Innovations Driving Small Business Lending Forward: It’s Not All About AI

Peter Renton 00:05

Hi everybody. My name is Peter Renton. I am the founder of Renton and Co., where I’m a consulting firm focused on FinTech and events. I’ve been doing FinTech events for 13 plus almost 14 years now, and I have been an enthusiast of FinTech for the last couple of decades. And I’m also a lifelong small business owner. I’ve only ever worked in a small business. My father was a small business owner before me, and he, he gave me the bug. And I don’t know what it’s like to have a real job. I’ve never had one, but I do love the you know what, what they were talking about, the freedom that it gives the ability to have much, much more control of your of your time and and often finances. But I’m saying that I’m also I’m a FinTech enthusiast who’s a small business owner, so small business lending is like this great blend of all of my interests, and it’s always been something that I’ve been personally very fascinated about and innovation in small business lending, which is what this this panel is about innovations driving small business lending forward. It’s that that is something that is very near and dear to my heart. So with that, I would like to hand it over to the panel to do quick introductions.

Malika Anand  01:39

Malika, sure. Hi. I’m Malika. I serve as Director of Impact at Community Investment Management. You met Jacob earlier, but we are an Impact Manager backing responsible lenders, most of them tech enabled fintechs here in North America and as well as a handful of emerging markets. So like Tim, I have the pleasure of getting to see innovations around the world, and my job on the investment team is to help us understand which models are responsible and are going to have long term benefits for their users, and which models we should be being more careful of and trying to keep up with the evolving landscape of predatory and responsible lending.

Jay Long  02:21

Good afternoon, everyone. Jay Long Co-founder and Chief Operating Officer of Parlay. Parlay is a loan intelligence system, and our mission is to help lenders get more small business done, loans done cheaper and faster. Within the context of this conversation, we feel pretty lucky to be both small businesses ourselves. We’re a startup and also supporting small businesses in their journey to access capital, which gives us a really interesting take on both the borrower side and the lender side of the equation and the way technology is shaping the broader ecosystem.

Elizabeth Ross-Ronchi  02:50

Hi everybody. Elizabeth Ross-Ronchi, and I am the Chief Marketing Communications and Product Strategy Officer at ACCION Opportunity Fund, and I have to, of course, recognize some of my wonderful colleagues in the room, Whitney, Eric and many other fans along the way. We are a leading CDFI lender that’s focused on not just capital but also education and technical assistance, and we really believe that those two things have to move together to do responsible lending in this country, and so our vision is to become the first mission driven, scalable, financially sustainable lender that’s doing both education and capital at the same time together.

Samir Shergill  03:40

Hey, everyone. Samir Shergill, I’m the Co-founder and CEO of High Beam. We’re a banking and AI cash management platform. That’s a lot of FinTech buzz words in one sense. So I’ll break that out a bit. So our customers are small business owners, actually, like Rebecca, who I will try to pitch shortly, but people who’ve started consumer brands who are trying to grow their businesses. And it turns out, for those businesses, cash management is mission critical, but really hard, and so a lot of times, don’t even understand how much cash they’re making. I say cash very precisely, not profit. They really understand their sales, usually pretty well. They may understand the profit margins, but not the cash, but that’s kind of the most important part of it. So what our platform does is, in one platform, you have banking, lending and cash management and planning for these businesses to kind of maximize the amount of cash that they get to maintain and retain.

Peter Renton  04:29

Okay, so I want to this is, this is about innovation. This, this, this session. And I used to run a conference called Lend It, which was all about lending innovation. And I’m thinking back to when it was about a decade ago, where we had so much enthusiasm, there was lots of money coming into the space, and we all thought that we were going to solve small business lending so that every single small business would have access to capital within a decade. That was a decade ago. It hasn’t happened. I think we’ve made strides. But that’s what I want to start off with. I’d like to get perspective from from each of you as you start off with, let’s going back over the last decade where, where have we made progress, and where do we still have a lot of room to improve? And Samir, let’s start with you.

Samir Shergill  05:24

Yeah, if I could just tie it back to the last panel, I find it really interesting that Rebecca’s point of view, like the access to capital has seemed to improve. And we see that a lot as well, where FinTech has seemingly provided a lot of access to capital for people when they need it. I think what’s happened, though, is in some ways, it’s very extractive as well. And so while people have now access to capital, they don’t necessarily have the tools to decide, is this the right product for me? What’s the impact on my business? Should I take this money or not? And these are tough questions to answer, but I think there’s increasingly this idea that because of the availability and the Internet and ubiquity of like access to these businesses, there will be people willing to provide them capital at some rate, but they typically will find ways to hide that rate or find ways to have the expertise on their side of the fence. And so I think there’s been this imbalance now where, for the first time, a lot of small businesses, the ones that we see have access to capital, but there’s a still large expertise gap on what is the rate here? Is this a good product for you? Should you take this money? What alternative do you have? And so I think that’s where, kind of we see the pros of access, but the cons of not being able to actually compare options.

Peter Renton  06:39

Anyone also want to jump in on the last decade?

Elizabeth Ross-Ronchi  06:42

Yeah, I’d be happy to. So I love that Samir. And what I would add is the digital platforms and fintech platforms have been tremendous in giving us confidence that we can solve this three legged stool that I talked about in terms of financial trilemma at the same time, though, for the vulnerable small business, more vulnerable small businesses and communities that we’re looking to serve, having that human interaction, to understand edge cases and to understand their context is incredibly important, and so I think that’s really important, and we’re still trying to figure that out as of course, unit economics are important, but also that human interaction is incredibly important to maintain.

