Transcript:
Penny Crosman (00:03):
Welcome to the American Banker Podcast. I'm Penny Crosman. Over the past year, Grasshopper Bank, a New York-based digital bank that is due to be acquired by Enova later this year, has been applying AI to internal workflows and its leaders say it's been speeding up processes without reducing headcount. Leaders say AI has absorbed prep work at scale, saving thousands of hours per month, while employees remain focused on decisions and exceptions. Here with us to talk about this is chief technology officer at Grasshopper Peter Chapman. Welcome, Peter.
Peter Chapman (00:37):
Penny, thank you. Thanks for having me. Looking forward to the conversation today.
Penny Crosman (00:41):
Sure. Thanks for coming. So some people, I think people in your bank even, have described AI as a confident intern. Do you agree with that description and is this a good thing?
Peter Chapman (00:53):
Yeah, I do agree with that description. It's funny. Two years ago when we first started on this AI journey, the way I described AI to the rest of the executive team and my initial pitches on why this was important was this is the best intern that you could ever have. And by the way, you don't have to pay them and they don't sleep. They will just go after a problem that you assign to them and go, go, go on it. So I think that is a fair moniker to apply out to AI more broadly. Now, are we doing some things here that are maybe a little bit more advanced in the intern now? Potentially, yes. But I think that's a fair term broadly to apply to AI.
Penny Crosman (01:36):
Sure. And what are some of the things you've been doing with AI just broadly?
Peter Chapman (01:41):
Yeah. To answer that question, let me share with you just a little bit about how we've gotten to where we are today, because I think that helps with talking about the use cases. So as mentioned earlier, we've been on this AI journey since the beginning of 2024. If you recall back in 2024, which is an eon ago when it comes to AI ages, we were all month to month saying, "Oh, this model's better than this model." And Gemini wasn't even really out in the market and OpenAI had a huge lead. And so things have certainly changed since then. But back in January 2024, we said, "Okay, we've got to get really serious about this. We got to find a way to deploy AI. We got to find a strategy around AI." And so we spent most of 2024 and probably the first half of 2025 putting together a framework and deploying AI out in the bank and listening to how people are listening to feedback from people that are leveraging AI.
(02:43):
So long story short with that, we are a big Google shop here. No surprise that in the early stages we decided on leveraging Gemini and then eventually Gemini Enterprise as our core AI model and deploying that thing out in that first year, year and a half or so. And so during that first year and a half, our focus was really on making sure that we figure out how to use it, learn a lot, education, training, communication, and of course the entire risk management side of this thing as well. So making sure there's proper infosec controls around it, making sure that there's proper risk management practices around this, et cetera. And so we spent a year and a half doing that. And I would say most of that time, or as I look back at the use cases and the value that we got out of that period, a lot of that was use cases that were centered around, let's call it standard productivity.
(03:41):
So using AI to help create a document, to review a document, to do emails, to stand up a PowerPoint, whatever that thing may be, Gemini Enterprise is baked in really nicely into the Google Workspace ecosystem, which we leverage heavily too. So that was great. And I think we got a lot of value out of that. But then starting mid last year into where we are today, I think we've stepped into a new phase of our AI adoption, and that phase is more focused on agents. And so what we have done is we have moved from standard Gemini deployment out across our enterprise into Gemini Enterprise, which is an enterprise deployment of Gemini that we've dubbed Hopper. We've rebranded the thing called Hopper, which we think is a nice play on our name. And so Hopper allows us to do a couple of different things that we think will start to take us to that next level of leveraging AI.
(04:42):
The first thing it does is it allows us to connect Hopper into other third-party data sources that are not just Google native data sources. And so as we all know, AI is really only as good as the underlying data that it has access to. So that's one really important thing that Hopper allows us to do. A second important thing that Hopper allows us to do is it has the ability to actually perform functions on behalf of the user. And so this is, what does agents mean? A lot of people will say very different things, but when I think agents, I'm thinking this AI, I give it some sort of direction and it's able to go perform an operation on my behalf. And Gemini Enterprise, a.k.a. Hopper, has the ability to do so. So I think those two broader capabilities have now taken us to a spot where we've moved on from leveraging vanilla Gemini, vanilla AI for standard productivity gains, which I'm not dismissing those, those are important, but more into new territories.
