In this episode of The Next Next, host Jason Jacobs speaks with Scott Weller, co-founder and CTO of EnFi, about their AI-powered platform developed to enhance the lending process. Weller shares insights on EnFi’s formation during the financial instability of major banks, their strategic use of AI for automating credit analysis, and the challenges of collaborating with traditional lenders. They discuss EnFi’s prototyping phase, leveraging AI agents to write code, and the role of ‘design partners’ in refining their product. Weller emphasizes the necessity of staying current with AI advancements and the implications for team structure and capital requirements. The conversation also delves into the broader impact of AI on the workforce and venture capital strategies.
Harnessing AI to Revolutionize Lending: A Conversation with Scott Weller of EnFi In this episode of The Next Next, host Jason Jacobs sits down with Scott Weller, co-founder and CTO of EnFi, an AI-powered platform aimed at enhancing the lending process. Scott shares insights into how EnFi emerged amid the financial crises of Silicon Valley Bank and First Republic Bank, revealing the fragility of the financial system. The discussion delves into Scott's journey as a serial entrepreneur, the role of AI in building EnFi, and the innovative way EnFi collaborates with design partners instead of traditional customers. Scott highlights the challenges and opportunities of integrating AI in lending, the process of prototyping, and the importance of staying on the cutting edge of technology. They also explore the future of AI in various sectors, the implications for venture funding, and the potential for community-driven knowledge sharing to accelerate AI development. This episode is packed with insights on how emerging AI tools can aid in building ambitious companies while maintaining a balance with personal life.
00:00 Introduction to Scott Weller and EnFi
00:13 The Genesis of EnFi
00:56 Scott Weller's Entrepreneurial Journey
01:13 AI's Role in EnFi's Development
01:30 Challenges and Innovations in Lending
02:18 The Next Next: A Learning Journey
02:58 Catching Up with Scott Weller
04:38 Scott's Background and Early Influences
06:35 The Birth of EnFi
10:04 Prototyping and AI Integration
16:53 Design Partners and Market Fit
29:51 Building the Team and Scaling
40:02 Fundraising and Future Prospects
44:39 The Impact of Raising Capital on Business Trajectory
44:59 Navigating Capital Light vs. Capital Heavy Strategies
46:09 Understanding Market Demand and User Access
47:03 Challenges in Different AI Application Markets
48:01 Balancing R&D and Market Entry
49:57 Personal Reflections on Entrepreneurship
56:39 The Role of Agents in Software Development
59:31 Implications of AI on Hiring and Industry
01:01:48 The Future of Venture Capital in a Changing Market
01:07:11 Building a Collaborative Community for AI Development
01:22:20 Final Thoughts and Reflections
Jason Jacobs: On today's episode of The Next Next, our guest is Scott Weller, co founder and CTO of EnFi. EnFi is an AI powered platform designed to streamline and enhance the lending process. Now, it was started by Scott and Joshua Summers, both seasoned entrepreneurs and well known in the Boston Startup ecosystem where I am and the idea for EnFi emerged during the financial turmoil surrounding the collapses of Silicon Valley Bank and First Republic Bank.
And as these events were unfolding, Joshua and Scott were actively supporting their network of startups, which exposed them to the inherent fragility within the financial system and led them to contemplate how AI might be able to revolutionize the complex credit analysis and risk monitoring processes in lending.
Now, it was really interesting to have this discussion for a number of reasons. One, Scott is a multiple time entrepreneur, and he doesn't need to keep doing this, yet he's still doing it. So that was a fascinating topic. And also, they're leaning hard into AI in terms of both How they're building the business internally.
I mean, they have a fleet of agents that are actually writing a bunch of the code, which is crazy. so we talk about that. And, Scott thinks it's only just beginning in terms of the output that's going to come from these agents writing code. But then they also are not only doing this in this emerging area, but they're doing it with these big lenders. Who are inherently, you know, kind of stodgy, bureaucracy, compliance, regulation, and so that's a fascinating topic and the way that they've gone about it is super interesting where they brought these design partners into the mix. And we talked about what the criteria were. To make a good design partner, we talk about the level of involvement of these partners in the process, and the collaboration, or lack thereof, between them.
And then directionally, we talk about where all this is going. Where it's going in banking, where it's going in lending specifically, where it's going for startups, and how these worlds are intersecting. One of my favorite episodes yet, and I hope you enjoy it. But before we get started
I'm Jason Jacobs, and this is The Next Next, it's not really a show. It's more of a learning journey to explore how founders can build ambitious companies while being present for family and not compromising flexibility and control, and also how emerging AI tools can assist with that. Each week, we bring on guests who are at the tip of the spear on redefining how ambitious companies get built.
And selfishly, the goal is for this to help me better understand how to do that myself. While bringing all of you along for the ride. Not sure where this is going to go, but it's going to be fun.
Alright, Scott Weller, welcome to the show.
Scott Weller: Hi Jason, how are you?
Jason Jacobs: I am sick, still. Anyone that's listening to the show, like this episode or the few before it, will hear that my voice is nasally, that's because, I didn't record these all in one day, this is just the same stupid, nagging cold that that I can't seem to kick.
But that aside really excited to catch up with you, Scott, and as we were just talking about before hitting record, We haven't caught up and we're doing it live and the show is not a show. It's just a content exhaust from catching up with people that I want to learn from. And I'm so excited to have this discussion and learn with, learn from you.
Scott Weller: I'm excited to be here. I think this is like the coffee chat we should have had, five months ago that we just haven't been able to get on the books. I'm excited to chat about all the things you've been writing about and posting about. And I'm deep in it right now. I'm really deep in I would say.
The onslaught or wave that would now is this like technology, technological shift caused by a I models. It's incredible. It's fast. It's hard to keep up. But it's fun.
Jason Jacobs: We're going to talk all about that, but where I want to start though is you've been through this entrepreneurial rodeo a number of times and. I don't mean to get personal, but you probably don't need to keep doing this and I know you've got kids and you're in a time of life where you're probably getting stretched in a lot of directions like me and you made the decision to not just do it again, but to do it in this big venture way.
And yeah, maybe let's just start there. Were you seeking it? Did it grab you? Like, how did you think about that?
Scott Weller: It's great. I think about that less than I think people do from the outside looking from the outside. For me my journey has always been Yeah. Like interest led, I was a kid who grew up on a farm. I probably had ADHD. My parents pulled me out of school. I was homeschooled.
Everything I did as a kid was interest led learning. And I think that crafted who I am as a human, as an adult, so I see something that I think is a challenge or a problem. I start prototyping because I'm a coder, I'm a software developer. I see ways of fixing that thing. And I just get sucked into that endeavor.
And and if I was as I was transitioning out of MasterCard, so session on my previous company was acquired by MasterCard. I became a head of product for one of the divisions there and towards the end of that journey, I'd made an explicit decision to transition out and take a break.
But break for me was like, just a way to get. Pulled into something else. And so I,
Jason Jacobs: Didn't we, I feel like that's the last time we caught up. Didn't we catch up when you were either, you had started the break or, and I, if I recall, I don't know that you were jumping to go start your next venture back thing.
Scott Weller: No. I was thinking of so I was trying all sorts of things. I did an experiment where I had made a bunch of angel investments. I did an experiment where I, ran a retreat for first time founders and operators. I was thinking about starting a venture fund around a certain thesis.
There
Jason Jacobs: I think that's the part I remember the venture fund part.
Scott Weller: Yeah, and you said just create a community. I think your advice to me was like, just start by trying to create a community. Maybe the fund will come after that. And so I took that to heart actually.
Jason Jacobs: That's a good lesson though. That's a good lesson that like, the advice you're going to get from people is just whatever they did. That doesn't mean it's the right thing, it's just whatever they did. So it's know who you're at, know your audience, know who you're asking from.
Scott Weller: yeah. And I also often, I often think my advice is not really worth much from my perspective, because of one person's experiences, another person's is could be useful, could not be useful. I don't know. I think I try to be as humble as possible when it comes to giving advice.
But yeah, I'm going to tell you exactly what I did. The journey for me, though, around this how did EnFi come about was really Joshua Summers and I were plugged into the same angel investments. And at that time
Jason Jacobs: that through that group,
Scott Weller: Angels is the group that we're part of.
And through TBD Angels, we had made, I don't know, 45 angel investments at that point. And together, and the, at the same time I was taking a break and we were looking at, I was looking at like an AI company a month. I was, I felt since I had tried the opening, I playground in 2022. I was like, hooked with what they were doing.
And in 23 and into 24, right when I was taking a break. I had a an epiphany that I should be just researching these AI companies as they're coming down the pike and seeing what they're producing, see what they're creating. And I think it was when mid journey launched that I was there.
