Speaker 2: Well Akash. Alan. We couldn't get Jamie Dimon and Dario. But I think this is the second best thing.

Speaker 1: It'll do, it'll do.

Speaker 2: Yeah, so we heard from them all about what the big banks and Anthropic are doing with entering into AI agents. This is something that we've kind of been doing for a long time with agents and banking. I think one of the people we didn't hear about, actually quite a lot of them, is the community banks, the regional banks, the fintechs. There's a ton of room there for AI agents to help in banking, both internally and externally. We've kind of been on this journey for a while and I thought it would be worth to kind of, you know, tell the people a little bit about what we're seeing so they can have that perspective. And you've been in this deep too.

Speaker 1: That's true. I mean, it's been like, I don't know, let's say two years of you doing this and about 14 months of me doing this. And obviously the space has like continued to evolve more than I think any other space I think we've been part of. And I think, yeah, it'd be good to hear from you. Like, why are we still doing this? And like, what are we doing in the first place?

Speaker 2: Yeah, why are we doing this? I think when people hear about banking and AI, immediately I think AI people just like shutter because they're like, AI people are so used to things moving so fast. They're so used to, in engineering you can go and ship a thing and cursor can come out and it's like, ⁓ who cares if this code isn't right? Look how fast they wrote this like piece of code. Then you go into it. banking and I think one of the things most AI people, like it almost doesn't seem so compatible at sometimes because it's, you can't break stuff.

Speaker 1: You can't be wrong.

Speaker 2: Can't be wrong. And I think a big part of our thing that we started off with was how do you get agents to start moving money? And just off the bat, we've seen just how long it takes to get that bill.

Speaker 1: Yeah, it's, I guess let's go down that thought, right? It's like, we're building AI agents for community, regional, fintechs. But it's like really what we're doing is giving AI agents access to money. And like, what have we learned along the way where you think like a lot of people don't really understand why those learnings are like so important.

Speaker 2: Okay, we can start off as like, what does it mean for the agent team and get access? Like, why did we start and why did we go to the banks? It's because we realized really quick what we were doing was trying to give agents access to wallets. And by giving AI agents access to wallets, you realize that like, there's no money in these wallets, there's no transaction data, like there's nothing that actually makes these agents valuable. So we asked the question of like, Where do people have money? Where's trust? Where's this transaction data? Where's the things that make the actual agent valuable? And that all lives out of bank, right? Or fintech.

Speaker 1: Or wherever deposits are. I've been in this space for so long where it's like, you don't want to be in the world to fight for deposits because people make such a conscious decision of where they're going to move their money. And they move their money where there's going to be some level of value. And then we heard from a friend of ours, he's been banking at his regional bank or community bank for the last so many years or since he's had his account. Yeah, and he has no reason to move it because there hasn't been anything.

Speaker 2: Or where, I like five.

Speaker 1: so crazy that's going to make him want to move that money because it's so safe there, right? And so we're going now with this technology to change that piece of the puzzle. So yeah.

Speaker 2: Yeah, so that was the thing. We've been building the tech for two years to essentially figure out how do you get agents access to money safely? And we would work with the banks to figure out how do we get these agents to use their bank accounts? And it was about, what, a year ago where we realized we got to actually figure out how do we get these agents into the banks? And through that, I think a lot of what we're seeing early on with the Anthropics, the OpenAI is doing all their deployments and things like ⁓ we're gonna get into the bank, we're gonna solve these efficiencies, and it's like, man, that's like low-hanging fruit. Yeah, you can go and do that, but can you get the agent to actually handle the money to move? I think the big thing we learned is like, you gotta build these agents. You can't let, maybe the big banks can. I don't even know if the big banks can, quite honestly, but the communities, the regionals, like, they just need the agent given to them.

Speaker 1: Yeah, it's it's like Claude, I think they did a great job here where they're like launching the Claude for small businesses in the concept of the people that really need this the most are the ones who are like spreading themselves the most thin possibly, right? And those are SMBs. Like where do these people bank, right? Where are they going? They're going to fintechs or to community banks that are either providing them some level of loan or some level of deposit account that solves so many other problems for them where we think AI could like actually accelerate. so much of that experience, right? But I guess, yeah, we were talking about agents. What's been the hardest part? Like even if someone gave you 10 agents tomorrow, what would be the hardest part to even use those things and manage those things?