Jay Long  07:29

If I could just add to I think, from the FinTech perspective, it’s been inspiring to watch the shifting perception from lenders and banks, where, I think maybe 10 years ago, fintechs were seen as an extraneous third party group, and now they’re seen as core enablers of operations and strategic partners, and we’re seeing emerging roles at banks, like Chief FinTech Officer or something along those lines. And so the interoperability between the two systems has been really helpful.

Malika Anand  07:52

Yeah, I mean, I think we thought that in solving the financial access problem, that we would see this unlock of small business some small business some small business longevity, small business resilience. And in fact, those numbers haven’t changed, right? You know, the panel before us mentioned that most businesses don’t make it to five years, and maybe only 25% make it to 10 years. And those numbers have stayed more or less stable, even as the FinTech revolution has unlocked all kinds of access. So I think the question becomes, what’s what’s going wrong in some ways and what’s going right in other ways, right the access, the ease, the fact that the button is right there for Rebecca, that she gets the money tomorrow. All of those are massive improvements, I think most small business owners would claim. But yet, we’re not seeing the commensurate sort of improvement in outcomes that we would hope to see, which is that people feel more capable in their businesses, that they’re able to operate them for longer, that they are not digging themselves out of holes months and months later. So that’s where I would, I would focus?

Peter Renton  08:53

And as Rebecca said in that last in the last panel there that she was, I think she she talked about Shopify, advance, paying back 10% every day. That was a little shocking to me, that but the problem that I see, and what I want to maybe Samir, I’ll go to you first, is this, the transparency and the information asymmetry that you know, she didn’t know what the APR was. Maybe Shopify provides it. Maybe it doesn’t. It would only be an estimate anyway, because it’s a it is just this advance that cuts coming out of you don’t know how long it’s going to take to pay off the loan. So What? What? What innovations are actually moving the needle when it comes to transparency today?

Samir Shergill  09:42

Yeah, and if I could just build on what Malika said there, what’s interesting to me is, like FinTech has not fundamentally transformed how small businesses make financial decisions right, like they might be writing access and the FinTech might say, Well, I’m done a lot of work to make sure if I get paid back or not. And I’ve done a lot of work to make sure if I get my money back, but there’s not yet been a transformation in helping the small business make better financial decisions, right? And this goes back to this information asymmetry problem where, you know, take the Shopify example, and Shopify is not the worst offender here. In some ways, if you’re you know the cost of capital is what what it is to some of these businesses. But let’s just take cash advances, and they’re preying, in some ways, on people’s intuition. So give you an example. Someone might say, this is a 10% fixed fee loan over the course of six months, right? And maybe you say, like, Okay, I’m pretty smart. I have the intuition that I’ll double that. And it’s a 20% loan because it’s 10% six months, 20% 12 months, expensive, but still fine. Well, actually it’s a 40% loan, because what they do is they start taking money every day, so your average balance is about half of what you would think it would be. So these are things that they’ve obviously thought about, but the small business owner is not going to think about, right? So there is information asymmetry about it’s kind of like the same thing with you’ve seen lots of other different spaces and gambling other places, right? Where you need to have fairness on both sides, or expertise on both sides. And so I think that a lot of the innovation that I’m excited about and what we hope to bring is, how do you bring the expertise to level the playing field for small business owner, and how do you in a way where they can make informed decisions on Yes, this is good for me, or this might be expensive, but still my best option. Or a lot of times, like, don’t look for a financing solution for an operating problem. Like, make the hard decision. Like, sometimes you have to make the hard decision to lay someone off or do something different. But sometimes that short fix looks like the right answer. It might not always be. So I think that’s kind of part of the problem here.

Peter Renton  11:38

Right? Do you want to join us?

Jay Long  11:41

Yeah, if I could just build off your point. I think when you think about transparency and data, and you think about like the data science pyramid, descriptive data is important. It tells you what the real cost alone is. I think where it gets increasingly fascinating as you look at the role of emerging technologies like AI is when you go into diagnostic and prescriptive and even proactive data. And I think within the context of merchant cash advance or other tools, a question might be, how do we get left of the dilemma where I need to get cash tomorrow or not, and what can AI or other tools do to allow me to understand the tapestry and terrain of my business so that I can be working in advance? And I think the education layers are much deeper than just this individual product, and it’s more, how does FinTech enable the strategy itself?

Peter Renton  12:22

Well, we are going to dive into AI in the second half of our discussion, because this, the sub topic for this is, it’s not all about AI. And so we want to talk about the innovations that are happening. I mean, obviously AI infuses absolutely everything today, but the core innovations that are happening that are not really as a result of what AI is doing. And one of the things I want to talk about is the education piece. And Elizabeth, I know at ACCION Opportunity Fund, that’s a really big part of what you do, helping helping small business owners understand finance better, because not you know, as you say, most small business owners don’t, they don’t get into small business because they want to become expert at managing finances. So maybe you could talk about how you do that and how it changes outcomes.

Elizabeth Ross-Ronchi  13:16

Yeah. Thank you, Peter. So you know access without understanding is an opportunity, and we’re really focused on this integration, integration of lending and learning, and it’s based on an insight. So we have several small businesses who have said, Don’t give me capital without giving me the know how, and for the populations that we serve, there’s oftentimes not as much inherited knowledge or financial acumen, and some of the traditional financial education can seem inaccessible and not grounded in the practicalities of what they’re experiencing on the ground. So today, when we have someone who goes through our pre-qual check, for example, or gets declined, we don’t view that as the end of a relationship, but still the beginning of a relationship. So for example, if somebody doesn’t make it through a pre-qual check or they get a decline, reason, we actually say it in plain English, and so that gives them agency to be able to take accountability and to focus on getting ready and what readiness means for them. And the way that we further help them do that is actually through the help of several whether it’s JP Morgan, Chase, MasterCard and others who have helped us really develop the ability to personalize learning. So right in that experience, let’s say you weren’t ready because of cash flow, then we’ll serve up a cash flow learning module right in that environment. Over fiscal year 25 we had about 900 small businesses take advantage of this and share that they felt like they mastered that experience. So you might say, okay, so fine. Why? Why are not, you know, more organizations doing this well for CDFIs, this is resource intensive. It takes investment, right? And so that’s a challenge, much less to do it through a digital environment. And for digital providers are really focused on unit economics, of course, and so it takes a belief that this is important, and then focus and then sustained accountability. So that’s the journey that we’re on, Peter

Peter Renton  15:40

Interesting, so Malika, I want to turn to you and something that you talked about in on our prep call a couple of weeks ago. You were talking about resilience solutions and how the industry is shifting from this growth, growth financing mindset to what to a more resilient financing mindset. Explain what you what you mean there.