They're like 2nd or 3rd release. And it was just like, completely blown away with the potential of these new. Models and like real applications for them, outside of being kind of science experiments or research experiments, there's this like transition to wow these things can actually be plugged into products or powering products.
At the same time, we're helping startups after the Silicon Valley bank crash, and we were trying to help startups move their debt. It's easy to move an account these days, but moving your debts much harder. And. We were watching him hit a ton of friction in the process, having,
Jason Jacobs: another reason to stay away from Venture Debt!
Scott Weller: yeah, so I think Joshua and I just asked the question of can this be automated?
And so I went really deep. I put, I put my engineering coding hat back on and I started spending time just going really deep on can this be automated? And so I started going
Jason Jacobs: what was this?
Scott Weller: The activity of initiating. A commercial loan with a bank requires you providing a ton of information about your company.
From the bank's perspective, they take that and then they have to learn so much more about the market, all the things that could affect repayment of that loan. They're trying to come up with, a risk score on that business and It's that activity at the bank, it creates a pipeline and like any pipeline, there's people sitting in that pipeline doing the work and it just creates a stack.
And so banks are throttled on the commercial lending side by how many loans they can move through that stack. And the challenge for commercial, certain types of commercial lending is that there's like an early assessment process when they're working with you, the business. But they have to do almost all of the underwriting up front.
It's a mini underwriting and just takes forever. And so we just asked the question can we do that with AI? Can we actually gather the data with AI, parse the data, go from the documents where most of this data sits? To something a bit more useful and just digitally transform that entire process.
And so the answer was, yes, after prototyping it, we figured we could create a better experience. But then we also started to realize we could bring in our experience, it would data in general, in terms of using public data, enhance the visibility of the market and what this company is potentially worth.
And what's happening in an upturn or a downturn scenario and then ultimately scenario modeling. And so we produced initial products, brought it into the market, and essentially I was re hooked on building a company through just peering into that peering into that opportunity.
Jason Jacobs: Got it. And so it really started with the problem first. And when you set out to evaluate could this be done, you said you prototyped it. Can you talk a bit about that process and how much did you did AI play a role in that prototyping as well?
Scott Weller: I think where I really started going first is I obviously always go back to your roots. And so I just, started trying to Address the problem in the simplest form possible. I didn't really want to overengineer a solution. And I also wanted to explore the newer things that were
Jason Jacobs: That, that's spoken by someone who's likely overengineered a number of solutions historically and has the scars to show for it. Yeah, that, that's what experience is. It's just like you've screwed up enough times that you've learned from your mistakes finally. That's experience.
Scott Weller: I've built plenty of houses without a front door, but they were awesome inside. We I think where I went first to answer that question, where I went first was really to also explore some of the new paradigms that were emerging. As well, specifically around agents, and I had started my early career in massively multiplayer online games.
So the concept of building like non player characters that interface with each other to solve problems wasn't foreign to me. And so I really gravitated towards this idea of designing agents to solve tasks for you. But what was different this time is that we could have the agents write code.
And I think that's the most. Interesting, most distinctive thing that I've encountered as a software engineer and probably one of the most humbling things is that these models have become really great at writing code. And so you can now integrate that into your prototyping, integrate that into your product development methodology and gain speed.
So as one developer, I was able to get started in a way that I don't think I would have. Been able to the last time around when we started the previous company. So the ability to write code prototype using the AI was, is a force multiplier, I think for a lot of developers right now,
Jason Jacobs: Now, I don't want to lose this train of thought on the Early and by prototype and how you coded it and the interplay between the coding it and getting customer feedback. And so I definitely don't want to lose that. I want to take a quick detour just to ask you, do you think these tools scale? Is it just for the throwaway prototype early on that you can whip up quickly or are they going to continue to be as valuable to you as N5 scales?
Scott Weller: I think early on, like we started like trialing packages like llama index and a variety of other. Packages that are out there, open source stuff that, that I think everyone should try. It's so accessible right now to go and just boot up your Mac book, install Python, try to build applications, even if you're.
Not really a developer. There's plenty of guidance out there that enables you to create a prototype right now using AI. So I took the same tactic. I was researching every new possible, AI open source project, trying those out. At the time, when we started the company a year ago or so lot of these things weren't production ready, so you could do your prototype and try to bring it into production.
It's just not going to scale the way, I make, I'm experienced scaling software for large enterprises. And I could see this is really an awesome prototyping environment and have to understand what it's useful for. And it's really easy to make an awesome prototype these days.
That looks like a real product. Thanks. But to take that to production is a whole another, there's a whole another phase of what you need to do. And so for us, like we, we had the N5 prototyping phase, to realize, is there, are there solutions to this problem using this technology, using these models?
And then you'd have to pivot at a certain point and actually rebuild that stuff from scratch. And so we were, for example, last year, trialing a lot of open source. Agent frameworks that we ultimately had to scrap and build our own because we didn't get the level of, accuracy. We needed to get out of those things when we started putting them into production against real production use cases.
And so we've had to rebuild a lot of stuff from scratch.
Jason Jacobs: And do you see that changing as the tools get more mature? I'm hesitant to even ask this because I want to get back to the prototyping for Enfy, but maybe just answer that quick.
Scott Weller: I think there are areas where, I think there's areas where we will have to make, we have made a bet on a specific aspect of our tool chain or our product or platform that we're going to have to build. We continue to build from scratch. I think our agent based system is one of them because we need we need to be able to benchmark and measure accuracy.
We need to operate a certain level of performance, all of these things that we need and
Jason Jacobs: All the agents you've built in house?
Scott Weller: All the agents that we've built and designed, we've built in house
Jason Jacobs: Have you talked to Melody? Sarah?
Scott Weller: haven't. I
Jason Jacobs: Yeah, did you? I don't know if you listened to that episode, but she's helping companies that build agents in house to see in terms of how their end customers are experiencing the are they getting done what they set out to get done? What's the success rate?
What's their happiness? What is it doing what it set out to do? Because, according to Sarah you can't if you build the agents in house, you can't see similar to when web developers built mobile applications for the first time.
Scott Weller: Right. Measurement is such a a difficult, it's a such an important piece like benchmarking and measurement or something that, especially since these are, AI backed, LLM backed in a lot of cases. Accuracy is such an important piece of the puzzle. And so when you build a prototype, it's very easy to build a prototype for like just the surface level of a use case.
But when you actually, I can push that prototype into the depths of that use case, you start seeing accuracy issues. You're like, okay how do we control for these accuracy issues? How do we fine tune? And so I'll have to connect with her to align on what she's doing. It's like on the customer side, they see the accuracy issues and it creates a huge trust.
Jason Jacobs: Huh. I can connect you. That's part of the next stated value proposition is when Double opt in introductions make sense. We put people together and we get out of the way. So there you go. I'll chalk that up as a little win for the next because I'm sure she'd love to talk to you.
Scott Weller: that's awesome. I love to talk to her as well.
Jason Jacobs: So coming back to the prototyping.
So you're using these tools to prototype and maybe talk a little bit about I mean you're building for these Big bureaucratic companies with lots of regulatory constraints and privacy constraints and lots of dollars on the line and this stayed way of doing things and humans involved and inertia.
How did you balance building versus talking to them and how involved were they, if at all in the initial building process?
Scott Weller: We decided to not have customers the 1st year of our existence and only have what we call design partners. And I think this is a common thing in the space right now where companies partner with their customers or their potential customers so that. You can go through a cycle of learning together.
There's a, I think in the banking space or the private credit space right now, there's a tremendous amount of FOMO around, around AI, like the, a lot of executives really want to know how to use AI and they feel like they're missing out, which is, I think they are missing out and on the flip side of that, there's such low confidence.
In the actual use of AI within their environment, and there's a lot of fear around. What if it doesn't work or what if it messes up? But I think there's a lot of I think there's a lot of opportunity and just picking use cases that have a surface area that don't have the level of risk to trigger concerns and problems like, when we enter with a bank, we're picking a use case.
That essentially improves process, improves accuracy on top of something that is generally very human driven, which leads to accuracy issues over time. Because it's such a human driven process and already has accuracy issues I think we're starting a place that where people can trust that the surface area is not touching upon things like, for example, we do not make any decisions for a bank. We are just making it easier to process data. And speed up a transaction flow and by speeding up that transaction, the number 1 priority of bank, I think, generally is to provide a great customer experience at the end of the day and make their, customers happy speeding up a process flow and improving accuracy definitely helps improve customer experiences.
So I think we're choosing something that, low risk. High reward and we've decided it doesn't make credit decisions, doesn't decide anything. It just enables humans to do a better job.
