Enjoying the podcast? Don’t miss out on future episodes! Please hit that subscribe button on Apple, Spotify, or your favorite podcast platform to stay updated with our latest content. Thank you for your support!
The Generative AI hype is not dying down. But where are the use cases in fintech beyond the basic chatbot? Particularly in fintech lending where AI has been used in underwriting for more than a decade. But not Generative AI. That is why I was very interested to chat with today’s guest who has created a patent-pending Gen AI use case.
My next guest on the Fintech One-on-One Podcast is Ryan Rosset, the Co-CEO and Founder of Credibly. They are a small business lender that has been around since 2010, but today they call themselves a “data science driven lender”. We find out what this means and much more in this episode.
In this podcast you will learn:
- The founding story of Credibly.
- Why they decided very early on to take on a bank partnership.
- How they have changed as the industry has evolved over the last 14 years.
- The suite of financing products they offer today.
- How much progress has been made when it comes to access to capital for small business.
- The speed of their underwriting process.
- What they mean when they say they are a “data science driven lender.”
- How they are using their patent-pending Generative AI system in underwriting.
- The data they are feeding into this Gen AI-system.
- Other places where they are looking to leverage Gen AI.
- Why they decided to launch their own business banking product last year.
- Ryan’s perspective on banking-as-a-service and the current challenges.
- The scale they are at today.
- What is next for Credibly.
Read a transcription of our conversation below.
FINTECH ONE-ON-ONE PODCAST NO. 494 – RYAN ROSETT
Peter Renton 00:00
Welcome to the Fintech One-on-One Podcast. This is Peter Renton, Co-Founder of Fintech Nexus and now the CEO of the fintech consulting company, Renton and Co. I’ve been doing this show since 2013, which makes this the longest running one on one interview show in all of fintech. Thank you so much for joining me on this journey. Now, let’s get on with the show.
Peter Renton 00:31
Today on the show, I’m delighted to welcome Ryan Rosett, he is the Co-Founder and Co-CEO of Credibly. Now Credibly is a small business lender, really interesting company been doing this a long time since 2010. And they call themselves a data science driven lender, we actually get into what that really means. And Ryan also shares some really fascinating examples of Generative AI and how they’re using it in their company today, something I had never heard of before. We obviously talk about the challenge for small business and how fintechs have been addressing that challenge. We talk about the business banking offering, we talk about the banking as a service debacle that we’re going through right now, and much more. It was a fascinating discussion. I hope you enjoy the show.
Peter Renton 01:26
Welcome to the podcast, Ryan.
Ryan Rosett 01:27
Thanks, Peter. Glad to be here.
Peter Renton 01:29
Okay, so glad to have you. Let’s kick it off by giving listeners a little bit of background about yourself. I know, you’ve been at Credibly for some time, but why don’t you hit on some of the highlights of your career to date.
Ryan Rosett 01:42
So just to give you a little bit of background, so I started the business in 2010, with my partner, Eden King. We originally started it, we had a joint venture with a smaller factoring bank. And the reason I say that is it was sort of interesting, just because at the time in 2010, the business was sort of the wild west. And we had to sort of grow up much quicker and become compliant, because we have the FDIC in our office once a quarter. So really, that was like so the culture of compliance from a fintech standpoint was like in place intact, really early in our business venture. And I think I would say we were probably one of the first if not the first to actually have a joint venture with a bank. When you think about the time, it’s 14 years ago.
Peter Renton 02:26
Right.
Ryan Rosett 02:26
You know, fast forward, we took on a private equity partner in 2014. We did our first securitization in 2018. We just completed our third securitization about two months ago. And we continue to grow and innovate and evolve as this business continues to mature, as alternative lenders are becoming a little bit more mainstream.
Peter Renton 02:52
Well, let’s go back to 2010 or before 2010. What was the reason and the story for founding Credibly?