Speaker 2: think this is the thing that we probably didn't hear and that people won't say because maybe everyone's biased. The online banking providers want their AI agents to live within their accounts. Maybe the Clauds, the OpenAI's want the AI agents to live in their interface. But man, the hardest part is getting the, even the internal agents or the external AI agents, the consumer or employee facing, how are people gonna adopt it and use it? We're gonna say, hey, you just have a chat text box and you can ask it and. voila, people will start using it. It's like, no, that's not, it even gave, it took engineers a while to trust the AI agent to use this. But what worked really well and what I think the point I'm getting at is you gotta get the AI agents where people are working. Like, are they working in teams? Is that where their loan officers are working? Is that where their tellers are working? Get the AI agent there. Maybe there's another platform they're working in. Like some tellers are working, let's say in like a platform like Finastra. where they're interacting, they've got one screen that just shows that. Now you're telling them, okay, now you have to operate another screen while I have limited bandwidth, but you have a customer standing right in front of you. You think the AI agent's gonna be useful to switch screens and that? That's what we learned. So I think this is what any of the AI companies that have been successful, like the cursors, recognized, hey, let's build this natively into a VS code. And this is how engineers already understand work happens. We're not changing much.

Speaker 1: do think that's why so many companies are now becoming services-based companies to go and make this happen. Like, let's go figure out where we need to put this thing because we don't even know where these people are working to actually solve this problem. Something that we know because we've been doing it for so long. Do you think that's what they're trying to learn from building these consulting companies?

Speaker 2: Well, I think one, they recognize there's a lot of opportunity. And I think they probably underestimate, especially in the banking space, like I always kind of think with the anthropic engineers, I don't, anthropic engineers are like different type of breed, right? And it's like going into a bank, sitting into a workflow, it's going to be a culture shock for them. to understand how do I actually go and implement this workflow. So the services companies... ⁓ It's hard, I think today, to be a pure services company and then recognize a workflow that can be built and replicated and not turn into that software type company that just does that workflow. ⁓ I think a sneaky services type company that's a software company is something like Sierra, where Sierra is going in banking. They're going into other types of customer service roles and they're recognizing we can build workflows around this. which gets to a very important point, which is if you want to be a services company or be like one of these, think everyone is now talking about what's this agentic operating system. You saw Fiserve announce it today, you saw IFIS talk about it with ⁓ Claude, and I think it's actually where, like you've done a lot of work building our agentic operating system. Why do you think there needs to be a like this factory that needs to be built? Like, what are the considerations that go into that?

Speaker 1: Yeah, it's like, I think the thing I like saying a lot, which you've heard me say many times, but it's like, can't, it's like there's a box of ingredients that everyone can go and buy, but they might not actually end up making the cake. Right. And I think this whole idea of here are all the pieces that you need to build something, to then go and do something, I think are like too many steps in a journey of like adoption. Right. And it's like, especially as a product person, it's like, How do you get them to value faster? And especially in a world where there are so many options now of choosing where to interact with an agent, where to talk with an agent. But then building an agent has so many things. it's like, everyone can claim to be the operating system, but if there's no actual flow, it's like being a payment system without any actual volume, right? Like, yeah, like you exist, but no one's actually using you today. And I think that's why building a studio or a factory that can actually deploy usable agents that can be plugged into where you work today will be the secret sauce because you're solving the adoption problem, but then you're also solving the retention problem. And we've seen that with obviously our curve of growth with banks that we work with today. It's sure, maybe the first day it was like one, two, three intents that came in, but now our Slack has more messages in those channels than it does in our I the challenge. And what was the difference there? was like, we didn't give the bank the answer of saying, here's a platform where you can build agents where you think the solution can lie. Because you don't always have starters or individuals who know even how to use that system. And change is hard, right, in a system that has existed for so long. Especially if they don't have a metric to become more efficient, then it's fine. And so it's like, that's where I think Having a system that can punch out these agents really fast that understand banking only are built for banking and then can deploy banking and then be used within hours I think is actually like the magic Superman I mean like Well, you know with one of our institutions. It's like you're on version 42 or 51 at this point and it's like it's the it's the

Speaker 2: How hard is that? Agent 42 or 51 of an agent. Of an agent. That just is supposed to do one thing, right? Right.