Jason Jacobs: So was this just you and Joshua at the time? What did the team look like in, in these earliest incarnations?
Scott Weller: We earliest incarnations, it was a collection of folks. That we're involved in the kind of TBD angels investment process. So talking about how we could improve venture debt, there was even a conversation of, Hey, should we start our own venture debt fund to help startups?
And, since we're involved in so many startups, maybe we'll fully automate that and that'll become a better way to get access to venture debt. But I think Joshua and I decided to emerged as the co founders of NFI and formed and started the company. And we had a third co founder, Michelle Hipwood, who's a CFO and kind of almost like the financial user of the product who came on board and started really advising on how this should work for businesses, how it should work on the other side of the equation.
It's 3 co founders, Michelle Hipwood, Joshua Summers and myself, and, came out of the woodwork, last April when we decided to raise, our seed round.
Jason Jacobs: In These design partners what was the pitch to them and how how did you go about deciding who the right partners would be and what the right criteria would be in the right partners?
Scott Weller: I think through my experience of, and I'm sure you real you experienced this as well when you were you created, RunKeeper and my climate journey, but like market fit's so important. If you don't have it, you can feel it as a founder, the team feels it. And if so, so we were a bit obsessed early on around just making sure the juice was worth the squeeze, and so even before we founded the company, I think we had 60 plus conversations through connections, through getting introductions, through trying to LinkedIn with people randomly across LinkedIn.
But we wanted access to subject matter experts in the credit space and really wanted to validate. So we saw this problem from the borrower's perspective with startups, but was it really a problem in mass? And was it something that had budgets to go after solving this problem? And so we marched through the industry, just having conversation after conversation to try to validate.
And we would have phone calls, which is listen, we're not trying to sell you on anything. We want you to pick apart this idea and tell and give us the brass tacks and what you think in relation to relation to automating the the commercial lending process or the risk management process.
And in that, we found a burning need, and it was different than what we thought it was going in. We also found a handful of executives at banks who had the reaction of Not only do I think this is, something that should be solved. I want you to solve, I want to solve it with you. And that's when we realized we should create these design partnerships because we didn't really know the total trajectory of the product.
The domain experts were saying, yes, this is a problem. And the ones that were saying, let's do it together. We felt we could definitely learn from each other faster. So we ended up forming some business relationships early on that. Brought these banks in as customers, but they really were there along the journey to help influence the product.
And if you get 2, of those, you start seeing the industry problem. And if you just have 1 of those, maybe you're designing for 1 company, which could probably be bad for a startup. So we also made sure that we're trying to get a wide enough aperture of these design partners.
Jason Jacobs: And don't disclose anything that you don't want to disclose, but how are these structured? Does any money change hands and also for any other founders out there who are building similar products that they're trying to sell into? These large enterprises and want to know early on, like you did is the, what did you say?
Is the juice
Scott Weller: juice worth the squeeze.
Jason Jacobs: Yeah, is the juice. Yeah, I've never said that before, but it sounds good. I think I'll say it again. I'll steal it with attribution. But but what advice would you have for founders as they might aspire to have similar types of design partner arrangements?
Scott Weller: I think there's a match between you first need a partner is willing to lean in, meaning dedicate someone a champion to the project or to the process. And so they have to have the bandwidth to engage with you and the willingness to engage with you. And so that filter probably reduces a lot of the companies you could work with.
You also need executive like an executive kind of, top level support as well as an owner within the organization. That's also hard to get but you do need that sort of barbell of executive supporting the initiative as well as there being someone in a role that's going to use the product, use the software every day.
That's also leaning in and sees the value. And so we spent some time trying to find that combo. And feel pretty blessed that we did in order to bootstrap the company advice I would have for founders. It's always a quid pro quo to establishing these like enterprise software relationships, even if it is a design partnership.
It's one of the things that we asked our design partners was in a time investment. So we we asked them for a three year contract. The 1st year, we would lean in and do this design partnership thing. Certainly it was metrics driven. So coming out of the 1st year, there'd be some success metrics.
But as long as those are tripped, we move into, a 2 year contract. It's relevant in the banking sector, because I think. Finance, nothing moves quickly because it is a regulated space. And so that time commitment is helpful to the startup because now we know, okay we really can lean in. We can put our resources against this. And it also demonstrates. To the startup from the corporate side that they're also, making a significant investment in terms of time. And so there's plenty of companies that would say no way this is unknown. We're not going to sign a 3 year relationship and those just aren't the companies we could work with at the time.
And but there were a handful that did say yes, because why we're in a time where they wanted to explore AI. They want to find efficiencies. We had, built a crack team of experts to go do it. And all of that aligned. So my advice is push on the levers that establish a healthy quid pro quo in terms of time investment.
Jason Jacobs: Now, were any of these design partners competitive to each other?
Scott Weller: I, we tried, I would say we selectively tried to navigate along the lines of loan categories. So if you, I think every bank and lending institution is competitive with another lending institution, but for our design partners, they their bread and butter sat in different loan categories.
And so I think that kept. Kept the design partnerships noncompetitive. But I think for the most part that lenders compete. But some ways, yes, they do compete and other ways. They do not because they're going after different types of businesses with different types of lending lines.
Jason Jacobs: Were they sensitive to who else was involved?
Scott Weller: We never ran into that. Actually we in some ways, I imagine a world. Where we are building a community of AI enabled portfolio and risk managers for lending, and we'll sit them around the same table in the coming year and they'll collaborate with each other.
Jason Jacobs: I was going to ask, has the interface to date primarily been siloed each one with you? Or is there any Collaboration going on with each other.
Scott Weller: today it's completely siloed. It is siloed on purpose because there's data silos, there's information silos, there's confidentiality silos. But I could imagine, I do, we aspire to create a community of AI enabled portfolio managers and risk analysts, and we want to bring them together, maybe even this year to have them sit down together, talk about strategies, what's worked, what hasn't worked other tools, other AI tools are using across their enterprise or helping in other ways, other areas and I think that's important because the confidence still is very low for these for practitioners in a lot of spaces, but in finances in particularly it's just been slow.
There's been a slow adoption curve in terms of where to apply and just for any given knowledge worker in the space their exposure to AI has been opening eye, or it's been Gemini, or it's been like the more common models that are infused into kind of the productivity products that are out there.
Okay. Outside of that, they're not quite sure where to go, what to do, and they're constantly being hit by so many vendors. So hopefully we can create a community that can collaborate with each other in this very specific domain space. Like, how do you answer the question of, should I trust OpenAI's deep research to do research for me at the bank? You look at the benchmarks. It's like 55, 60 percent accuracy. If something is returning an accuracy rate where it's wrong 40 percent of the time, is that useful? Maybe it's useful, but maybe it's not. And so bringing people around a table to discuss that openly with each other who have the same domain experience, I think could be really powerful.
So it's something we're going to do this year and we're trying to do it in a non competitive way so they don't feel like they're sharing their secret. They're sharing their secrets with each other.
Jason Jacobs: So I thought we were going to getting beginning into building agents and how to select whether to do it yourself versus off the shelf and how to get them talking to each other and what the hard parts are and what the surprises are and all that, which I still want to cover. But I'm fascinated by these design partners.
How do you find a common elements across so that you can confidently build and satisfy? Oh, I would imagine, especially if they're in different areas of lending, that they had wildly different needs. Was that a challenge when determining what functionality to provide in a V1?
Scott Weller: What was challenging is our design partners came on in a staggered way because like when you're building a company, nothing just happens at once and you're hustling as a founder, doing founder led sales and things come in. So our first design partner started six months before the next one.
And so you're seeing, things through that lens and I think we were always wondering is this is the feedback we're hearing applicable just to this 1 institution? Or does everyone have this problem? So we kept doing that flywheel of talking to subject matter experts.
We didn't stop the the approach of getting on the phone. With a subject matter expert, telling him to rip apart our idea and put their hopes and dreams back onto the table. But I think, once the, the 2nd design partner hit, we started seeing commonalities and where we saw commonalities are really in how humans process these loans, the workflows.
Across the institutions are very similar. The type of data they're using to make decisions very similar. What they're asking of ours very similar across multiple loan categories. And so we started seeing these patterns and that started that drove us for towards how we were designing the product, how we were designing things as a founder.
We're not done. I think we're in the very early innings and we're still learning tremendously from every customer we bring on to the product. And we'll continue to look at it. That's a great, it's hard. I think we're still haven't hit a sample size where we feel like we're at the industry, but we certainly are beyond just one, one voice helping us drive the evolution of the product.
Jason Jacobs: So when it came time to start to to actually bring it to life and automate these human tasks, what did that look like? What did that look like in terms of where to start? What did that look like in terms of what tools to use? And just any color you can provide on how that journey's been would be super interesting to hear.