(05:49):
And I think there's probably two or three really big buckets of new territories that we're exploring with these agents, with agentic AI. One of them is hyper-efficient data access. So can we use Hopper in a natural language query to go out and pull from our data lake information about a client, information about a portfolio, et cetera, as opposed to a standard banking model where you'll put in a request with the data team and they'll get back to you in 10 to 15 days. So something like that. And then I think the other bucket is operational workflow replacement, which is just the standard agent type of capability. So leveraging it from everything from compliance workflows to reconciliation to a little bit of everything in between. So anyways, Penny, that was an extremely long answer to your question, but hopefully that helps.
Penny Crosman (06:46):
That was great. Yeah, it brings a number of other questions to mind. But one is to that last part about using it in operational workflows. Does this touch on the SaaSpocalypse that people have been talking about where AI is going to take the place of traditional software that's in the past has handled many kinds of workflows? Do you see that happening as you deploy this more in your operations?
Peter Chapman (07:17):
I think you can't ignore it. Now, I am not a doomer one way or the other on AI. I don't think the Terminator series is going to happen, and I don't think it's a bubble, and I probably don't think the SaaS apocalypse is going to happen either, but I think there's going to be a massive impact. And we're seeing this in real terms here today at Grasshopper Bank. So obviously when you're talking about SaaSpocalypse, you're talking about that Citrini Research paper that came out, what was it, a week ago, once again, ancient history and AI world. But one week ago, they put that thing out. Two and a half weeks ago, we actually saw this happen at our bank. We've got, and I can dive more deeply into this, but we've got these different teams we bring together to get feedback around AI, what could we use it for, et cetera.
(08:08):
And so we source ideas. And so one of the ideas came up that we've got enhanced due diligence requirements for some of our customers as we onboard them and some of our higher volume businesses. And so that process looks like we've got a person who goes and we'll pull the application and we'll go out into OFAC lists, government watchlists, et cetera, all sorts of different external data sources and basically write up a report and make a decision on should we onboard this person or not? Standard enhanced due diligence. So pretty standard stuff for a financial institution. And as we looked at our budget for this year, we had actually looked at a vendor to essentially automate that workflow for us. Now, once again, it was the automation of the creation of that report, not the full automation of the decision. That's a very important distinction for us.
(09:02):
It's getting the report in front of that person so they can quickly make that decision. We have a big human in the loop thing here. And so we ended up not going with that vendor. Let's say conservatively, that vendor would've cost us somewhere six digits annually, but we brought that idea to this AI group that we have, this internal working group. And in the last three weeks, we were able to very quickly stand up a working model using Hopper to be able to do exactly that. So we were able to go from idea to delivery in under three weeks and we were able to bypass going out to a vendor, which would be the traditional way of going about and doing this and saving ourselves that 100K or whatever it may be, and essentially deliver this thing internally. So I think there's something real about that SaaS apocalypse.
(09:54):
I think that report was working towards maybe a conclusion they've wanted to work towards. I think it didn't talk about potentially any of the other efficiencies or gains or new opportunities in the market, and I'll let the people way smarter than me figure that stuff out. But I think there's no doubt about it that what you see from the larger AI models right now and the agent design capability they have should have people thinking about, "Hey, could I leverage what I have here internally to stand up an agent quickly versus going out in the market and buying something or building something from a more traditional development perspective?"
Penny Crosman (10:33):
That's very interesting. Yeah, that story does give traditional software vendors something to think about for sure. So one thing I've heard about you and Grasshopper Bank is that you guys asked everyone in the bank to come up with some way that they could use AI and kind of report back on what they're using it for. Is that true?