Malika Anand  16:02

Sure. I think when we, when I first came to the microfinance industry, 20 years ago, we talked a lot about startup capital, right, or growth capital for small businesses. And what we’ve seen is that people start businesses right? We heard that just earlier this morning, people have they’re able to find the resources to start businesses, but those businesses don’t survive, right? So only 25% make it to 10 years. And when you ask them, What happened, it’s not that they didn’t have a great product, it’s not that they didn’t have customers, it’s not that they didn’t have suppliers. 82% of business owner of closures in America happened because of cash flow management problems, which to say, sort of like Rebecca was talking about, there was an expense, there. They need to stock inventory, and they’re not going to get paid for a 60 day cycle, a 90 day cycle. Meanwhile, you have bills to pay, employees to keep on payroll, have to keep the lights on, and that mismatch between your income streams and your expense streams just create these gaps, gaps in liquidity, as Tim mentioned, and it’s almost a perfect problem for FinTech, right? Because what it means is that there’s these kind of you can see in cash flow where the money’s going, who’s getting paid at what time, in a pretty predictable way, and you can see what money is coming in from from customers as they swipe cards or they purchase on on Amazon or Marketplace or whatever it is. And these are oftentimes short duration gaps, which is also eminently financeable. But many of our financial service providers are not set off to meet that kind of financing need. But what we’re seeing more and more in our portfolio, and what we’re really excited to see in the innovation piece of this panel is the sort of variety of cash flow management solutions that are coming up. We’re very proud to have High Beam in our portfolio. And Samir will talk more about how their match solving this problem for small business owners.  But we have companies that are solving this based on invoices, companies that are solving this based on trade agreements, companies that are solving this based on revenue that based on so there’s a whole, there’s a whole, I think, industry and community forming around understanding the cash flow and cyclicality of small businesses, and they’re there in crafting these very tailored, very particular solutions. And much like we heard, they’re able to create experiences that feel easy and seamless. 10% feels like a lot, but it also means that a little bit is going every day. You don’t have to remember to pay it. It’s not happening on the 15th, which might be a Sunday. In some ways, it’s if you don’t want to focus on HR and you don’t want to focus on bill payment. These can be very natural, agreeable kinds of arrangements for your small business. But as Samir was mentioning, it can be very difficult to understand how much they cost, what the obligation for your business might be over time, how it might square with your long term profitability. And so this line between what responsible resilience solutions look like and what predatory solutions look like becomes increasingly difficult to to distinguish.

Peter Renton  19:20

Right. And let’s, so let’s, let’s dive into that. And I, I want to turn to you, Jay, because one of the things that that you’ve really focused on is this cash flow data that’s available through open banking. I mean, that’s an innovation that is, has been groundbreaking, I think, for for the growth of FinTech, and it allows for, you know, cash flow analysis at scale, which was simply wouldn’t be possible without it. So maybe you can talk about how, like, how you’re how you’re using it, how you’re making this available for lenders, and what, and how it’s improving the not just the. The borrower experience, but the lender outcomes.

Jay Long  20:02

Absolutely.  A theme that we’re hearing a lot today, I think, is that cash flow injection is not a way to mitigate against operational challenges. And so what we end up seeing, and it’s brought up, cash flow management is important, and operator expertise varies a bit wildly. Where we found a lot of success is Parlay in partnering with MasterCard is that providing lenders with the ability to have on demand access to cash flow data allows them to see a lot of variances in the way that a business is managing cash. So they can be much more than just a lender where you’re negotiating on rates, but they can be a capital stack advisor. So we found for some of our CDFI lenders is they’re able to turn on the open banking data, API call per inquiry at the point of need, and so as they’re doing the initial assessments, they’re able to understand cyclically and at scale, how are these small businesses leveraging cash? There’s a few things. One we’re finding it helps them better understand and surface if MCAS or other injections have been used, and it’s less of a gotcha game with the small business and more an opportunity to educate them and become a partner. And so we’re seeing there’s a lot of really tailored education instruction. A theme that’s come up consistently this morning is the importance of being a really good technical assistance provider, so our partner is able to use that data and look at a small business with a lot of focus on the unique components of its growth, and have really tailored conversations about not just this individual inquiry but the overall tapestry of products that could support them. So we find that’s a really powerful tool. Open banking doesn’t just empower the lender, but also empowers the borrower by giving them more insights. And we’ve seen a lot of really cool opportunities to use the data in the application process to increase awareness for both. So we’re able to generate dashboards back to the applicant, for example, so they know where they are, which can be really helpful for small business owners.

Peter Renton  21:49

Okay, so then Elizabeth, I want to turn to you, because one thing that I wasn’t aware of until we had our call recently was the research you’ve done with Ripple, which I thought was super interesting. And so tell us about that research and the digital tools that are available for these for underserved small businesses that that haven’t been available before.