Speaker 1: Just do one thing. Exactly. And it's like, I think that's where everyone sort of like the beauty of AI is that it can do so much and that it has this ability to solve so many problems. But to test and iterate and to make sure that making one little change doesn't break the other 99 changes that exist, I think is something a lot of people are missing. And it's like the difference between saying, should this be in the prom? Or should this actually be a skill? Or should this be a capability that could be turned on and off? Like all three of those have different outputs that are going to happen. And I think it's easy to say, ⁓ it's like a quick thing to build. It's an agent that has a prompt that you can just tweak and it's going to have the output. But it solves like a very binary problem. But if you're really building an agent that can do things like we are, it's like, actually, it should be solving so many problems that banks haven't even thought about that it can actually solve. And I think you see that from the intents that are happening. And it's like, Even as someone who didn't know AI that well two years ago, it's like, man, my definition of what an agent is is completely different than what the world is putting out there today. And it's incredibly difficult. How about you? Why do you think it's so hard?

Speaker 2: ⁓ I think that if I get technical, think a lot of the current LLMs are trained on a lot of engineering. And a lot of these internal workflows are different per bank. They have a different way of operating. They have different docs. I think some of the tech isn't really... strong enough to just solely be used. You can't just hook up an LLM to like some kind of document and rags gonna work magically and never be wrong. So then you have to go in as you've done a lot, which is like, all right, we got to test these intents and we think about intents a lot. Intents are just like, what are people gonna come in with and ask the AI? And does it respond the same way? Like in banking, you can't mess up. So is this agent gonna give the same flavor? Like, or is it gonna start like, maybe you're looking for ice cream and it gives you like Twix? Like is crazy. I think that has been the most surprising thing. ⁓ I think the other hard part has been seeing like how expensive some of these things can be. you can't, I mean, what we're on, like the... for any like bank entertaining this, you have to, think one great thing to ask your provider would be like, are you covering my token costs or am I going to pay my token costs? Because it's not going to an engineering budget where it's like, hey, token max, right? Like spend as much tokens as you can. Banks are operating on like very thin interest margins, especially these community banks. it's like getting in and blasting tokens is not always the answer. So how do you build a system that can go and like make this cheaper? and faster.

Speaker 1: Yeah, the last thing we want is for it to be shut down before the magic is actually created. And it's funny, it's like, you know me, I spent two months building an OpenClaw agent that I had to shut down myself because I got randomly charged like 200 bucks on Google Cloud for not monitoring that agent, right? And it's like, yeah, it was great and simple to build it, but the reason that it was 200 bucks is because just like planning a Google calendar event, it did what we call loops forever and wasted so many tokens, right, for something very simple because it couldn't find someone. And it's like, I don't know, it's like those things that, yes, implementing it is great, managing it, I think, is the hard part that, like, no one else is really thinking that deeply about.

Speaker 2: The other thing that's been like hard, which is kind of, heard one thing I really liked, I think his name was Mario at the Claude talk said like, at first I was trying to push everyone uphill and now I'm trying to hold the car from going downhill because everyone's asking for so many things. And how much did we experience that with the banks where it's like, oh, can you do, okay, we have the teller agent. Okay, well now actually this is doing really good. Can we do a loan agent? Can we now do, an AI agent that can be customer facing, right? Can we now can the agent move? It's like, can you support as naturally people recognize how valuable this is, especially in this industry, can you support that growth with the agents and ship them out? And can you do it safely? And can you show them?

Speaker 1: Yeah, exactly. It's like, and I think they're also asking that because the first agent was that powerful and that successful, right? There's so many stories, I'm sure, about implementing something and then it not moving forward. I think that's why that joke about AI kind of exists about prototype to production is there because it's failing so much when it actually does the first or 10 things. But for us, it was getting it right equaled that trust. And so like, I don't know, it's like, it's an interesting point. It's like the question moves from can it do this to what it can't do, right? And we know that that's been something that's been asked of us so many times. And so, yeah, I guess how are we, how did that shape how we built what we ended up building? Because we never want to say no.