Ryan Rosett 03:03
Both my partner and I are lawyers by education, I never practiced, my partner did. I was in real estate development for 15 years, and then got into short term bridge loans, sort of during the Great Recession, when there was opportunity, where banks were purging assets and borrowers with income producing properties needed money to buy them back from a bank at a discount. So we were doing short term two to five million dollar loans, kind of mid-teens type yield. And it led me to this business, small business lending, which was at a contrarian time, because it was almost like this whole conversation about a double dip recession and things of that nature were out there, but the access to credit wasn’t. So small business just could not obtain capital. And so we thought this was an interesting time to start a business because there was such a need for capital as businesses needed to grow. So really, it came from a genuine place of trying to help small businesses secure capital. And we always were an on balance sheet lender. So we never started the business as a broker and converted to an on balance sheet. We started as on balance sheet and tried to, for better or worse, figure it out as we went along and learn.
Peter Renton 04:28
And so did you put that that bank partnership in place from day one? Or when did that bank partnership start?
Ryan Rosett 04:34
Yeah, so it was about nine months after we started. So we started and we had our own money in the business. We raised a few funds to finance the business. And I recall very, you know, clearly we had the small office in downtown Birmingham, Michigan, and I remember having this bank come into our office. And we were too young of a company to forge a relationship like this, but they really liked the idea and the concept and us. And we were seeing it took a long time to negotiate. But you know, we were successful in negotiating this joint venture. And it proved to be like a really interesting funding source for us to grow the business, and it was really profitable venture for the bank as well.
05:26
Right,
Peter Renton 05:26
Right, right. Because I imagine the bank probably wasn’t able to — a lot of banks pulled back from small business lending, right? Particularly during the financial crisis? So how long did that bank relationship last, or is it still in place today?
Ryan Rosett 05:39
It lasted through 2014. Really, what happened is we were a single credit for them. So we created an SPV as a joint venture between the two and we became too large of a concentration risk for them. They were like a half a billion dollar bank. I think at the at the time, I think we grew it to about $25 million. And they said, listen, with the returns, they were getting there, they were taking like 12 or 13% return on average assets, which a typical bank, I think makes one or two percent return on average assets. So it was a very healthy return for them. They actually really liked the relationship. They just they said, “Hey, you have one year, go find senior credit facility, take us out, we love your relationship, we’ll be an advocate for you. But, you know, this is something that we need to do for the business”.
Peter Renton 06:29
Okay. And so you replace them with obviously other funding sources?
Peter Renton 06:34
Correct?
Ryan Rosett 06:34
Correct. Correct.
Peter Renton 06:36
Okay. Well, let’s maybe talk about, you know, you were very early on, before fintech was a thing, and then there were others out there, like OnDeck and Kabbage where they slightly predated you, but fintech lending wasn’t a thing. But from doing this for 14 years, and how you operated back then versus how you’re operating today, tell us some of the evolutions that your company has undergone as the industry has kind of changed dramatically in the last 14 years.
Ryan Rosett 07:08
I mean, let me say that the evolution has been, if I can go back to when we originally started, we really were only doing split processing. So, you know, it was a merchant cash advance company, we were the first company to partner with First Data at the time; it was First Data before it was Fiserv. And we were doing split fundings. And so we were really targeting the First Data customers. I remember very clearly, you know, looking at my partner and saying, “Why does everyone collect bank statements?” We didn’t understand that. So what sounds elementary, for what we were doing at the time, it was, like, an interesting time in the business, because we were learning as we went.
Peter Renton 07:56
Right.
Ryan Rosett 07:56
And now, when you look at the landscape of how advanced we’ve become, and how much automation we have in place, to handle the flow that we see, everything in the beginning was obviously underwritten by hand. It was all keyed in, you know. Once we understood how bank statements played a role, those were all spread into our own, you know, kind of makeshift Excel spreadsheet, which is the way we started until we evolved to create our own loan origination system, because there was no off-the-shelf loan origination system that worked for our business, to then moving to our next generation loan origination system that we created, and are still on, and, you know, continue to invest heavily into. So just the technology automation, where we started to where we are today, we’re two completely different businesses. From a scale perspective, from, I would say sophistication and staffing to how our applications actually work, how the automation works. And then, you know, I’m sure we’ll touch on Generative AI, but like, even today where we’re starting to use real scenarios around Generative AI that are extremely helpful, and we have a number of other kind of initiatives that we’re working on that I think are gonna just continue to help us make for a better customer experience when they interface with us.