Scott Weller: So a couple of interesting things about that journey. I think there are You know, there's the speed at which the newer models have entered into the market and the advancements that have occurred, I think is unparalleled for most of the people. Most of the even experts sitting around the room, like the people that you would hire.
So when we started building the team and we're going out there to find experts, I use like this drift or lag on anyone's ability to be up to speed. Unless you've been sitting at OpenAI or been sitting in a research institution and at the exact forefront of these models, you're lagging.
And so that like Delta is something we had, we have to deal with. And as we started hiring the team, it's, that, that sort of experience gap with these models was something that we needed to address in terms of how we built software together. And the agile software development process is the first thing that kind of it sits in front of your ability to actually create an environment of experimentation. So we had to, we had to like. Almost enforce an environment of constant experimentation so that our team could learn faster around where we should be focusing our time and energy. And so just give you an example, like our 1st prototype. Involved building essentially a traditional data processing pipeline on AWS using standard like it was more of a, I would call it like a data science approach to data processing and it prove some things to us.
But it wasn't an agentic system by any means. And so we had to scrap that. We spent like a couple months developing and building on top of like a traditional data pipeline architecture. And then we scrapped that and we brought in our initial approach to agents, which is much more real time and far more reactive.
And so what it was a real learning moment for us though, to recognize that, as practitioners, we had to up our game in order to be on top of the state of the art. And what that meant is, coding during the day and getting your features out and building your features in that night, prototyping, learning, reading, assessing, and bringing that back into the environment.
Everyone on our team for the last year has just been heads down working 24 by 7, but living in these. In this, in these 2 worlds, 1 of like researcher, prototyper, experimenter, and in the other world of oh, I actually have to put stuff into production. I have to build software. And so that duality is was challenging for a lot of the engineers that came on to the team early, but they've embraced that culture and we're still trying to figure it out, to be honest with you.
We bring on a new engineer generally has a certain experience, certain background, et cetera, but generally often has a gap with the newer state of the art. In the industry, and we need them to catch up. And so they also have to go through that process as well.
Jason Jacobs: It's funny. So much of the chatter and AI is about what are we going to do when the AI is just doing everything? But the way you're talking about it is that it's so hard to stand top of that it almost increases the workload because you can't assume that the tools you're building with are static.
The opposite of that. They're moving under your feet faster than ever. And so it's a full time job just to stand top of it. How as an organization, Can you have processes that do both, that get all the work done that's on your plate on the hamster wheel but also stay on top of that without going at an unsustainable place pace and just torching the whole organization by burning everybody out.
Scott Weller: That's a great question. I think when you join a startup in its first year, we're just over a year old. A lot of it's like setting the early on just making sure that people coming on board are know what they're signing up for and that there's incentives. But one of the incentives of joining an AI first company is that a lot of the people who filter through that process, they want to spend their spare time learning and getting ahead the same way, someone professional sports would spend their spare time outside of the games.
Weightlifting, eating optimizing kind of their regiments so that they, when they get on the field and there are in the game, they're actually performing really well. So there's plenty of analogies out there to say your job doesn't start at the job.
And this is a moment in time where you have to reach the limits of your intellectual capacity as a software engineer, and that's going to require you to study. And practice outside of the game, and that's what we ask of people. And so that's like a self selection process that people come through that filter, want to do that because they want to advance their careers and they want to get ahead.
A lot of them already have somewhat of an AI or data science background or a data background and automation background. And so they've got some experience to put to bear. But I think what everyone in the industry is unprepared for is the speed at which things are moving and where to find tactical advantages.
And it's stressing current methodologies. When we first started the company, we were like, yep, we're going to be an agile shop, we're going to have these specific ceremonies and a cadence. and we still have that, but we have ways of, we created ways of violating that enables the team to surge and innovate in a way that's fundamentally different than a standard, agile scrum process.
And the reason we have that is because someone can come back from research and be like, Oh, I found this really interesting thing. That's going to accelerate our agent system. And it's going to give each agent memory and with memory, we'll be able to do X, Y, and Z because this paper was just written.
And we want to implement that paper and test it. Now, we certainly could like, plug that into the sprint process, or we could allow that team or subgroup to go off, crush that in a couple of days and come back with some results. Thanks. And so we've got to create this hybrid process that enables people to learn, put their learnings into, production, essentially do some testing.
And then we can rationalize what that means later on for the rest of the development process. So
Jason Jacobs: and do this experimentation, there's so many different tools out there and it, I mean it, a lot of it seems like just comes down to personal preference and I've been hearing anecdotally and I'd love to hear what's going on inside and five, but that there's not a lot of standardization today that different engineers have a different stack, if you will, of tools that they prefer.
How do you think about that consistency at N five? Does it matter? Does it exist? Does it need to exist today or directionally? And what advice do you have for others trying to figure out that same question for their organizations?
Scott Weller: what I'm about to say is not meant to disparage all the great, awesome work that's happening out there with different types of products, tools, packages, et cetera, especially in the agent based systems community. There's some awesome open source projects and some awesome commercial companies emerging, but for us, we've had to keep we've had to stay as close to the bare metal as possible to control for 2 things. 1 is our own ability to move quickly, but also our ability to assess accuracy and do benchmarking. And so if we, for example, leveraged a off the shelf. Agent system there's a bunch out there, whether they're open source or commercial we're just introducing a lot of middleware essentially in our environment that we don't understand, and it's hard for us to tune for.
And so we either can commit to have engineers learn that and spend the time to learn it. Or we can build what's barely like our bare necessity for what we need. And all of the things around that, that then allow us to measure for its accuracy and measure for its. It's consistency. And so I think our biggest learning is that we had to scrap a lot of stuff that we thought was going.
We're going to be forced multipliers that were essentially open source packages and things that we thought were going to be accelerators. They were accelerators in the prototyping process. But when it came down to the brass tacks of operationalizing the system doing benchmarking, iterating quickly, like they became inhibitors.
And so we have to scrap that and build from the ground up and only take with us exactly what we needed for the job, the mission that we're on. And that would be my advice to a lot of startups is you don't necessarily need to be a build at all environment. And we certainly use model off the shelf models, for example, that are that we don't train, those are very helpful that we don't have to train those models, but there's other things that we're doing that we have to build from the ground up.
And so one of those things was our agent based system. Like we started with a, when we rebooted our approach we started from the ground up and started building very simple mechanisms to run agents. Very simple mechanisms where agents could have tools where agents could have sensors for their environment and a very simple framework, a very bare bones, lightweight framework for our engineers to start iterating on the development of agents.
And it was so simple that it enabled innovation, enabled like thinking outside the box. There wasn't the weight of having to learn. Like a large framework that someone else had created, and so it actually created speed for us internally and from day one, we could build in the underpinnings for one of the most important pieces, which is benchmarking, like the ability to measure the effectiveness of your agents over time so that you could communicate to your customers and say, here is the accuracy of this particular task or this particular thing that we did, which is so important in our environment.
Jason Jacobs: So as you were heading down this path at what point did you start thinking about capital and to the extent that you can share how many rounds have you raised, how much have you raised, what kinds of investors you have at the table and at what point in the process did those intersect the capital and the traction, if you will.
Scott Weller: For us, we're really blessed in the fact that we were in the capital environment when we started the company as angel investors, we had a lot of, I'd say discussions with contacts and friends around what we were doing and getting their feedback early on. I've learned in the fundraising process over several companies, like the fundraising event happens after you've partnered with someone for a period of time. And there are plenty like capital sources, VCs, angels that want to be involved beyond just the capital money. They want to know that they can help catalyze success for the business in some way beyond just the capital.
And so always start your conversations with Your pitch starts with just getting feedback from some trusted sources, I think in a
Jason Jacobs: Just like with the design partners.
Scott Weller: Exactly. So
Jason Jacobs: noticing a theme. I bet that's how you hire too, Scott.
Scott Weller: certainly is. It's a it's not natural for everyone to have relationships with VCs. And so I know for a lot of founders it's one of the reasons why I You know, Randy, the Seguin venture summit was because a lot of founders just don't have capital networks. They don't have connections to people with capital.
But when you do, it's the normal experience for founder is get introduced to a VC, do your pitch. And it's like very kind of tactical and functional, right? Where. I think success for a lot of founders is go to a VC and say, tell me I shouldn't spend the next 10 years of my life doing this and tell me why and build a relationship.
And the VCs that really lean in are probably the ones that most interested in potentially investing in you and supporting you. So my advice for a lot of founders would be start your journey that way. Start your journey by trying to find VCs who are going to.