Speaker 2: think we, and the space is so young, so it's worth knowing all the stories and how it keeps evolving because... I think a lot of, especially what we built and how we saw the industry unfolding and how it probably will unfold is the AI agent will be interacting on behalf of the user with their bank account with things. Eventually the user will just trust it to do things. We're a little ways away from that, but I don't think banks disagree. I think they're a little skeptical on when it'll happen, but it's kind of like that feature will happen. So in the meantime, I think to get that future to happen, ⁓ people need to be able to trust you that you can actually get these agents out in some fashion. I have a lot of empathy for banks because they get pitched a lot of tech that doesn't work. And it's very fair. think if you look at even our tech a year ago, it's like, we pitch, it's like, OK, we're getting there. So I think what's stopped a lot of that from happening was, think today, especially us, deployed in multiple banks, getting more banks all the time, and can actually put in a working product. What changed for us?

Speaker 1: Always.

Speaker 2: was that we had to actually build a system that could handle not just the agent that can handle the money, but the agent that can work inside of a bank. ⁓ You know this from building like one loan agent. What do you have to go do? It's like, it's not, hey, let me just spin up this prompt and put in a loan document on how to make a presentation. It's like, I gotta go and figure out how to get the APIs from Trends. Union, trans-union so that I can run credit reports, right? Because this is the process that you have to sit down and understand their process. Okay, you have to do this, then you search on the internet. Now I need to go and figure out how to integrate Brave into this. And now the agent has to be composable enough to where if I want to do this for another bank, now I have the skill and the capability. ⁓ It's also interesting because a lot of this is almost open-claw-esque and so many banks don't even know what clawed-cowork is. They can't use it. Yeah, let alone

Speaker 1: for them to do or they're not even like that's not the enterprise solution that they have today and so you know they're not going to have the opportunity to do that unless they use it in their personal lives.

Speaker 2: How much have you had to sit with bankers and just run through intents to make them comfortable?

Speaker 1: I mean, like, it's not even, it's like sitting, talk about sitting, it's like sitting in their home, right? And like actually almost going to a place that they trust to then show them that they should trust us with something that actually is going to change the trajectory of their bank, right? Like everyone's looking for that 10x bet, like, or that thing that's going to help them grow. And if you're not growing, you're just trying to cut expenses and you're trying to like build margin, right? And it's like, Wow, we've come across something that can actually do the opposite. And it's like, what is this going to take? It's like, ⁓ it's going to take me to show you the agents that exist internally at my company to that help me build the agent that your company is going to use that then your end customers are going to use. And it's, I think the beauty of it is like, there's not a list that you're proving and saying, ⁓ it can do these a hundred things. It's like, actually, what are the hundred things that your customers ask today? Or if you banked there today, what would you ask it? And the agent being able to respond and complete that intent in front of them makes them feel really comfortable putting that in front of somebody else. And so I think it's a risk, right? Everything that you build is a risk of the brand that you've built. And I think community, regional banks, fintechs, all these companies have built such a brand of like, wow, you are trusting me with probably the highest asset that exists in your life. They never want to ruin that experience and we have to treat it with that utmost thing. And it's like, that's why I'm sure you get asked this all the time, Tom, like, why are we so focused on agents moving money? Right. And it's like, ⁓ it's cause it's the hardest thing to do actually. What do think about that?

Speaker 2: It's so hard. It actually makes every other agent, I feel like we could build pretty much any other agent at this point, but ⁓ it's been like, why we even started it this way was because no one could build a good AI agent that could handle money. It's like everyone wants to say building an AI agent is easy. Yeah, I agree. It depends on what AI agent you're building. If you're building an AI agent to be trusted to move money out of bank, yeah, why isn't everyone doing it? But when you can do it, when you can go and sit with the OCC, when you can go and sit with the Fed and you're showing them things that they haven't even thought of, you're like, okay, this is nice. And when a bank's deploying it to their customers and everyone trusts it, ⁓ that's a very hard-won thing. And I think we just had, it's like,

Speaker 1: exciting.

