit box, we still have our credit box, this is going to be an overlay to it that may be an additional detective control. I think that what I hope is a couple of years from now that it's something that allows us to be much more thoughtful about how we expand access to credit. It should be allow us to be much more thoughtful about how we give things like early wage access, how we do things that really allow people to live their life a little bit better or move up on the socioeconomic ladder.

(16:10):

But it's going to take time to get there because these tools are still very, very early and they need to be vetted. And if we remember, the hype cycle is where we are right now. And I think that if you go back four years, everything blockchain is going to change the world. And inevitably it changed a lot of things, but we're all still using credit cards and those payment methods, and maybe long-term it will, but I think we're somewhere in the hype cycle right now where it's making lives better and it's making our daily work a little bit more efficient. But I don't think any of us are ready to go to our regulators and say, we're replacing X with Y. We want to tell them that we think this can make us serve our customers even better, rather than saying, we're going to use this as a replacement for an existing process.

Ian Moloney (16:59):

That makes total sense. And I guess one thing that I know I've come across a number of folks that are piloting projects, they're running startups, they're very new in the venture. And I think just given your position, given the work that customers is doing, how when it relates to ai, when you have those conversations with some of the vendors, what are some of those? How are they going and what do you want startup folks to take away from a conversation with a bank?

Jesse Honigberg (17:33):

So I do a bunch of work with startups generally dashing their hopes and dreams when they tell me they're about to change the world, which is a little bit fun, no, but try to get a dose of reality. I think everybody thinks they have an idea that will change the world, and some people do, but there's an appreciation. There's usually a lack of appreciation of how difficult it is and all the things that we have to do to work with somebody, and candidly, the expectations that we're held to. So when I talk to startups that are thinking about doing something in AI or Hey, we are X for Y, the real thing that I ask them is, what's your value creation? So where is ai? Why is someone paying you for this? And especially with how much you think it's worth, are you generating that much equivalent value for the bank or for whoever's licensing it?

(18:36):

Right? And inevitably the answer's like, well, because we could eliminate X or we could do Y better. And I was like, well, what happens if I can't eliminate X, but I can just do Y better? How much would you charge for it? Does this still make sense? Inevitably, it gets them to start to think about what the overall value trade is of what they're doing and why it matters. Sense is one of the startups I work with is in Y Combinator right now, and I was talking to the founder, Oop, and he's coming up with this process and he's telling me about what he's doing and it's all about more effective compliance management, which is something I think we all need. And he was like, well, I'm going to replace all of the compliance people in the bank, and I was super as a friend, it's not a good sales pitch to go in and say you're going to eliminate all the compliance people if you're trying to sell it to compliance people.

(19:36):

And so we kind of helped him change his approach to say, it's a great tool, but how do we use that as a way to make the compliance people feel better about what they do, feel more confident that they can take on different types of partners and that you can expand those learnings and make them repeatable and institutionalized knowledge for the bank. That's often at smaller institutions rested in a handful of people. And so when I look at startups that are going to be successful here, they will take something that's revolutionary. And I do think that AI is fundamentally revolutionary, but they'll figure out how to turn it into something evolutionary. And so how we end up is a process, especially with banks. And so the startups that have these really compelling interesting ideas need to figure out a way to get it so that we can take it within our organizations and we can get people excited about it and not scared about it because once people are scared about it, they just shut down and these are things that should be making their lives better, not candidly looking at it. It's going to be the Terminator. Yeah.

Ian Moloney (20:46):

Yeah. I think probably all the staff in the compliance department that you mentioned, they would be scared about that, but then if they see it as what it actually is, which is a tool that they can use, then they'll be excited about it and there's a better uptake. There's the first line, second line folks can feel more confident in the work that they're doing, which is ultimately, I mean hopefully going to keep them around. And so that allows you to attract and retain better employees. So that's fantastic.

Jesse Honigberg (21:17):

I was in Omaha, Nebraska, and if you want to talk about a difference between weather between Florida and Omaha, it was about 20 degrees and 30 degree winds on Monday and Tuesday and was there visiting a client. And the client runs a pretty decent sized call center. And we were talking about this and I was saying how I'm coming down here and I'm going to talk about this with Ian. They're like, who's Ian? And I was like, oh, don't worry about it. He's a nice guy.

Ian Moloney (21:39):

No worries.