Jason Jacobs: And so at what point did you raise the capital for EnFi? And I know that was the lagging indicator of the relationship but what did EnFi look like when that capital came in?
Scott Weller: So N5, we when the capital came in, we were a very small team. We were at prototype stage. We had one design partner. We had signed a, we had signed that initial contract for that design partner. So we were off to the races in terms of really building out and scaling the product for that design partner.
We're working on our second design partner at that time and conversations. And we could see a pipeline building. Of. Interest in using this type of product and so we had that pipeline building which allowed us to talk to potential the potential of this product. We had that prototype. And I think that's when we really started in earnest, having serious conversations around funding because we were watching.
The whole market take off and, every day, like a hundred companies being created in the AI space. And so we felt if we weren't well capitalized, we might as well sit it out because it's really difficult to be an undercapitalized AI company. At this, in this moment, in this moment that we're in, especially when the largest foundational models are raising 60 billion dollars.
And so we felt we had to have a substantial fund. We'd have to have substantial interest, substantial market fit signals. And we're gonna have to raise something and take more dilution than we normally would take early on in order to really see this business, launch. And so we went out to go raise 3 and a half million seed round. After about 2 to 3 months of that process, we had. Generated about 9 million dollars in interest and as founders, we felt like that was a bit ridiculous. We felt very blessed. We had that much interest, but we paired it back to 75. We ended up raising 75 last April. So very substantial seed round. But at the same time, when I look back today, some of the expenses we've had.
The cost of, operationalizing AI, the cost of hiring resources don't think that set that 3 5 would not have got that 3 5 would have put our company into some sort of founder engineered trough of despair. That would have been unfortunate for the business, if that makes any sense.
So raising the seven, five actually put us on a trajectory of getting our product live, getting additional customers on the product and getting to a point where we felt like we could viably scale, the business. So not every founder has the opportunity to raise seven, five seed round.
But I, when I look back, the three, five wouldn't have got it. They wouldn't, it just wouldn't have gotten us here.
Jason Jacobs: So there's a lot of talk in the AI world about how, the, don't try to compete with the models and the models are on a, on a path to, I don't know if commoditization is the right word, but like they're going to duke it out with each other and that chip has sailed and that the applications are where it's at, but on the application side, because Defensibility is harder and moats are smaller and competition emerges faster, maybe try to get by with less capital, which you probably can because you can, smaller teams can go further using these tools to build, but it seems like the general feedback is that capital light should rule the day, but I'm hearing the opposite from you.
And so I'd love to understand. As a potential founder who might be evaluating building a company in the space, how to determine when it's you should fall on the capital light side and try to raise as little as possible and when you should try to raise a war chest to ride out the storm, otherwise don't get in the game.
Like, how do you know which one you've got on your hands
Scott Weller: That's a tough question. Might require another. Three hour podcast session. My reaction to that question is just instinctually, I feel like it's got to be based on the market you're tackling and the demand signal and. And your access to your user. So I think something like cursor, for example, I think is a great example of a product that has an organic demand signal because it knits into the life cycle of every software developer. It's installs on your self installs on your laptop. So your distribution model is actually directly correlated. It's tightly correlated just to interest of a developer wanting to improve their productivity. So they just need to find you or you need to find them. And then that developer can walk up.
So very few barriers, the biggest barrier there is can cursor get in front of every developer in the world who might use VS code essentially, which is what they built their product off of. That's a lot different than saying we have a verticalized AI application for air traffic control, right?
Where just imagine trying to break into the air traffic control business. There's probably three vendors that serve that entire market. They're probably government contractors. It's just big and it's hard and it's hairy. But you, maybe you're going to build the best AI air traffic control system.
I don't know. But that that I think is such a different, and so for that endeavor, I think you need backers who are willing to go the long haul. They see the big opportunity other side and a lot of the risk, but they're willing to do it because they believe in the space, believe in the opportunity.
And so I think you have to go a little bit bigger. When you're going into spaces that are slower, there's bigger walled gardens in those spaces, but, you're serving a different type of user who acquire software differently or acquires product differently, and you may have a bit more friction.
And so at the same time, too, I think. I think you have to look at your R and D, like how much R and D you need to put in this product to really make it work for that market. So for financials, like we're purpose building a system for financial services. Like security and privacy first and a lot of other tenants around FINRA and all these things that we have to care about in order to check the boxes and be relevant for our buyers.
Whereas if your cursor and you're just plugging into it like an I. D. E. Relevancy is like how well can you generate code for me in my user experience of writing code? This is like fundamentally different. So I think that's also a characteristic of a sort of what are the what's the overhead to getting into this market for this product?
We purposely saw a problem in a market that has more overhead than other markets. And I couldn't in good faith. Incubate a company that didn't have enough capital to overcome some of those hurdles. It's almost like my fiduciary responsibility to myself, co founders and our now investors to make sure that we're well capitalized to go after and do that.
And now I mean taking more dilution early on, but it sets us up for success, in theory, farther out. Whereas I do think, we will have companies of two developers who create a great product where the barriers are much lower. And they're, they don't have to be as capitalized to go into those markets. And so I would say find the problem first. And then maybe your gut check second is is this a market I want to live in? Like the idea, the problem is interesting. Is this a market I want to live in for myself? I was like, I'm up for the challenge of working in financial services. I've been there before.
I understand that marketplace, whereas someone coming in completely cold might just feel it's like not a very attractive space to operate given some of the overhead. Now, what are you thinking? Like, when you think about this for yourself are you looking for low overhead, like low overhead, low barrier opportunities?
Is that one of your like criteria you're still learning? We're just curious
Jason Jacobs: I'm still figuring it out because I say that. Flexibility and control are paramount. I look at all the crap that I have on my plate and it's not crap. It's, it's like stuff with my parents, stuff with my kids, like stuff with, trying to stay healthy and get in shape and just just more stuff than I'm used to having. And it's I don't know if I want the stress. I don't know if I want to manage a big team. I don't know if I want to deal with all the BS. But then at the same time, it's I'm like a moth to a flame of the most ambitious stuff. And it's like, all right like, how can those coexist? And that's the journey that I'm on is figure that out. It's like, all right, I've got this massive ambition, but also relative to any other time I've built in my life, I have a lot more constraints. And and so how, what is the right recipe?
What is the right rhythm? What is the right team? And I think it's like small team. Doing really cool stuff in flow state a bunch of the time, not having to do, a bunch of the things that I don't want to do having a lot of flexibility and control around my schedule and having a team who likes to work the same way, but then find a way to build products we can be proud of, serve customers in a way that delight them and hopefully have some purpose and a mission to it that, that feels good as well.
But I don't, would that be enough? I don't actually know because then, when, what fills me up with air, the most ambitious, the heart, climbing Everest is what fills me up with air, right? And so
Scott Weller: Yeah.
Jason Jacobs: I don't know, Scott
Scott Weller: I feel like we, a lot of times entrepreneurs feel like they're always made for more, always trying to solve, bigger problems. And, whether it's climate change or whether it's curing cancer, like there's some big problems to solve out there.
Then I think the for our time here on earth, why not, like, why not go after the bigger things? I think we tend to see, I tend to look at things as projects. It's the only way I get by in this like massively stressful world where we have a lot of responsibility. I think we tend to. Underestimate the force capacity and leverage we can create for solving that bigger problem by just starting with the smaller steps. And so we're like how could I possibly do that? Really challenging thing today as who I am. And with the leverage I have right now, the answer is you can't, but you certainly can start and over time, if you're good at creating that leverage and that force capacity, you certainly can actually accomplish that really hard thing.
But I would say, just get started get started and see where you go from day 1. If you want to have a certain lifestyle. You want to have a certain kind of like approach to solving that problem. Set those constraints. Now build that as the foundation of what you're doing. It doesn't mean you cannot achieve that really, challenging thing.
It just, you have to do it in between. And it means you need different structures of force, capacity, and leverage in order to go. Accomplish that thing. And so I feel like if you have certain like desires around your lifestyle and you also want to go do big things, just set that tone now, because what you'll end up doing in the first iteration of that project will be building the infrastructure you need in order to have both of those things.
And it's really hard on the other side of that because I think a lot of founders just go and they race and they make huge sacrifices and they sacrifice I, like for me personally, the last company, I sacrificed a lot of things. I sacrificed time at home with my kids. I sacrificed family time, it was time I was on an airplane, 180 days a year to go, see customers and do customer implementations, personal health sacrifices.
With this one, I have like very clear. Guardrails around the things that I feel like are important in my life. And I just recognize like I'm going to have to build different structures to create leverage to achieve the bigger goal and maintain some of those things. So guess what am I might need to be more ambitious around how I use my time and like more cute.