Speaker 2: Dude, there's so much value there. And it's like, yeah, I think the future is definitely hopping on my phone and like talking to an AI agent to handle all my money. Like, how is that not the future? Why doesn't that exist? And the only reason that doesn't exist is because, like you said, everyone wants to give the ingredients to make the cake. And we were like, dude, we just got to make the cake.

Speaker 1: And the nicest cake that's out there, right? The one that has ingredients that you have to go around the world to figure out and get, right? Because that's how hard this thing is going to be. And I know like some, know, it's like, wow, like money moving. Oh, that's just like an API call or like, why don't we just like put an MCP server out there, right? Well, you've been doing this for, I don't know, the last two months building with everything else that's out there in the world. And, you know, it's helped improve our products too, because there's a great solutions out there too. Don't get me wrong, but. What have you learned the most in that experience?

Speaker 2: ⁓ it doesn't work. think that's, banking data is messy. There's a ton of banking data. How much do we have to clean up the banking data that comes in just so that the AI can read it well? like, MCPs are just, here's all the AI, here's how you can access the data. Yes. And go get it. And I mean, I was shocked. There's like big companies in banking that release MCPs that just don't. really work well and I think one of the reasons why a lot of people are doing that is because they're like, well, we don't want the liability of an AI agent to go and do this, so we're just gonna ship the product over there and let the customer do it and have a bad experience. Let them have a bad experience in Cloud and not in our app. that goes against every product instinct in my bone.

Speaker 1: And there's so many barriers to that, right? Which is like, to add an MCP server or a connector, you actually have to have a specific plan that's like above the regular plan. So when you think about like 900,000 daily active users or monthly active users, that number becomes so small. And then you're expecting that subset to then also do something that's so hard. That's not the hardest problem that they're trying to solve in their lives, right? And we're saying, Let's just go to the place where they already solve one of the hardest problems in our lives, which is their finances. Yeah.

Speaker 2: I think that's like the... like hitting on both sides, it's like there's the external use case, which is like consumer facing, which I think we, know, being the part two of the Dario and Jamie conversation is like, you know, this is the real world of what like agents are trying to, it's like, what are the consumer agents? What are the internal AI agents? Why is the MCP like not a very consumer friendly thing? ⁓ It's everything we've talked about. And I can tell you probably if there's a banker listening to this right now, they are probably Googling what is an MCP. You know, it's like this industry also doesn't know what an MCP is. It's like, think about the tech we're dealing with at some of these banks. That is like soap APIs.

Speaker 1: Yeah. Or don't even have APS.

Speaker 2: And COBOL is like, they can't even get cloud cowork in the bank, right? It's like not used out. It's like the bank has a separate thing that only approved apps can go and be used. then.

Speaker 1: Which is crazy that we are now an approved vendor at a financial institution that has to live up to the highest regard of security today. And that's how high of a barrier it was for us to break into this thing. Not at one, not at two, not at three, we can do the whole LeBron thing. ⁓ it's, I think setting the bar so high allowed us to achieve what we're doing now today. And you're right, people don't need to know what an MCP server is in the first place, right? Like we're both product guys. No one thinks about what payment rail.

Speaker 2: Yeah, that's awesome.

Speaker 1: they're doing when they make a payment. No one's talking about ACH or wire. It's like it just needs to be an experience.

Speaker 2: think that's a good point. We talk about internal AI agents, we talk about AI agents handling money like... Is the AI making the payment? what do you think when people hear that, what should they take away? Like an AI agent making payment, how does that happen? This might be the first time anyone's even hearing about that. What actually happens?

Speaker 1: That's really funny. Yeah, let's demystify that a little bit. So it's like, one, let's start off with the payment rails are still the existing payment rails that have been built today. ⁓ And the future payment rails that will also be agentic, if that's what we want to call it, will still be the payment rails that would exist for humans and fiat money to move or crypto money to move if they needed to be. So what is the AI agent actually doing? It is the orchestration layer. It is the understanding the intent, taking how we naturally speak. and how we naturally want to do finances, and then determining what actually to do on the underlying hood. Right? And so it's like, the agent is the same way a human is deciding that they want to do certain actions. But now the agent just makes that experience way more powerful and way more possible because it's not confined by what exists within an application or what some bank offers today. And so, yeah, it's like, understanding that, I think, makes the load of taking this on a lot easier, right? And so, yeah, mean, like, what would you say the AI is doing before it's moving money?