Elizabeth Ross-Ronchi  22:15

Yeah, absolutely. And of course, we’ve not released this research yet, but I will share some early headlines, and you can look forward to seeing the report soon. So we conducted a national survey of about 600 small businesses, the majority of which are underserved small businesses. And interesting. You know, you might think that underserved small businesses are more tech followers as opposed to leaders. But what we found is, of course, it makes sense for operating leverage, they’re actually more urgency, right? Like tech, technology is urgent, and so interestingly that digital tools and technology was actually statistically significant, found more important among underserved small businesses than all small businesses. No surprise that financial tools, payment tools, were the top category of importance. But what was super interesting, as we’re talking about AI, is that AI was by far more important for underserved small businesses than all small businesses at this time. And I think again, that really speaks to some of the pressures even that Rebecca talked about in terms of whether it’s operating leverage, and we even experience as a as a small business ourselves and how we’re using AI to give us operating leverage. So I you know some of the interesting findings, though, what are the barriers, right? What are the opportunities for us to think about as we’re designing products and solutions. First cost, right? So any new technology needs to be accessible, and then the second is intimidation, right? Intimidation in trying to figure out how to implement some of these new tools and technology. So anything that we create thinking about how it’s accessible in terms of affordability for different segments of customers. And then second, how you can make it really intuitive and make sure that you’re being customer first in your design. So those are the some of the things that we’re thinking about. And again, we haven’t published this research yet, but those are some of the early things coming out of it.

Peter Renton  24:43

Okay, so I want to, I want to switch gears to another innovation I think that’s been groundbreaking for a lot of small businesses, particularly when it comes to capital access, and that is the embedded lending piece from you’ve got in the vertical SaaS space. Pretty much every industry now has a vertical SaaS player or multiple vertical SaaS players, and what what that has meant is that they understand their niche better than any generalist bank could ever understand it, and so they’re able to analyze the numbers in ways that have better predictive power, and so they’re able to lend in a lot more aggressive ways. And what I think this has done has been it’s moved a lot of the financing outside of the banking system. I mean, often banks are at the end of it, but you know, you’ve got companies like CIM providing financing for for a lot of these things. So I want to open this up to anybody. I mean, what tell us about what you think of this sort of, this trend towards embedded lending and this expertise inside the vertical that these software companies have, and is this a threat to banks and CDFIs, or is it an opportunity. Anyone?

Samir Shergill  26:04

I’m happy to go first, I think it’s probably both. But I do think that it’s really interesting that, if you think about it, that most financial products are horizontal and that they’re consumed by all businesses. Everyone has the same credit card, everyone’s the same bank, everyone has the same access to the financial products. It’s not specialized by business, right? Whereas that changed completely for software. So what happened in software was to help you run your business, you have to understand what the business is doing. So Shopify helps consumer brands, right? Because understands that its business is selling consumer goods. Toast helps restaurants. It understands that my business is, you know, a restaurant. But if you’re a lender to the business and you don’t fundamentally understand what the drivers of that business are, how are you making underwriting and servicing decisions to the best of your capability? You’re not. And so if you think there’s a one size fits all. I’ll just look at the accounting data and do the same asset based underwriting, or whatever underwriting I’m doing across every single business type. That doesn’t work. It doesn’t scale. So to me, it’s more if you know, lot of times technology or FinTech or startups will race to where the opportunity is, it doesn’t mean that doesn’t exist for, you know, banks or other players. It’s just a matter of, are they willing to augment their underwriting with an understanding of maybe we should consider these augmented data sources. Maybe we should change how we think about things. And you know that pressure of will we want to do that? We’ll see. But I do think that what’s happening now is there’s an increasing understanding that the underlying business data to order, to process that and understand the health of the business or the quality of risk, you have to understand what the business does, and you have to understand how that data relies on that for just to give a quick example, in our segment, if you’re selling consumer goods and we’re lending to these businesses, if your first order profitable, or if your return rate is super high, that’s a bad sign, right? So I know that if you’re selling a product, but your return rate is spiking, that’s a really bad sign, and I’m going to stop lending, or, you know, pull back. Now, if you’re a bank and you don’t know this, well, obviously you should, but you know that data is available. So I think that it’s a matter of, will the industry evolve such that you’re not just using accounting data signals, and you’re using these signals that now are available by business type.

Malika Anand  28:24

Yeah, I don’t know if I agree that that banks have been so agnostic about sector or type. I mean, I think we’ve, we’ve had a long history in financial services, both in the US and outside of the US, of quite specialized lenders, whether that’s ag or project finance or infrastructure. I don’t I do think that that whether it’s formalized in the language that we use now, in terms of vertical or horizontal, there might be a debate there, but the notion that you must understand the fundamentals and the mechanics of a business to be able to lend to it well, I think has always been true in the financial services sector.  The fact that we now have a range of what we used to call alternative data, and now might call open finance. Means that there’s making be maybe much more specific about especially maybe in the small business sector, but the notion that business fundamentals matter in underwriting, I think, is well established.

Jay Long  29:22

I think my only contribution would be Mike. I would imagine there’s few cases where having more options is generally bad. There’s times when speed matters, in which case that can be an incredible resource. If you know that you’ve got a big company investment coming up, then some of the balance sheet strength of a traditional lender and the rates that they can offer accordingly might be advantageous. So my my current hypothesis, the way we think about this in Parlay is by creating more access to opportunity and more access to financial resources that can be bespoke to your need, you’re going to have, hopefully, a broader ecosystem of businesses that can become future partners for financial institutions across their journey. So I think it’s net additive. But we’re definitely in a storming phase as a collective.

Peter Renton  30:02

Yeah, that’s fair. Okay, let’s talk about AI, and we have to. Yes. So there’s, there’s obviously a lot of hype out there. There’s also, there’s also a lot of incredible use cases that are, that are happening. Maybe, let’s start with how you are deploying AI today, beyond just doing coding, which I think has become a pretty well established use case. Now, if you are not coding with AI, you are way behind all your competitors. So let’s talk about what, what? What are the use cases, and what are the concrete differences making? I want to hear from everybody. Samir, I’ll start down with you.