Where I think, the past company, I spent I wasted a lot of time did not even realizing I was wasting that time. And so with this endeavor, like time is more precious. And so the way I use time is different, but I think it is achievable to go after the bigger things.
Jason Jacobs: but, the concern is that it's easy to say that when you're flush with cash and you have plenty of runway, right? But but let's say things go sideways with the design partners and and you don't have the metrics. To do the big next round and you're running out of cash and the insiders are capped out and you have all these people on the payroll and you feel a duty to, to, to be the steward of the ship in the storm and are all those things that were uncompromisable going to all of a sudden be up for grabs because they have to be for the company to survive.
Scott Weller: I think probably. I think an emergency room operates differently than a hospital, right? And there are moments in our lives where we have to be the emergency room. And we have to do triage. And it means 100 percent intense focus for a period of time. One of the things I realized at the last company is sometimes we treated things like the emergency room all the time.
And it didn't need to be. The emergency room and it just we did that because they had a lack of experience. Some of the issues we address definitely were emergency moments and definitely required 100 percent of our time and sacrifices. And, hey, I can't go, I can't go do this with the kids, or I won't be able to be here after work.
But I think fundamentally I learned through that experience that you can be more prescriptive around what, how to address an emergency versus how to operate business as usual. With your business and if you're in emergency mode constantly maybe it's time to reflect, like starting a project get yourself out of that for a moment, get advice.
Maybe you're over the tips of your skis and you need to hire someone to do your job. There's a lot of ways to address that feeling or that condition. And if you're out of cash. That is a moment that every founder I think has experienced at one point in their journey.
And yeah, you're going to have to, you're to put everything you can into making sure that you find the right outcome for the business, the investors, et cetera. So I think it's what you're signing up for. In some ways, when you decide to create something from nothing and you decide to bring in partners with their money and their time into the endeavor.
Jason Jacobs: Yeah it's I think that, that's why whatever I end up doing and, I might already be doing it because there's ways, there's paths where this just, you know, this and a business model could go a long way but but I think, right now I'm just focused on learning and relationships and, build, building a tribe of like minded people and informal collaborators and it goes where it goes.
But I'm sensitive to time my, I got just a few more questions to tuck in. One is if you look over the next call at 12, 18 months, what are the key priorities for ENFI?
Scott Weller: One of the things that I, we go through these innovation cycles, like we changed our process and we have a champion every week who on the team was going and experimenting new stuff and bringing that back to the team and reiterating based on some of those learnings.
One of the most recent learnings we've had, I think will play a big role in our technology. This next year and so we were able to implement agents that actually write their own code for the product itself. And so we did this experiment over a weekend where, since we have this agent system, we were like if it's producing data and documentation it's completing tasks for these workers.
Why can't it complete tasks for us? And so we gave an agent the persona of a product manager. And we gave it a set of tickets and we said, we want to implement new charts and graphs for our analysts in the analysis tool and finalize this tool and in basically within 20 minutes, we had the feature built.
Jason Jacobs: Wow.
Scott Weller: And we were able to display charts and graphs in a different way for the analysts right in our tool. The front end code, the back end code. And when you look at this code, you're like, it's not rocket science. This is didn't require research. But this ability to almost seed a product idea with a core agent whose responsibility is to start building out the basic code base with the influence of the developers and the product managers to influence the direction is something that's different than how we traditionally build software. and to see the results so quickly maybe realize that agent systems on the cutting edge are agent systems that write and execute their own code in real time. And that's substantially different than where agent started, which was a bit more, I would say, deterministic, less dynamic.
And so I think, the future is for NFI is going to be agents that actually can solve problems that we did not anticipate. So the analysts can come to the NFI dashboard with a new research task or a new analysis they want to conduct. And if we haven't designed for that as product managers, engineers, they can instruct the system with these primitives on how to do that.
And if the system can't do that, it's going to write some code to do it in real time. And that's going to extend our ability to accommodate use cases we would never anticipate, never see that are highly relevant to that user. Which is awesome. And it's just like blows my mind that we could be at a place where the user could inform the software and how it could write itself to be better for them, which is just nuts and awesome at the same time.
Jason Jacobs: So is that going to have implications on how you hire, how much you hire and capital requirements for the business?
Scott Weller: I think so. One, I think it's going to enable us to hire individuals with more progressed experience. So the types of folks that we want guiding that type of system are individuals who have a ton of software development experience. Who have seen a bunch of, generally on how to operation, how to operationalize and write for us a real time system that performs like this.
And unfortunately, it means it probably means we hire less junior engineers over time. I don't know what that means for the industry in general, in terms of this knee capping that can occur where everyone just wants the most senior engineers, most experienced engineers. And it's really hard to get in.
To the industry.
Jason Jacobs: and yeah, and then, and also what's the farm system for future senior engineers?
Scott Weller: I think that's going to happen. I honestly do people worry about AI taking jobs. And I think, one of our missions is eliminate the mundane tasks and activities. As a, as an early software developer, I did a ton of maintain, like ton of mundane tasks and activities. That's how I learned, right? So we all generally learn in the spaces we were in.
But I am concerned around this kneecapping that occurred, whether it's in financial services, whether it's in software whether it's in accounting do think like wherever there's knowledge work is this potential to eliminate the need for the entry level? But over a five year cycle, every industry will lack the mid level experts. And so what do we do? And so maybe that's why the the largest foundational models are racing towards AGIs because they like ultimately feel that with AGI or some concept of AGI, it'll they'll have true market fit for AI, you won't need a practitioner to deploy AI.
It will deploy itself. I'm getting really esoteric though, and I'm getting into a space that I haven't thought a ton about, but I think there is this reality around a kneecapping cross knowledge work that we need to design for. I don't think as a society we can accept not having junior people move up through the ranks and learn and experience the things they need to experience to be experts.
We need that. We need those farm teams. We need that farm. We need that process. think organizations are going to have to design themselves for that.
Jason Jacobs: Gosh I wasn't planning to ask this one, but it just makes me think, what are the implications for venture? If you If you were a VC deploying capital in with this emerging landscape with this accelerating pace, how would it change how you deploy that capital and what criteria may make a good investment?
Scott Weller: It's a great question. If your users are changing, if the market is changing really fast then it might, ICP like an ideal customer profile and market fit also might be changing really fast, right? In theory. And it's I think for a lot of VCs, they're going to have to do a lot more data crunching and research. to hone their thesis and the idea of finding market fit might, might require more ongoing support. So VCs have to be ready, I think, for these early parts of these businesses being just more capitally intensive when it comes to figuring out market fit and figuring out the go to market aspect of the company.
Like I for example, the. The whole construct of software as a service is fundamentally changing. So this idea that, there were a lot of VCs that built their thesis around SAS because it was a predictable financial model in a chaotic environment. I think that's really shifting quite a bit.
I know a lot of companies are gravitating more towards transactional pricing or pricing that's on demand pricing versus buying capacity that they haven't used yet. And that's changing the dynamic of market fit. Whereas assess. Software provider could find solace in in yearly subscriptions and in tiered based pricing.
Now you're held to the whims of being much more transactional, much more on demand, which reduces your. Predictable revenue coming on the other side of that. So I do think we'll be Caesar going to have to adjust how they measure market fit and how they adjust how they view the revenue potential of a company.
In this changing market,
Jason Jacobs: One aside, Scott, that that the work that you guys have done around getting the agents to write the code and build features is that I'm not suggesting at all that as a startup, you get distracted from your laser focus. But is there IP there? Could you productize that? Could that be a company itself helping other companies deploy that
Scott Weller: about that because because we've been very bare metal on building these things as we need them, like the first, before we even got to can we make this right code? We really focused on, can we stimulate, can we fine tune our process and model use to simulate the steps required to solve any complex task? And so we designed and optimized a model that produces a plan. So now we have these planning agents. You can give these planning agents, and it's for knowledge work, and we've constrained it to financial services, knowledge work. It could be any kind of digital knowledge work, but they produce. A state machine that then can be executed to solve the steps to complete a task.
So like a simple, like just think of the most simple task agent task might be I want to book a, I want to book a reservation at a nice Italian restaurant in Boston tomorrow night. That's a prompt I could give one of our agents. The planning agent would go off and say, Oh how do I solve this problem?
And it would ask the first question what tools do I have access to and what data is in my environment that I can leverage? And so the first thing he does is it creates a bunch of plans and it's, and it just tries to figure out which one of those plans is probably has the greatest degree of success.
And it chooses a plan that planning agent comes back and says, I think this is the best plan. I think I should go do a Google search and then find the top five Italian restaurants in Boston. And then I should go to Rezzy and I should use the Rezzy API to figure out if they are on Rezzy. Oh, and then if they are the subset of those I'll check times.