Peter Renton 09:31
Right, right. I do want to talk about Generative AI, but I’m going to leave that to later in the conversation. Maybe you could just talk about what you offer today. Do you still do MCA? What is the the suite of products that you’re offering?
Peter Renton 09:44
We to
Ryan Rosett 09:45
Forty percent of our business is MCAs, 60% is loans. You know, we do business in all 50 states, just depending on the states that where we offer MCA versus loans, but that there too, it’s all automated within our system. So whatever product that we have available, we offer to the customer if it’s available in the state. Again, each state has their own guidelines, as I’m sure you’re well aware of. And so staying very current from a compliance standpoint is key to this business, because it’s something that you can’t fall behind. And if you really want to stay as a top alternative lender, which we believe we are, those are the things that you need to do in order to just continue to compete.
Peter Renton 10:34
So then I want to sort of step back for a second. Because, you know, I feel like I’ve been having this similar conversation, really, since I started doing the podcast 11 years ago, with a lot of different small business lenders. There’s a lack of access to capital for small business, that is a given, we know that that happens. But that was true in 2014. And it’s still true in 2024, it seems. And fintech has really, I think, attacked the problem fairly aggressively. And I think we’ve done a fairly good job. But I’d love to get your perspective of how much progress you think has actually been made for the typical small business owner who is still struggling with access to capital?
Ryan Rosett 11:17
So progress, you know, here’s the reality and why I think this business exists, in my opinion, is that our average loan size kind of hovers around $70,000. If you think about it from a bank perspective, banks aren’t equipped to lend $70,000 to a small business. So while they can, and will, the economics associated with it, versus the risk of actually lending $70,000 I would say just the risk outweighs the benefit. So in scale, I think, you know, when you look at the businesses across and why alternative lenders exist, I just don’t think banks choose to compete in this environment right now. Nor do I believe they’re equipped to. And when you think about also, the macroeconomic headwinds that come, in various stages over a decade, banks tend to pull back, and then they’ll go a little bit too long, in a certain lending, and then they pull back, and then the pull backs are usually like, longer periods of time, where we can adapt in hours, you know, where if there’s a forest fire or a, you know, a natural disaster, we have the ability to really just kind of adapt in real time, where like, if we need to do a site inspection, or whatever the case may be, if it’s in some area that’s been affected, those are things that we can do very quickly. I just don’t believe banks today are equipped, nor do they have, you know, I would say the interest to enter into this. I would say, it’s a tricky lending space, because you are underwriting an entity. Yes, you’re underwriting a small business owner. But at the end of the day, it’s shielded by an entity. And really what we’re lending against is cashflow. So where and that has like been part of our evolution as well. So we are, you know, I would say 100% cashflow lenders. So we’re looking at average daily balances, we’re looking at negative days in the bank, how many NSFs. And then we lay in some credit profile as well of this small business owner, to ensure how we can risk base price a specific merchant. So if you’re a high quality merchant that’s been in business for 15 years, and you have good credit, you’re gonna get great pricing from us. And if you’re a six month business, that has spotty credit, you know, we’re gonna price that accordingly as well to accept the risks that we may take on.
14:00
So
Peter Renton 14:00
So then would you say that, in particular I’m talking about the sub $250K segment, as far as loan size goes, is it fair to say that the small business owners are better served today than ever before?
Ryan Rosett 14:12
You know, I think they do have options. I mean, there’s no doubt that we are more expensive than a bank. We borrow money from, you know, we have asset backed securitizations, we have senior credit facilities from banks, so no doubt, if they can access a bank, anyone who asks me, they should. However, if you need your money, tomorrow, we are a great resource for that or you need your money in two days, or whatever the case may be. I think we’re an excellent resource for that. So I do believe that small businesses have the ability to access credit much easier today than say a decade ago.
Peter Renton 14:53
So just take us through the underwriting process. How much of it is automated? How long does it take between application and approval?