Samir Shergill  30:49

Sure, I’ll put in two buckets. There’s, how are we using AI internally, and how are we using AI in our customer facing offerings? And both have evolved quite a bit internally. I think it’s a lot of it is, can we more efficiently clean and categorize this kind of unstructured data the small business generates in order to have a standardized data set to do underwriting, to provide business logic, to do other things with it? And that, I think, is actually a game changer here, which is the way small businesses run their finances a lot of times. It’s kind of crazy to me is that there’s all this innovation around it. You talk about Shopify or Toast or these other platforms, but how they run their finances is still very much month end accounting. I get a random P&L balance sheet. I’m supposed to make sense of that most people actually just look at their bank balance and guess that’s usually how most small businesses run their business day to day, or they build some Excel model on the fly. So what we’re really focused on is, can we take the data from the small business and kind of short circuit that accounting loop by extracting the data and then in our customer facing offerings, putting that data in a format that helps them make better financial decisions? So rather than giving them a generic P&L balance sheet or cash flow statement, give them something that feels true to them, helping them manage their finances, something that shows them the impact of what happens if your return rate drops, or what happens if you take this loan. What’s that likely to impact your cash so that’s the part where I’m excited, where I think AI can increasingly drive additional financial expertise and awareness into the small business, which then has tremendous knock on effects for everyone.

Elizabeth Ross-Ronchi  32:28

Thanks Samir, yeah, so we think about using AI one. First, how are our customers using AI to discover and find us? And then, how are we using AI? So first, our customers, we’ve seen a significant shift. We’re so focused on discoverability, and we’ve seen a significant shift in customers coming to us from large language models. And so that means we’re really focused on studying that and studying what they’re looking for, so that we can create content so that they can find us more readily.  And then internally in terms of how we’re using it. Really, for us, it’s always about managing the cost. There’s so many exciting solutions out there, but we only have so many dollars that we can invest right and there’s a tremendous amount of data sources and solutions that we would love to get our hands on, but that’s a challenge, so we really have started using it to reduce our own cost of operation, whether that’s marketing, servicing, underwriting. I mean, the list goes on, data, data analysis, and more and more, we’re looking to service providers versus doing things ourselves, whether it’s Zendesk with servicing or whether it’s using getting to machine learning models within CDFIs, which is over the past couple of years, that’s a big deal just for a CDFI to use a machine learning models. Of course, banks have been using this for decades, but it at the end of the day, it really gives us the operating leverage that we need to become more financially sustainable on our lending.

Jay Long  34:12

From our end, I think there’s a couple layers, both as an internal company and then we’re seeing in the market and both sides of the platform that is the borrower and applicant. On the borrower side, like our company is called Parlay, and for those Pirates of the Caribbean fans out there, there’s that moment where they’re going to have the negotiation. It’s a really tense conversation. I think oftentimes for small business owners, walking into a lending institution can feel pretty intimidating, because they know they’re about to be assessed pretty rigorously, and the default answer usually is no. I think four out of five times it’s actually no so by decreasing barriers to access and helping them navigate the process in abstracting way the need to interact with forms and instead asking them like straightforward, plain language questions, while models in the back end can orchestrate the data that’s necessary to support lender underwriting, massively decreases time it decreases barriers to confidence and execution. So we’re seeing is it’s much more approachable on the lender side, what we’re finding is as you leverage these tools to increase your capacity to structure, clean, originate and automate a lot of the document collection preparatory stages of underwriting, your cost curve shifts left. And so we’re seeing is by taking the manual toil out of gathering documents and making API calls and cleaning information and structuring it is economically feasible to lend into segments of the market that previously were cost prohibitive. And I think, you know, one of the constant themes we see, or at least I hear, is a conversation around we wish lenders would be more proactive in reaching into some of these underserved communities. Often, the intent isn’t so much to neglect communities, so much as they can’t actually make it economically viable to engage. It’s like a cost problem, and so we’re seeing with AI is if you can remove the majority of the work that takes time that, in turn, builds in cost, you can suddenly expand reach effectively, and you can also mitigate against fraud by accelerating your capacity to take proactive steps, so you’re increasing volume and scale while decreasing risk surface area, and the net result then becomes a much more fluid and effective and permeable capital market where you’re able to leverage AI to increase outreach. For our company internally, I come from military background, so all this was net new to me and our CEO, and I will do things like cash flow analysis and sensitivity analysis based on business outcomes that if I had to know how to do organically without the benefit of YouTube and Claude would be quite challenging. With enough time Red Bull and prompting, you can get really like, nuanced understanding of where are we now, what happens with these different business decisions, and how does this better? Let us steer the ship. And so I think it’s decreased barriers to expertise and entry, while increasing your capacity to really effectively navigate these challenges.

Malika Anand  36:49

Yeah, I think I might skip how we use AI internally. I don’t know, say that much about small business, but I can, I can tell you a little bit about where we’re seeing AI get used across our pipeline and our portfolio, and I hesitate to be like Debbie Downer, but a lot of what we’re we hear is a lot of theoretical use of AI, how it’s going to make customers finding us and engaging customers so much easier, so much more not so much more natural. How it’s going to make the onboarding process, sharing your data, being able to identify the right product fit, that it’s going to get easier there, that it’s going to bring down the cost of underwriting, and therefore the cost of financing, that it’s going to make portfolio management so much better. You’re going to identify fraud sooner. But it might be that it’s coming, that all of a sudden I’m going to see a decrease in the cost of capital, and that all these people are going to unlock, you know, unlock access, and that fraud rates are going to drop. So far, we’re not seeing that. I don’t, I don’t know that. I can think of a few very interesting exceptions where AI is really being deployed in a transformative way. But right now, I think, from where we sit, it’s still in the realm of exciting potential that hopefully will manifest.