Peter Chapman (10:55):
It is. We have very much asked a lot of the bank to give us feedback. We try to be a bank that doesn't just build cool products or technology for the sake of building cool products or technology. We like building cool products and cool technology, but there has to be some sort of reason for it. And so we partner very, very closely with the business, with operations, with the rest of the bank to make sure that what we're doing is actually meaningful. And so we've built out, as I hinted at earlier, what we call our AI cross-functional team. We're a relatively small bank. We've got about 160 people or so, and we've got about 20 people that meet every other week, come together, and we've got probably half of those folks are our data and technology folks, and half of those folks are business folks.
(11:43):
And those business folks represent every single business line, every single shared service as well. And we tried to pull into that group from the business side and from the shared services, what we dubbed AI champions. And these are people that we know that are leveraging AI a lot in their daily workflows. And so it's their job to come to these AI meetings, share ideas that they've sourced from their teams to get us thinking about how could we leverage agents and/or Hopper and/or other AI capabilities to really do something meaningful for the business. So yeah, I think the net of it is, Penny, that everyone at the bank is certainly tasked with providing feedback to this group to make sure that we're working on the right stuff and delivering stuff back to the business that provides value.
Penny Crosman (12:34):
I've heard of companies going even farther. One company that makes people report every week their status on what they're using AI for. I've heard of companies monitoring employees' use of AI, and in some cases, even determining who's going to get laid off by how much people use AI. Would you ever go to any of those extremes?
Peter Chapman (12:58):
We would not. We have tried to build a culture here that reduces fear around AI adoption. And I think if you start implementing things like that, all it's going to do is make people anxious and make people not want to be involved. And so we've taken a very different approach where back from the beginning of our AI journey, we said broadly to everyone, "Hey, this isn't here to replace you. This is here to augment you. " And look, we all in our day jobs have things that we do or things that we've traditionally done that are routine, that are boring, that we don't like. And so we kind of set out to the bank and culturally we've said, "Help us get rid of those parts of your job. Help us think about ways to automate away those parts so you can spend more of your time using your brain and doing some of the more higher level things that happen in your regular nine to five." So we certainly have the ability to, and we do look at across the company, broader AI adoption, team AI adoption, et cetera.
(14:10):
But we try to do that in an optimistic way where we're going to teams and saying, "Hey, is there a way that we can help you? " Less, "Hey, you must report in this Friday: what are the nine things that you did with AI or else?" So there's just something culturally here where we've tried to approach it a very different way, and we think that has given us really good positive results. One of the other things we're proud of when we think about who's leveraging AI here at Grasshopper is we do go look at some of those reports and then I can see who are the top 10% of users. And we go look at the top 10% of our users, 70% of those users are individual contributors. 20% of them are managers, 10% of them are on the executive team. So I think that's really important too, is we've tried to, once again, culturally deploy this thing and saying, "This is a tool, this is a framework that we want to provide to you. We want to use it, give us feedback. Is it helping you? Is it not helping you?"
And we think that helps not just with adoption and fear, because there shouldn't be any fear around this. We should be excited about this thing, but it also helps with the growing massive canyon that is happening out there in businesses and enterprises beyond ours where executives expect something out of AI and there's just a massive gap between what they expect and what the people are actually leveraging it, see and experience day in, day out. And so we think that there's not that gap here at Grasshopper because we got people that are closer to the work, engaging with it, using it and coming up with ideas and feeding it back to all of us so we can work on those ideas and ultimately make them successful.
Penny Crosman (15:58):
Jack Dorsey at Block recently announced a layoff of about 40% of the company staff and he cited efficiencies due to the use of AI. What was your reaction to that?