I'm going to send A user input request to the user asking what times are best for them. I'll try to align them the available times. And if the user picks one of those times, I will book that. So just like the idea of creating that optimal plan with those tools. And then measuring its success is something that we had to solve.
That alone probably can be its own company for any industry. In the agent based space. So you do your question, like sometimes we pluck these things and we're like, wow, is that a bigger idea? I'm not sure. Is that a more, is that more applicable to other people building agents? Probably. At the same time, who's the user? I think when we're solving this problem for, in financial services, for a portfolio manager and a risk analyst, like we really know who that user is and what their needs are, and we can fine tune. That planning agent to be really great for them, right? And so I sometimes I do think the focus helps.
But there are ideas that kind of jump out that sometimes we're wondering, is this a bigger, is this a bigger play?
Jason Jacobs: that do I have time to ask you one other question?
Scott Weller: We have plenty of time.
Jason Jacobs: Okay, yeah, and I don't actually care if this is boring to listen to because I'm just learning so much but But here's something I've been thinking about. I'm increasingly, because of the show, And because I'm, I'm like posting about this stuff, thinking about it all day long, putting idea starters out there on social media I'm getting a, I'm building a tribe of people like you that are smart, hungry, intellectually curious, getting their hands dirty with the tools, and at various stages from just coming in, to actually where you guys are, which is getting a lot done, have a bunch of learnings, have already scrapped, four different, four different ways of doing things, have iterated to a way that's now working, at least for now, until the landscape changes under your feet, right? And it's changing so fast that it's hard for everybody to keep up with, right? Therefore Collaboration will get everybody there faster.
And so there's incentive to collaborate at the same time as ENFI. What's your incentive to share anything? Because your knowledge and those hard earned battle scars are your competitive differentiation in a landscape that's, where moats and defensibility are harder and harder to come by.
And so your moat and defensibility one of them is just experience. Like, Why are you going to give away all your experience for free, right? And one of the things I'm thinking about like with MCJ in the early days, We had all these listeners on the show. They were clamoring to sort through all these to have a peer group to sort through all these nuanced topics.
They came from wildly different backgrounds and skill sets. We just set up a slack community, right? And and they took on a life of its own during the pandemic where there was all, there was all these like brown bag sessions and book clubs and study groups and people teaching each other stuff and knowledge sharing.
And it was like really helpful and inspiring, right? And what I'm wondering is as the next audience and tribe grows, what does that look like? Because it doesn't feel like it looks like just another slack or discord, right? Especially there's all these tools getting built. Is there a way to actually enable knowledge sharing, tool sharing, like almost like modularizing or like pulling out little clumps of things that people like, what you guys have built for N5 might be super wrong for another company in unrelated space.
Is there any incentive for you to share that, if you could get stuff back from them? And could we build a community of people that are just sharing stuff with each other, right? I'm no open source expert or anything, but I'm just I feel like there's something here and I don't know what it is.
Scott Weller: I think about this all the time because I, within our domain space, I like to walk around saying we are, we genuinely want to create collaboration between portfolio managers and risk analysts that sit across all these different financial service institutions. And up level their knowledge on how to use AI, whether it's N5, whether it's another company, but the future are AI enabled knowledge workers, no matter what happens, unless an asteroid hits earth, which could happen, right?
But low probability. So take that aside, take a disaster aside. You're going to have to be a, an equipped AI expert in your field in terms of what you do. And like in the future, it's going to feel very similar to using Microsoft word and. And Excel, the tools are going to just start feeling more and more familiar engaging with a digital, a digital coworker is going to start feeling natural.
It's just what I do as part of my job and dependencies will be built. So I think community is so important in terms of up leveling. And that's why early on, I mentioned one of the hardest things we had to do is like up level our skill sets towards the, if you're, if we're surfers, we want to find the right sets to surf.
And we want to be in front of the waves that are the best waves. And not only does that work, that requires a lot of things, it also requires us to have a tremendous amount of experience surfing. And the current way that software is built within a company is almost designed to slow down. It's designed to create predictability, which is fantastic. But in this moment in time, you have to find an augment to that, to uplevel everyone's game, and I think community, guilds, getting people together, to your point, is how you do it. To the extent that like a small company doesn't feel like they're giving away their IP I'm totally up for making sure we can pull the right people together collectively, a group uplevels their game in this environment.
I often think What else does a company have or, there's two things a company can have. They either can have ongoing success and press and occasional notoriety with their customers and, or they can have a tombstone. And most, most, in most of history, your existence is I would say, tracked by your tombstones, right?
It's like how we learn about our ancestors. It's oh, there's a tombstone. What was that tombstone I think Unless there was a community around that activity. The community is what kind of persists the evolution of that thing over time. So yes, we should build a community. Yes.
It should be AI enabled and yeah, sharing tools. I don't know what this community is, but I think you're on the right track and some at some level, it helps everyone exist beyond just their future tombstone and allows you to share and be better at what you're doing. I don't even know if that analogy works.
But that's that's how I think about it.
Jason Jacobs: Huh. Yeah, maybe for starters, it's just like a Google Doc of guests on the show on an opt in basis with their, with a bit about their expertise and contact information. Are they open to be contacted by other guests from the show? Maybe we just start there. Yeah, because I'm already finding like guests from the show, like the, I've only published, I think a dozen episodes and.
I'm already making several introductions a week. We already just found one on this show, right? I'm sure Sarah from Melody will love to meet you. You introduced me to Sean from co work, right? And also I'm seeing how these pieces fit together, right? It's oh a couple companies, Enfi and Cowork, who are building a bunch of agents in house. And then Melody, who's who's helping companies that build agents in house to have better analytics. And and it's, I'd love to talk to some folks that are trying to figure this out in the big enterprises that you guys do business with, right?
That'd be fascinating. Or, in Sean's case, the big PMO offices, or so it's, yeah, it's like a puzzle. But to Climbing the learning curve, to your point, is a full time job. And so just how do you make that as efficient as possible? And also, how do you make those, how do you know who to go to for what, right?
And that's a big piece of it too. And I feel like whatever I'm building can help with both of those things. Like it can help make, it'll help make knowledge more widely spread, like the latest knowledge more widely spread. And it can help people better navigate who to go to for what. And if I can do both of those things at bigger and bigger scale more and more effectively, then it feels like there's plenty of business models we could layer on top to put food on the table, for me and for whatever the team looks like, families, right?
And it doesn't need to be a unicorn. It just needs to cover our nut. Anyways,
Scott Weller: think you just inspired for me, like just this concept of if your mission is around is around knowledge sharing and accelerating. Knowledge sharing and accelerating some of these qualitative qualities of communities, like how, how communities can influence each other with that, in a way that's hard to measure.
Why can't you do it with some agents? You could be sleeping to your point around. To your point around wanting to build something that is also healthy and doesn't require like. 24 by 7 attention in the age of AI agents, like you certainly should be able to set an agent, have a conversation with someone like this and say, Oh, there was something interesting in that conversation, this topic, like I can either a go obsess over that topic, or I can have my agents go off and pull together a ton of information about that topic and actually simulate back something that we could share with the community.
Or bring the community too. So there could be a way to leverage AI for you to scale,
Jason Jacobs: it sounds like it's a it's a, it sounds like to use the N5 example, I need some design partners.
Scott Weller: exactly.
Jason Jacobs: Yeah. What do you think the profile of those design partners should look like if you're me
Scott Weller: they probably look like me. They probably look like Sean and a bunch of the other people you've interviewed, I'm guessing in terms of like sharing and creating like a fluid way to see what's coming down the pike, from other people seeing trends, like seeing like I often wonder, like we had to engineer this really sophisticated way of reasoning a plan. How many other companies are trying to solve that right now? I don't know. Are we on the cutting edge or are we lagging? I actually don't know. Cause we're so in it that, we don't have time to pick our head up and say who built a model that's really great at task planning
Jason Jacobs: but here's the thing you need to keep in mind, there's tons of companies out there who started before all this madness, that have just been doing things. The way that they do them. And now they're like, and I'm not mentioning names, but some people that I know we both know, right?
That are like, What do I do? Like experienced founders. They're like, what do I do? Because I have this core business and hamster wheel and it is a treadmill that I'm on and there's not enough time in the day and we need to get a lot done with a little, we need to stay focused, but at the same time, like all this other stuff's happening over there that I'm just ignoring and I'm increasingly not okay with that.
What do I do? Yeah.
Scott Weller: and everyone. So recently it was, I was involved in a and this is less on the startup side, more on the corporate side. But this survey that went around a business group that I'm in two executives around AI and the two biggest things that came out of it was ton of FOMO, and then also a huge lack of confidence on what it is, how it works and where we should use it in our business.