Peter Renton  38:11

You bring up a good point, because you, when we all saw in the news, Block has laid off 40% of their staff. They are a huge lender to small business, one of the one of the largest in the country. I wonder if all this reduced cost now is going to lead to reduced cost of capital for their customers. I would probably guess, though

Malika Anand  38:36

Joyce is laughing because anybody is curious?

Peter Renton  38:38

Yeah. I mean, I don’t want to pick on Block, PayPal, shop, Shopify, they all have Stripe then. I mean, they’re not necessarily going to reduce the cost of the small business, right?

Malika Anand  38:54

You said it, not me, but yeah, one would hope, if you’re dropping your if your operating costs are dropping by half. Samir, to your point, could you not now offer a substantially cheaper product to your user? But without being facetious, I do think that we can’t lose sight of what the outcomes are for the borrower, right? Whether it’s about what information or coaching or intuitive process we offer them. The truth is that we don’t yet know, or we haven’t spent enough time or care in understanding how that creates resilience and opportunity and well being for the small business like there’s just unbelievable, I think gaps in our understanding of what is the real unlock for them in terms of their financial service, access and experience, and one will hope, actually, the AI might be a part of that solution, being able to enable us to collect more data more frequently, in greater depth, and be able to analyze it in a more detailed, nuanced fashion. But, but yeah, I mean, I think the pressure, the potential for AI to decrease costs, we need to, we can’t lose sight of the fact that we would hope to see that benefit trickle down to users.

Peter Renton  40:09

Right, right? Yeah, that’s, that’s, it’s well put. So I want to turn to the lenders we have on stage here. And before we go. I think one of the things that we have to keep in mind with AI is its potential to automate bias that is already in the system, that has been in the, you know, in underwriting, for decades or centuries. Shall we say? How are you guys thinking about it? So it’s Samir and Elizabeth, I’m talking to you. They are the lenders that we have. How are your organization thinking about governance and accountability when it comes to this AI driven decisions and the potential bias that it might bring?

Elizabeth Ross-Ronchi  40:54

I can go first. So in terms of automating bias and governance, I mean, we are not using AI directly in our underwriting to the degree we have machine learning models that we’re using. We work with Experian that helps us really increase our data sets so that we have more statistically significant data sets. But in terms of AI driving underwriting, your decisioning. We’re not, we’re not there yet, but in terms of governance, and the way that we think about it is, I think that there, there’s danger, and we have to be very careful about binary decisions, binary use of AI to complete a task, and then just taking that task, as opposed to using AI to solve strategic problems and opportunities, and making sure that we’re editing, we’re reviewing, putting it back through against the strategic problem that we’re looking to solve, and then reviewing again, and that’s really across the board, no matter no matter what. So Samir, I don’t know if you would.

Samir Shergill  42:05

That’s well said the when people talk about AI right now, that’s usually a proxy for using an LLM, a large language model like Claude or Gemini. And these models are, you know, non deterministic black box models, right? And so they’re not necessarily best served for most tasks. They’re better machine learning models that underwriting statistics have done for a long time that are better served for those tasks. So I think this is one catch all people think that you just gotta ask Claude like, Hey, should I lend to this business? And that’s not how it’s gonna work, right? Where it can be useful is in the accessing, cleaning, categorizing the data. And for us, it’s making sure that we’re going back to the I’ll call it, like the raw data sources, like the sales data, the transaction data, these, these kind of like, you know, primitive data sources that are available in real time. Can you process them and then use that as a way to make the decision, not through the LLM necessarily, but through machine learning models, or even, we have human underwriters for every underwriting that we do. So I think that it’s important to have your pipeline understand where AI will fit in and then where it will not. And for the foreseeable future, I don’t think for most responsible lenders, that they’re outsourcing significant chunks of the underwriting decision to an LLM. I do think as new models emerge and there’s more kind of white box machine learning based models, then you know, we’ll reassess as that comes along.

Elizabeth Ross-Ronchi  43:28

Samir, I have to just build on. One point that you made, is that this combination of using AI and human interaction, particularly for edge cases, and where you’re lending on the margins, this is just incredibly important. It’s also important from a relationship building perspective, certainly in with our audience.

Peter Renton  43:51

Okay, so I want to, I want to touch on the LLM piece that you, that you mentioned Samir and Jay. I want to turn to you because I think it’s something that I think is one of the really interesting, things about Parlay, to me is that all of the application data that happens for every small business loan that is this is data that, you know, it’s, it’s, it’s real, it’s available, but the LLMs don’t have access to it. It’s dark, as far as they’re concerned. So how useful like, like, maybe you can talk about the data set that you that you are building, and that is being built, expanded every day, and how that can help in you know how that is really helping lenders be able to underwrite better.

Jay Long  44:39

Absolutely. So I think the first thing we’ll call out is the goal isn’t just making bad decisions faster, right? What you want to be able to do is have a really nuanced approach to decision making. The challenge with a lot of foundational models is that they’re only trained on what can be acquired. And for a lot of CDFIs and community banks and regional lenders, a non trivial amount of your data might be in PDF’s, or might be in file cabinets or might be in email, and the net result, then is the information you need to understand business growth over time is not accessible to a model, and no amount of pouring money in or building data centers changes that, like the AI is a function of what the data was trained on. And so what we did was we said, how might we originate data sets from first contact for anyone coming in through the front door, and allow that to be aggregated over time. So the lender’s unique approach to lending exists, but also they can ask new and better questions. So in the past, if I’m looking purely at automation, I can say yes or no faster, but if I’m bringing in hundreds or 1000s of inquiries over the course of a period of time, you can start asking questions like, what products should I be offering for these segments? And so now you’re able to, like, make really refined judgments on not just this individual application but also your broader go to market and how you’re serving the community. And that works only if you can effectively originate and structure data sets, which, as has been mentioned, is like, really time and labor intensive. If we’re doing this manually. If you’re having machines do it, it’s incredibly intuitive and really fast and really powerful. So we’re seeing and one of the things we’re most excited about is, how might we originate a small business lending data set that transforms understanding at the borrower at lender level, but also from like a governance perspective or even policy perspective. We’re seeing certain industries struggle early, then we can be proactive in doing that. But right now, it’s hard if everything’s trapped in email and workflows that are static. So the origination of data sets that train models allows us to be much more inclusive collectively as a community.