Ryan Rosett 15:05
So we have a few different channels. We have a direct sales channel, we have people who find us online, they can go through an online experience, which has full automation. So if it’s coming through a wholesale channel, typically from application to decision, so if it’s a decline, it can be seconds. We have a bunch of models running, and if it’s declined, it happens almost automatically. If there’s an approval, usually like two and a half to three hours from app to offer. And then depending on if there’s any conditions to fund, which sometimes it might be just like driver’s license, voided check, it can fund, you know, in an hour after. So, and if they’re eligible for, you know, what we call an online checkout, that’s a 10 minute process, and their funds are deposited into their bank account. So it just depends on how complex or hairy the deal is. From an underwriting perspective, does it have multiple owners, is there any issues that pop up in the underwrite that potentially cause further questions that need to be identified? Those are where it takes a little longer, or we may require additional documentation that an underwriter needs to review. But we are, you know, from from a, I would say, a happy path deal, where it’s clean, and there’s really no unanswered questions, it could be a 10 minute process.
Peter Renton 16:40
You’ve called yourself a data science driven lender. So maybe you could sort of give us a little bit of color there. How are you using data science in your underwriting?
Ryan Rosett 16:52
So we use data on everything we do. We’re using data from the origination perspective, like from marketing, all the way through portfolio management funding, and certainly, on the collection side. When you look at, obviously, our scoring models, and how we pull different levers that we need, if for example, if we see certain industries that are underperforming, we may pull back, where the term may be shortened, the rate may increase a touch to cover any potential losses so that we can make up, those are examples where, you know, the data that we’re looking at is very vast, and, you know, now we have several billion dollars of collection data. So we know what’s happening on industry type seasonality, and we’re able to price for it better. So, you know, I think with the data that we have, it allows us to be more aggressive where we want to be more aggressive, and more conservative where we need to be more conservative. It’s never ending. So it’s like it’s a constant change, because industries change. For example, trucking was amazing during the pandemic.
Peter Renton 18:16
Right.
Ryan Rosett 18:17
Okay? And then until it wasn’t, so plenty of companies got caught, stuck with a lot of trucking in their portfolio. We were fortunate that we did have trucking in our portfolio, and it took a while to, you know, kind of get it out. But that’s an example where an industry kind of flipped upside down. And you can say the same about construction, you know, with mortgage rates, where they are. So there’s some macro trends that we look at. And then certainly we look at data that happens within the portfolio to make better decisioning.
18:53
Okay,
Peter Renton 18:53
Okay, so then let’s segue from there to Generative AI. I read that you recently applied for a patent, around GenAI powered underwriting. Now, I’m curious about that, because obviously, AI has been used in underwriting in fintech lending for a decade or more, and I haven’t seen much in the way of Generative AI and when you talk to some of the other fintech lenders, they don’t really talk much about it. So maybe explain what the patent is and what you’re doing exactly.
Ryan Rosett 19:22
We have a few patents that are pending. One specifically I’ll get into is if you look at, for example, NAICS code vs the SIC code, which is like basically an industry classification and there’s over you know, a couple of thousand industry classifications. And so, what historically, what we’ve done is we’d have an underwriter select the industry type, okay? And if I use an example lik Peters Electrical LLC. Okay? Are you an electrician? Are you an electrical supplier? Do you manufacture electric switches? Don’t know. And so we’re historically relying on the underwriter to put in whatever code they determine. What we’ve done with our provisional patent that’s pending right now is it allows us to really go through a workflow, leveraging large language models, and identifying through a waterfall, what is the best NAICS code by prompting questions for this specific business. And we found a 25% lift of accuracy over humans on the selection of the correct industry by using this Generative AI tool. So that’s just an example. Not only does it save us time, we get more accurate. And then when we’re more accurate, it removes some of the subjectivity. And it makes it more objective where we can be, I would say better at calling risk on certain businesses that otherwise we’re making mistakes on because we didn’t know if you were an electrician or a supplier. And there’s a huge difference between that in terms of how we underwrite files.
Peter Renton 21:30
Right. That’s really interesting. So then the data that’s being generated here, the data that the large language model is using, is that, like, who’s answering those questions? You said that there’s certain questions that go up. Is that the borrower that is answering those questions? Or how does that work?
Peter Renton 21:48
Know,
Ryan Rosett 21:48
No. The way that we’re creating this workflow is it goes through a pattern of questions, and then it comes up with a ranking, and then it reranks it and then reranks it so that we get the best answer that comes out of the GenAI. For the large language models we’re using, so it has nothing to do with the customer whatsoever. We’re not looking for them to supply us any information. We are looking to identify the accuracy of the industry code.