So that alone, if you're able to tackle that problem alone, it just started making sure that like people in leadership roles can confidently speak to how their business could potentially use AI that alone is massively useful.
Jason Jacobs: So I think what I may do then is in this week's newsletter, I'm going to talk a bit about this discussion and about how I want to design a system, but I don't want to design it in a vacuum. I want to design it with a working group of collaborators and if anyone wants to be considered as one of these initial, I'll call them design partners get in touch.
And yeah, I. Anyways, I'll have to think about that. Is that something I put out to everyone? Is that just for show guests? Is it cause I don't want to be exclusive, but at the same time, like I really want the right people around the table and I want people who like I, as much as AI is supposed to empower people like me, and it does in a lot of ways when it comes to actually building real software that does real things and can do it in a robust, accurate, reliable way. I am not that guy, at least with the, at least with the tools in the current form. And it would be a full time job to try to become that guy.
But before I could even become that guy, there's so much underlying prep I would need to do that would derail me from all these discussions like this that are moving the needle so much for me and for what I'm building that I can't do that, right? And so it's I actually need some real technologists around the table.
And guess what? I have no business model. So then how do I pay them? How is it not just taking, right? It's what's in it for them? I guess they get to be part of the community, whatever I build, right? But it's this is the chicken and egg that I'm sorting through.
And then there's stuff like like Slow Ventures has this creator fund, right? Where it's look the future Mr. Beast of the world. And it doesn't need to be Mr. Beast. It can be a small and mighty community of hammers. That still has value, right? But it's the person, it's like the brand is around the individual.
And the individual is the Nikes of yesteryear. Are the Jasons of whatever right and like you go and build like the Jason brand or the next brand or whatever and then there's ample things you can do around it like it's not media is not interesting to me I don't want to act like maybe a little sponsorship to pay the bills but like I don't want to muck with that like the real gold is you know if you have the knowledge and the relationships and the unique perspective and the following and the trust and credibility there's a lot you can do around that right and so it's oh like you know All I want is like a hacker and a producer great, I'll just, raise a bit of money from this creator fund and then that'll get me what I need.
It's yeah, but then I need to track like a public company and then there goes all my flexibility and control, right? And so it's I don't want to sign up for that. I want to just organically let the water flow in an unbridled way and not have to force it to be anything, right? And.
Yeah, so and because I because it's not a business. I'm not about to you know I can't fund out of pocket hiring a team in good conscience because I'm already feeling financial pressure for my family Without that expense, right? So yeah, I don't know it'll fit Italy. It'll figure itself out I'm as excited as ever like I'm doing the right thing I'm gonna stick with it.
But like these chicken and eggs are gnawing at me, you know Which is it's classic just early stage founder stuff
Scott Weller: Yeah if your superpower is building a community then I'd say find that one tool you can build for your community. After talking with a lot of people and understanding like maybe this common thread, this common problem, just one tool, just build one tool
Jason Jacobs: Yeah, and it's not like the lazy just set up a slack room and see for at the time that we did it, five or six years ago, whatever, that was the right thing. But like for this world with these tools and knowledge and the rate it's changing and agents and that is not like everything in my being says that is not the right thing.
No.
Scott Weller: I'm excited to see what you do. And I'm here if you want to have another coffee talk.
Jason Jacobs: Amazing. Yeah. Yeah. And and I'm sensitive to time commitment and stuff, but obviously you'd be a fantastic design partner whatever that means. If I tried to do that, Joshua might. Come after me with a hammer because I'm doing anything to distract you from, from trying to build this big thing you committed to with all these fancy VCs on your cat table.
Scott Weller: Yeah yeah that's my number one, that's my number one goal. But if I can be helpful, let me know.
Jason Jacobs: Anyone you want to hear from, design, future design partners, any key hires you're trying to make, any, anything you want to shout
Scott Weller: I think you should I'm going to introduce you to some cTOs, CEOs that are not startups. They are in enterprise and they're trying to figure this out too, from their perspective.
Jason Jacobs: Oh, you're talking about
Scott Weller: for the show. Yeah I think definitely why I'm sorry, maybe I missed what you were talking about.
Jason Jacobs: Oh, I was just saying for you who do you want to show up in your inbox? Yeah.
Scott Weller: I want to talk to who would I want to show up my inbox? One great AI engineers that want to work in financial services on helping to, provide tools to analysts and people who manage trillions of dollars in credit. And then I think always want to meet that next.
Financial services fund, like private credit fund or fund manager or portfolio manager at a bank who's just fed up with. All of the minutiae they need to deal with in order to create a great customer experience, like the files, the data, the annual reports, the OCC regulatory filings, like all that stuff sits in front of, it's all important, but it all sits in front of creating a great borrower experience.
And if they're fed up with it, and they want to be on the other side of that, we're also hiring people. As subject matter experts within NFI to bring NFI to the institutions they work for. If you want to get on the other side, you're in this mess and you want to get on the other side of the mess, but you want to build something that solves the problem.
We're also want to talk to those people too.
Jason Jacobs: Great. And we can talk about some of those introductions for the show offline, but any, anything I didn't ask that you wish I did or any final parting words.
Scott Weller: One, do we live in a simulation? Would have been a great topic. I'm joking. What's the impact of the, impending tariffs? I know this has been really great. I think it's been really comfortable to chat. Really enjoyed it quite a bit. I felt like I talked.
About myself way too much. But I guess that was the point.
Jason Jacobs: You shared so much. No, it was awesome. But like I, you definitely don't, and again, I don't know. I don't know if you and Joshua are philosophically aligned in this, but like you, you didn't, you shared a lot of gold nuggets in there that I think some, some founders are just really squirrelly about information and learnings.
And and I thought you were just really generous with the learnings that you shared today, which helped me immensely. And I'm sure it's going to help listeners as well.
Scott Weller: one last thing to end on, which I think is a core tenant for N5 and Joshua and Michelle and myself. We're increasingly trying to do to the degree in which we're not giving away IP. I'm not going to describe to you how the, how we got to an agent system that could plan really well.
I'm not going to show you that code today. But because that's our IP
Jason Jacobs: We'll save that for the next community session for members only.
Scott Weller: But I do think you're increasingly, you have to do things in the open, like we're in a time where I think it's important to share to your community, to your buyers, to your customers why you exist in the world and what you're doing. And if you're not doing it in the open, in some capacity, you don't exist and in the olden days, maybe 5, 6 years ago.
That was like going to conferences and doing presentations. And I think a lot of those formats post COVID are broken. Sitting at a conference, listening to a panel, it's fun, but man, for every hour we spend at a conference, sitting, listening to panels. As practitioners, this goes for both the companies, the vendors, as well as the, there's definitely knowledge sharing.
It's massively inefficient knowledge sharing. Like you're in a time where AI is like taking off like a rocket ship. And if you're going to spend a day sitting, listening to panels, like you're left behind every day. You're not advancing yourself. You're losing a week in future value. And so I think there's gotta be.
That's why I'm, like, really excited about what you're trying to figure out. There's gotta be different formats, newer formats, for information exchange that empowers people and also gives you the sense That I'm not a snowflake. This is my, these are my people. This is my community, whether that, whether you're an actuary or you're dealing with air traffic control or you are in the waste management business, conferences matter because it brings people together for information sharing, but they're just massively inefficient.
And so that's why we're trying to do everything we can in the open using different platforms to say. Hey, here we are. We exist. We want your feedback. We want to hear the great things and the bad things you think about our product. And we want to have a conversation and we want to build a community around that.
And so you're going to see more from us as a team in that respect. And I'm hoping like whatever platform you create maybe is a helpful way to help companies like us get out to, just to show up in a different way.
Jason Jacobs: And like that agent coding, by the way with you guys and productizing that, right? That's another way the next X could go, is we could do this stuff for our own content and community flywheel, and then productize it for others, right? With the stuff we build for ourselves. Who knows? I'm not worried about making money.
If we can just keep the flywheel spinning faster and faster that part will Take care of itself. But anyways, Scott. Thank you. Awesome. Best of luck and ongoing dialogue for sure
Scott Weller: Talk to you soon. Take care, Jason. Thanks a lot.
Jason Jacobs: Thank you for tuning into The Next Next, if you enjoyed it, you can subscribe from your favorite podcast player. In addition to the podcast, which typically publishes weekly, there's also a weekly newsletter on sub stack at the next next dot sub stack. com. That's essentially for weekly accountability of the ground.
I'm covering areas I'm tackling next and where I could use some help as well. And it's a great area to foster discussion and dialogue around the topics that we cover on the show. Thanks for tuning in. See you next week!