Peter Renton  46:34

Okay, so we’re going to hit we’re going to go to audience questions here in a little bit, but before we do, I want to talk about something that is potentially coming. So I hope you don’t mind, Malika, we’re going to go into theoretical possibilities. Now, Louis and I were chatting at dinner last night, and we were talking about how AI agents are going to come to small businesses at some point, and they’re going to have much better information, the small business owner isn’t going to need to understand the difference between this MCA product that’s 10% over nine months or a fixed or a fixed term loan or a line of credit or an equipment finance or whatever, because the AI agent is going to have all this fantastic information that will make it, will it, will understand it all for you. So where are we with that? Samir you’re, I think the one here that’s, I think, closest to you know you’re working on some of this stuff. I mean, how close are we to getting AI agents that are actually helpful for small business owners?

Samir Shergill  47:48

Yeah, if you’ll let me, I’ll first paint the picture of where we could be and where we could be actually not so far in the future is imagine if you were a small business and you had access to a world class CFO at your fingertips, and this world class CFO could always help you make the best decisions. You know, there’s still art to this. It’s not like precise, but as a lender, if you knew a small business had a world class CFO and a world class financial operations and world class data, does that change your assessment of the risk of that business? For us, it does, right? And so where we’re saying is, if we are the banking solution, and we are providing the AI to you, and you use it, then that’s this virtuous cycle whereby we then feel better lending to you, right? And so where are we on that journey? I think we’ve gotten to a point now where you can and this is what we’re building fundamentally improve the cash operations and cash planning decisions, so that they’re not constantly under the gun to say, like, Oh my God, I need $20,000 I didn’t realize I needed this, right? That example that we had earlier, which is, how can you get ahead of this to plan for the right cash solutions? How do you clean the data? How do you actually for and this is why we focused on one vertical first, the agents now are able to help you plan and then take action on that plan. So you know, if you have a plan to say that I want to maintain this much balance and pay this bill in the state, then the agent will do it for you. Now it’s early days, and it’ll evolve, but I do think things are moving quite quickly, and so our hope is that kind of, it’s kind of like the intelligence coming back in and being run by the agents, but managed by the humans, so that, I think we’re not quite there yet where the strategic decisions can be made by so you don’t have the CFO yet that can be provided, but the financial analyst and the treasury person and the accounts payable person and accounts receivable person, all these jobs that would have been done at a large company by experts can now be given to a small business.

Samir Shergill  49:40

Now, in some ways, the key is not, it’s not the it’s that’s the vision. Of course, the setup is hard, right? The setup requires replatforming and moving your operations from what you do today onto a new system. And that’s what we find. Is actually the biggest barrier to entry, is actually uprooting their existing way of doing it. And pushing it onto a new system. Now the good news is for SMBs and mid markets, this actually is easier to do than a large enterprise, so this is kind of leapfrog opportunity. So what I always tell mid market small business owners, it’s the best time to be a founder, because you’re able to move faster and uproot your old way of financial planning and management, that you hated anyways, and move on to the system, then someone that has 100 person finance team is going to struggle. So to struggle. So I think that in the short run, there’ll be a lot more adoption in the SMB mid market of these solutions. And that’s why I actually think there is a there’s, you know, pros and cons to all this stuff, but I think there’s opportunity to really level the playing field when it comes to that.

Peter Renton  50:35

How far out are we from having from a small business or middle market business, having a solution that will actually help?

Samir Shergill  50:42

We’d like to think our solution already helps. But I think that, you know, depends on what you mean by help here, if it, if you mean, like, just do it all for me, press a button. I don’t think about finances ever again. No, if it’s press a button and, well, do a bunch of work for us, then press a button, and then at that point, you know, I have a system now that will help me make better decisions, help me track my cash, run my money for me, so I can run this entire operation with, you know, five hours a week, where, before I take a three person team, we’re already there.

Peter Renton  51:19

Okay, that’s, that’s, that’s great. That’s encouraging. Okay, questions from the audience, do we have, do we have a microphone? Okay, great. There we go.

51:32

[Cameron Dawes, Milking Institute.] This is for Elizabeth. With the customized training for loan denied businesses. What has been the percentage of loan approvals once the firms has completed your training?

Elizabeth Ross-Ronchi  51:48

Yeah, thank you for that. So we have just been building this unified system, and we’re just building that data set. So it’s a hypothesis and a belief, but some early indicators is that we’re tracking borrowers who do learning, and learners who then borrow or borrow, and people who learners who have completed a learning model show higher approval rates when they go apply for a loan. So those are some of the early indications that we’re seeing, but we’re looking to track this over the coming years. Thank you for the question.