Peter Renton 22:18
So you’re using external data sources for that then, right?
Ryan Rosett 22:22
Yes.
Peter Renton 22:22
I’d love to just get a little bit more color. This is completely new to me, and I think it’s a fascinating thing. I think the audience will find it interesting. And I presume you’re doing this in a way that it’s a closed system, right? You’re not going out and waiting for a response from someone. Maybe you could talk a little bit about the data that’s being fed into this.
Ryan Rosett 22:46
This is an example of like,
Ryan Rosett 22:48
We have a lot of information about a business. Okay? So we may know, for example, Credibly, it’s like we are located in Southfield, Michigan at this address. What do we do? You know, so some companies might say we’re a financial services, some may say, we’re a fintech, some may say we’re a data driven something. Where do we fit into the couple 1000 codes. And so that is what it, once we have a bunch of information about our customer that gets fed in as the same way. If you’re looking for an example, a restaurant review using Google, you’ll put in XYZ restaurant in Detroit, Michigan, and it will then come up with that restaurant, and then you can look at the reviews or what it looks like inside or whatever the case may be. We’re using a very similar way of identifying the customer to get more accuracy around what exactly they do without using humans. Which, say this process took us 12 minutes to do on average for a typical underwriter, but when you take 12 minutes across 1000s of applications, it’s meaningful from a time perspective, it creates a better customer experience, and we get a more accurate answer, which is more important to us. Because if we get a more accurate answer, we can have what we deem to be better probability of default, and more profitable business coming through our business.
Peter Renton 24:30
Would you price loans differently, given the different industry classification? Because some industries are just more risky than others?
Ryan Rosett 24:39
Of course, 100%.
Peter Renton 24:40
That’s where the advantage lies, right? You now know, with better accuracy, obviously, it’s probably not gonna be 100% but you’re doing a lot better than before and so now you have much more confidence that the industry that you’re getting is correct.
Ryan Rosett 24:54
That’s exactly right. And I said it before but it creates a more objective, determinative answer that we can have more confidence in than before. When I say even in our 25% lift, is getting better as it’s continuing to improve. I mean, it still has machine learning pieces to it. So it’s getting more and more accurate, which creates more accuracy around our ability to call risk.
Peter Renton 25:19
Any other GenAI and initiatives you’re willing to share at this time?
Peter Renton 25:23
We
Ryan Rosett 25:24
We have a few on the bank parsing side. And we also have a few on the marketing that we’re working on as well. But, you know, at this time, that’s the one that’s been installed, we’ve been leveraging it for maybe three months now. And, you know, again, it’s Generative AI, it’s new. So, you know, we’re learning. I am certainly, you know, not a Generative AI expert, but I find it fascinating. And I think that there’s just a world of opportunity that we are exploring, leveraging in our business, and I think we’re a really good business to use it.
Peter Renton 26:03
Okay, so I want to talk about banking, and specifically business banking, because that’s something that I saw you launched last year. What was the thinking behind that? And can you tell us about the traction that you’re getting there?
Ryan Rosett 26:18
We see, any given day, it could be 1500-2000 applications a day, and about a third of them don’t have business banking accounts. And we don’t lend to a small business that doesn’t have a commercial bank account. We’ll lend to partners we’ll lend to sole proprietors, but they have to have a business banking account. Otherwise, it would be deemed as a consumer loan, and then you have the CFPB, and you have all the other criteria that you have to follow. Yeah. So you know, we’re a small business lender. What we’ve done now is, those 1/3 of applications that come through that get rejected, we can give them an option to sign up for a business banking account. We have a partnership with Green Dot, it’s a white label solution for us. We’re not a bank. But we do get visibility into the bank accounts that are signed up. The traction has been growing. And the take rate has been, I would say, what we expected. It’s a growing number; we do have a share or interchange fee. We don’t view it as a huge profit center for us. But if we can bring a customer in, nurture that customer track their volume, and then proactively push an offer out to them that you’ve been pre-approved for? That’s the interesting part of the relationship that we have on the business banking side.