Peter Chapman (16:13):
There's a cynical reaction that I had to it right away and then a, okay reaction. And I think the cynical reaction that there's going to be companies out there that are going to be using AI as cover for doing some sort of downsizing. It's just going to happen. I'm sure it is happening. Now, do I know if this happened specifically in their scenario? I don't know. What I do know is that over the last three years, they staffed up massively and now they've cut down massively. So I don't know, maybe it's a reaction, maybe it isn't. I have no idea what's going on there, but I think there's something real in this. I have no doubt that there's something real and there's probably something broadly for overstaffed companies that they're going to have to deal with it.
(17:00):
Now, thankfully, we're not in that scenario. We actually, I love working at Grasshopper and I love that AI came about right when it came about because we are a young, hungry, growing company that is scaling. And it's once again, going back to our culture here, it's far easier for us to say, hey, this isn't about we're going to reduce our workforce by 40%. We only have 160 people. This is about us scaling and maybe not having to add headcount the way a traditional bank would who is inefficient. No.
Penny Crosman (17:35):
That makes sense. So of all the ways that you're using AI throughout the organization, what are some of the tangible or measurable results you've seen so far?
Peter Chapman (17:48):
Yeah, so there's a bunch of them. Going back to the journey last year, we had a goal that we set for ourselves and our goal was that we were going to implement essentially AI models that saved us roughly 2,000 hours a month. And so we had ways of calculating that were unsurprisingly imperfect ways of calculating that. But essentially we calculated how many prompts someone was leveraging an AI, what was the actual use case? We have integrations for Gemini into certain different applications within Google Workspace. So we had some fuzzy math that we were using. We blew that goal out of the water last year and we blew it out of the water by July. So our annual goal, we had beaten and beaten pretty quickly. And so once again, those were imperfect measurements of what I called earlier, kind of standard productivity use cases.
(18:53):
So I'm in Gmail, I need help managing my inbox or creating an email. I'm in Google Drive and I need help standing up a Google Sheet and connecting it into something else. I need to do some research, so I'm going to leverage Gemini to go out and do some research or tile that stuff in together. So those have been great. Those have been really good standard productivity use cases, and those aren't going away, and I don't want to dismiss those. Those things are real though. But since then, we have started to stand up some other things as mentioned. So I already talked about the enhanced due diligence for client onboarding. That's very real. We have also, for a longer time, had an enhanced due diligence process that is somewhat automated by a partner who leverages GenAI for existing corporate clients. So that's kind of one area.
(19:46):
Another area that we've spent a lot of time on working on and we're still continuing to perfect is how do we leverage GenAI to do ongoing compliance monitoring in our lending portfolio? And then how do we leverage GenAI to perform annual reviews and ultimately a more efficient credit memo for that lending portfolio as well. So I think those are some really specific use cases that we have spent a lot of time looking at, focusing on, et cetera. And then back to what I mentioned earlier, which is what I'm really incredibly excited about is the data access question. And I think we're about 30 days from going live with our initial connection from our Hopper AI system into our Google data lake, and I think that's going to net some massive efficiencies across the board for the bank and help us make better decisions for our clients.
Penny Crosman (20:45):
Making it easier for people to find information that might be buried in some reports somewhere or something like that. That's right. Well, a lot of those do sound like things that people don't necessarily enjoy doing as part of their job. So if AI can get beyond the confident intern, we were talking about at the beginning, what does it become? Does it become a mid-level career banker or what would you call it?
Peter Chapman (21:13):
That's a really good question. And I think the tighter, the more specialized you get and more focused you get on a specific task, maybe the higher up we could rank that thing. It could become a professor. So I think the tighter you get that thing, the more capability that it has. But I think that it is certainly graduated in certain pockets here at Grasshopper from being just a really good intern who can do very menial tasks into what I'd probably call a valued early career colleague. Maybe that's what I'll call it.
Penny Crosman (21:58):
Okay, fair enough. Great. Well, Peter Chapman, thank you so much for joining us today. And to all of you, thank you for listening to the American Banker Podcast. I produced this episode with audio production by Adnan Khan, WenWyst Jeanmary and Anna Mints. Special thanks this week to Peter Chapman at Grasshopper Bank. 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.