Jay Long  52:33

If we have a gap in questions, I’d just like to potentially build off the earlier conversation. One cautionary tale on this, the AI can pick out the product. Piece that I think is really relevant is the models are only as skilled as the questions you ask it. And I think it’s worth underscoring that making like bad decisions faster or incurring unconscious bias or not sufficiently framing the question you’re asking for it to deliver the product for might incur risk. And I think all that to say, as we’re looking at the future, the foundations and fundamentals the present, which is effectively training these borrowers, becomes really important, because the model will give you an answer, but if it’s not the right variables you’re solving for, is it money now? Is it best rate? Is it long term, resilient to the business? It will unintentionally take you down a path that we don’t know where it leads, but it’s probably not optimal outcome success. So I think even as these become more ubiquitous as tools that doesn’t negate the need for human judgment and shaping what ends up happening,

Peter Renton  53:32

that’s fair. Okay, question over here.

Jennifer Spaziano  53:38

Hi, Jennifer spaziano from Ascendis. I have a question for High Beam and Parlay in terms of the adoption of the tools. What are you seeing the biggest barriers are to small businesses as a small as a nonprofit organization, we have tested many different models with many different funders, and it’s hard to get small businesses to use these tools, no matter how good they may seem.

Samir Shergill  54:06

it’s a great question. What we found is you have to, and part of the reason we bundled our solution is you have to be able to provide the capital as well. So providing the lending was a core part of our value proposition, where we’re not just providing the intelligence and providing you this kind of, you know, idealized version of it’ll help some abstract decision making, but as part of that, will also give you the money today to help you grow your business. And so I think for businesses where the need capital need exists, there is that kind of pressure to make a decision, and then we’re one of the considerations set and they’re more likely to move. So I think for us, at least the bundling of the solution, so the intelligence, capital and banking all is one platform has helped, kind of with the value propositions of the business that you’re kind of picking a new financial partner, and that’s us, and so that kind of cognitively helped them, I think, make a decision on what we. Were versus some abstract AI intelligence tool that they don’t really know how to process, how to think about.

Jay Long  55:06

On our end, our answer is probably reflective of the business model. And so in that case, we extend our solution as a white label to lenders, who then bring on their small business owners that way. So the small business owner, naturally, if they’re working with the given fi kind of go down the path. So then the barrier to adoption becomes the FIS themselves. And I think here, what we’ve observed is it’s not dissimilar to any other product adoption curve. So if you think, like Jeffrey Moore’s book Crossing the Chasm talks about, you’ve got, like the far left end of the bell curve, that as soon as idea comes out, 1% of the population gets it, and they’re on board. And then a little bit larger group, still a little bit slower, understands what it can be, and they’re willing to work through bugs. Big gap. Early Majority is interested only if there’s validated business cases behind it. Late Majority moves only because of FOMO, and laggards still want rotary phones, right? So the reason that’s relevant is, I think we’ve seen that the cognitive framing has to be taking someone into the future by recognizing the opportunity cost of the present. And I think there are some people that are hesitant to displace what is known and comfortable and convenient for what is definitely the right business solution. So for us, it’s been a function of finding the teams and the institutions and the boards and this whole tapestry of decision making that happens within an fi that recognize the power of moving to the future in a way that lets them better take care of their customers in the present. And so I think once they’re willing to take the journey with us, we’ve had tremendous success, but we’re seeing the standard adoption of realizing that this is foundationally changing, not just the product you offer, but the processes within your fi and also how you train, hire and manage your workforce itself. Hopefully that’s helpful.

Peter Renton  56:47

I think we have time for one more question.

56:51

Two versions of this question. Back to the how far out are we? One is, how far are we out? Because stuff stops being free that all of our thinking about AI is because, like, Claude is free for most of us, for most use cases, or close to free. When are we going to hit the end of that runway? And second is, how much is it going to cost you to know that that world class CFO is a Chief Financial Officer and not a Chief Fraud Officer?

Samir Shergill  57:20

Yeah, it’s a good question. I’ll, I’ll give you maybe a parallel in what was true in our own product development, which is, you know, I was software engineer by training. We employ software engineers to write code. That’s our product, right? And as of November, I was still in this kind of mode of like, Yeah, it’s nice. It’s proof of concept. This AI, our best engineers are much better all this stuff. And then December, early January, these new models came out that were suddenly game changingly good, and had this emotional reaction, like, whoa. And like our best engineers now are just, they’re still involved in the process, but they’re using agents. The agents are writing the code, and now you’re much more of an Agent Manager. And this happened like very quickly. And this is with like Claude 4.6 and then you open X models, etc, and like, what’s happening now is like 4.6 a lot of it was written by 4.5 so the models are helping write the next model, right? And so the growth curve is just accelerating. So the heuristic I have is, if the model is somewhat good at it today, kind of clunky, but kind of not in the right work. But the curve at which it’s improving in the next three to six months is rapid. So just to answer your question directly, I don’t know how that applies to financial services directly, and there is a risk sure that the black box learning on the CFO, like it’s what was it training on its training on the best practices from 50 books that people have written on how to be a good CFO, etc, etc, and Gemini and Google and Claude, they control that corpus of knowledge. What we can do is we can see how well it’s working. So all we do is we try to take the top three or four models, train them on the day that we see, see the recommendations they give, assess if we believe those are correct or not, and then we roll them out accordingly. But I think there no one knows like it could be that it takes us a year for these models to get better at financials, financial operations and services. I think each functional area, as long as performed on a screen, it will come at some point. But I don’t think the good, I think the true answers is, I don’t think anyone knows when the models will be excellent at that. If you look at like, you know, music, you would think that music, something is a creative enterprise that, you know, would be last in line. But a lot of the we’ve seen these platforms now where digital music on a prompt is actually good enough that people are consuming it, and it’s competing with human composed music, right? So what does that mean for, you know, something like financial operations? I don’t know, but I do think in the next 3, 6, 9, months, there will be models that evolve that’ll be just like, whoa. This can be a good proxy for a CFO, of course, with the caveat that we’ll have to assess the quality of the advice, etc.

Peter Renton  59:56

Okay, I think that’s all we have time for. So thank you very much. Everybody. Appreciate your attention. [Applause]