Peter Renton 28:04
Right, right. That’s interesting. And I should should clarify, we’re recording this on June 28, and the banking as a service space is under duress, shall we say? There’s been some major challenges with some of the leading banks that have been doing this for a long time. You’re using banking as a service for your for your business banking? You’ve obviously been following this? It’s been like this dumpster fire that has been slow moving over the last several weeks. But do you have concerns or what are your thoughts on the impact of relationships like the one you have with what’s been going on?
Ryan Rosett 28:45
So I think banking as a service for companies that have partners with banks, and they’re offering banking services to others white label, but they’re not a bank themselves. I think those businesses are problematic. You know, when you look at banks that are white labeling services that they already offer, and Green Dot, for example, they do the banking for Walmart, you mentioned Intuit, among many other large players that they’re sort of the bank behind the curtain. And so I’m not as concerned about their business model as some of the other neo banks that pop up and don’t necessarily have, I’ll say the balance sheet or support or the insurance that everybody’s looking at. So, for example, that is why we use Green Dot and I’m not opposed to saying Green Dot, even though it’s a white label solution. At the end of the day, we’re not a bank. And so if a small business opens up a checking account, I want them to know that the checking account’s with a real lending institution that has a balance sheet and it’s publicly traded, and the FDIC insures their deposits. So I put myself in the shoes of a small business owner, I, too, would want that. So, that’s the way I sort of look at it. We’ve never found it to be overly interesting to be lending as a service. That’s not what we think is a very interesting business model. And banking as a service, also, depending on the business itself wasn’t overly attractive to us, and I know that there’s been a lot of turmoil. I think it can work on some level. And I think, you know, Green Dot’s an example where I think they do make it work.
Peter Renton 30:44
Okay, so can you give us a sense of the scale you guys are at today? Can you give us any kind of metrics there?
Ryan Rosett 30:51
Yeah, this year, we’ll fund our average loan size kind of hovers at $70,000. This year we should do between $450 and $500 million of origination. This year, we, by design, kind of slowed our growth a little. And when I say that we, we kind of cut off the lower tiers of our four different bands. We have an A, B, C, and D band. We cut the C and D band off about nine months ago.
Peter Renton 31:25
Wow.
Ryan Rosett 31:26
And that was, I don’t know, 25-30% of our business, we have picked it up in in B’s and C’s. But in the same token, we saw performance issues with the C’s and the D’s. And we brought back C, the C band. But there was a period of time where small businesses, and I still would say, I don’t know if they’re still in a recession or not. But, clearly, I think small businesses were in a recession. Even though the general economy wasn’t and I think small businesses were experiencing a lot. Because, you look at inflation and things of that nature, it was affecting small businesses and their profitability.
Peter Renton 32:10
Okay, so last question, then what’s next for Credibly? What are you looking forward to for the last half of the year and into 2025?
Ryan Rosett 32:18
Yeah, we continue to innovate around our technology. Like I mentioned, this online checkout is something that’s new for us. We just came out of beta. And the reception has been really strong. We think it’s sort of best in class, you know, we’re the newest to do it. So we had the benefit of seeing what others have done, what worked well, what hasn’t, and tried to take the best from everyone and kind of build our own that really works well and protects us from a risk standpoint. We talked about Generative AI. I think if we can continue to scale the business, and not add more heads, I think that would be a huge win for us, and also allow us to manage our risk better. So we have a philosophy, like we’re not trying to do a million things. We wanted to just be really good at what we do. And we think we are pretty good at it. So you know, it’s not as if we’re trying to offer all different types of products like banking as a service, or this or that, like we just, we want to be very good working capital lenders. And I think we’re doing a good job at that.
Peter Renton 33:32
Right. Well, it’s a good note to end on. Ryan, it’s always great to chat with you. Thank you so much for coming on the show today.
Ryan Rosett 33:39
Thank you, Peter. I really appreciate it.
Peter Renton 33:42
Well, I hope you enjoyed the show. Thank you so much for listening. Please go ahead and give the show a review on the podcast platform of your choice and go tell your friends and colleagues about it. Anyway, on that note, I will sign off. I very much appreciate you listening, and I’ll catch you next time. Bye.