Jesse Honigberg (21:41):

But he were telling me that they just started using this new tool that integrates to their call center and it goes through the recordings that they do today and does sentiment analysis on it and then generates an automated report for them every day that kind of gives them calls that they should look into. Right? Alright, and what's the general sentiment trending? And they get calls in on different trunk lines for different clients. And so they could say there more negative calls or positive calls for clients. And that's the type of thing that we all aspire to do. Like, oh, I would love to have that if I ran a call center and have it just come to me every morning so I could see a dashboard and then it recommends calls I should look into. And that's augmenting their existing process. They had managers who would sample, listen into calls, they had agents who would identify problem calls and then they would have a procedure for escalating this takes that and makes it better and makes it scalable. And that for me is a great use case for AI and where I think ultimately will generate the most benefit.

Ian Moloney (22:38):

And I think from a research perspective as well, I mean it has great implications for the bank, they can then take that it's a generalizable sample as opposed to a non generalizable sample. There's good opportunities to, because really the employee that would be able to do a similar thing has to be there for what, a decade probably. Maybe

Jesse Honigberg (22:58):

It's also picking for a needle in a haystack. Right? Exactly. And you think about what's the story of how you're using AI with your regulator and you tell that like, oh, I'm using it to more effectively detect customer complaints. Right. Well that seems like a great use. You have a good process now. You have a great process. Yeah.

Ian Moloney (23:15):

I would wager that folks at the CFPB would be very happy to hear.

Jesse Honigberg (23:19):

I'm not making any comments about it.

Ian Moloney (23:21):

I know you aren't, but I would wager given my candid conversations with folks at the CFPB that the markets division and some of those folks would be very interested and see some of the benefits around that because ultimately it's serving the consumer in a more effective manner. So we've got a few minutes left. Do you want to kick it to questions in the audience?

Jesse Honigberg (23:44):

I'd be more than happy to answer any questions as early as possible.

Ian Moloney (23:46):

There we go. We've got one right up in the middle area.

Jesse Honigberg (23:50):

Yes, you're the lucky winner. Let's hear. Go for it.

Audience Member 1 (23:56):

Broken eye. Oh great. I think I'm loud enough. Okay. Quick question. So I mean, you're talking about some of the things that you guys are thinking of. Has customers bank delved into anything internally when it comes to some of the tooling for ai? I noticed that with the other banks that they're doing a lot of internal stuff. Are you guys playing with anything?

Jesse Honigberg (24:17):

So we are running within the cone of secrecy, right? We do run chat GPT Enterprise. And so we're one of, I think there's a couple of dozen banks right now that license chat, GPT from open ai. And we run into private tenant so that we isolated and we have a pretty regimented approach for how we give people access. And so before you get access to either copilot or chat GPT Enterprise, you need to go through training and you need to acknowledge that you understand the right use of it as well as get periodically reviewed and refreshed on that. And so the key is these are very powerful tools, but you need to understand what they are. And if I give somebody, I always joke, how often do we use a knife to try to screw a screw? Screw in, thank you. And we're sitting there like, oh, I just about to get it.

(25:20):

And then inevitably you should have just gone to the toolbox and gotten a screwdriver. And so I think we want to make sure that people don't think like this is a knife and I'm just going to try to screw everything in. If you are trying to get something that needs a screwdriver or you're trying to do deep data analysis, this is not the tool for you. Not right now. It's a great way to give you purview into things. It's a great way for you to explore information that's out there. We encourage our folks to load things like contracts that we may get from vendors and ask questions about them for certain meetings. We use Microsoft, kind of the copilot recording feature that does summarization, but we're thoughtful about when and where we do that. And so a lot of these tools we do use internally and we are exploring new and interesting ways.

(26:11):

Legal is a fascinating one, but it's also one that you need to be very thoughtful about. But the idea is to start. And so if you never start, you'll never go anywhere. And so take your first steps inevitably, hopefully you won't trip, but probably you will. But just take a small enough steps that if you trip, you stumble and you keep going forward. And I think that's the key for all of these lessons, whether it's in AI or payments or whatever else. If you try to change the world, you're probably going to be disappointed. But if you try to just make people's lives a little bit better, I think that you can have a real impact.

Ian Moloney (26:47):

Fantastic. Alright, I think we've got time for maybe one more quick question if anybody has it.

Jesse Honigberg (26:55):

Oh, come on. That hurts.

Ian Moloney (26:58):

Well, you guys can have plenty of questions for you.

Jesse Honigberg (27:01):

 I'm sure. Yeah. But I'm sure we'll let everybody.

Ian Moloney (27:02):

Yeah, they want to go to happy hours, they want to catch flights. So I think with that, it's been wonderful chatting with y'all and hope everybody has a great rest of your day.

Jesse Honigberg (27:11):

Thanks guys.