Speaker 2: think you brought up a good... It's like when you're telling a person that's a teller at a bank what you want done and they're just like listening to you and clicking the buttons. Yes. And today when you expose an AI that can handle money, it's listening to you because you're talking to it and it's figuring out which APIs to call, what information to collect. Right. The very important thing to make this work is what information to collect based on what the bank's looking for of an AI under their control. Right. We'll take Middlesex as an example, one of our banks, like they have an AI that works exactly how they want it to collect money. And it replicates maybe even more than a human has to go collect. it doesn't change where that money's sent. Like send a wire to Akash of 50 bucks. Maybe the agent knows, okay, let me check my policies. Can I even send 50 bucks? Okay, I can, but I need to get approval. Well, let me check if... ⁓ Well, the bank also has this policy where if it's like over a certain limit in the checking account, maybe they'll like if it's multiple transactions, they'll charge a fee. Let me let the user know. Hey, are you good with this fee? All of that information gets collected. Hey bank, here's how the AI payment shows you this is how the AI made this decision. Right. If anyone wants to ever dispute this with the AI, you have all your information because this is all the information that the AI agent collected and The API just got clicked as they did. There's another person that manages the wires, does those things like...

Speaker 1: That's so, I like what you said there. So we think about the future world of how this is going to go down. Like having a trace, which is like what happened in the system of understanding what the agent did and having an agent that actually understands how the money moves in the United States or globally one day, hopefully actually makes the future problem that much easier too. Like for the payment nerds that are listening to this, like ACH has rules. Our agents know all the rules. So our agents would know exactly how to handle if those rules were ever broken or disputed. And it's like, that's an automation that's naturally going to happen through a payment AI agent that exists today without ever having to implement another AI agent. like, and we go back to the financial agents that are being put out there in the world, the solutions that people are trying to solve. Like one of my mentors used to always say this, like, if you cluster all those problems together, like you have to just change the actual product. today we're just putting a bunch of band-aids. on processes that shouldn't exist possibly in the first place. And we've had to put people to solve those processes when they could be solving other problems, right? Building things like we are for that same institution. So, I don't that excites me like a ton of what could be there.

Speaker 2: You've been a payments person your whole career. Yeah. Right? Like, do you get scared of an AI handling your money? I think you probably send more money with AI than anyone on the planet.

Speaker 1: It's because they're never, they're not lazy, right? They only go down if the systems are down, which hopefully doesn't happen. And good thing we have a bunch of fallbacks. But it's like, if you wanted to do 50 things before it does the actual action, it will do that every single time without skipping a step. How am I supposed to, like when I go to a bank to wire money for real estate, I'm like writing on a form. I'm maybe putting in a code.

Speaker 2: How many mistakes can you make?

Speaker 1: Yeah, and then you're like so anxious the whole time when you leave the bank because like a five figure or six figure amount of money is supposed to leave at some point and go somewhere where it's irreversible. And you're like, I don't even know if that person did everything correctly. At least now I am put the onus on me to do everything, right? It's like we want to do everything and then have the system check everything before it actually goes out to.

Speaker 2: Or have the system help you do it. You're telling me the greatest technology ever created isn't going to be helpful and be safer as a co-pilot, as an assistant with you at your bank. It's like, yeah, I wish I could just take a... I just had to do a wire transfer for a real estate transaction. Man, I'm checking that number 20 times. And I would have felt a whole lot safer if I had our AI agent just working with me to say, hey, this all looks right, double check, I've got the same thing, here's the approval. ⁓ Thank you. Yeah. it's still all on the same rails. mean, goodness, just the side tangent. It's like AI agents handling money seems a lot safer than like writing a check.

Speaker 1: Yeah, ⁓ for sure. Absolutely. Like that's, man it's, it's one of those things where it's like once you do it, it's like when people say when you travel first class, you never go, you never want to go back. It's like, I don't think I want to interact with anything else that can handle my money at this

Speaker 2: Yeah. think a lot of the bankers don't want to either, but I think you've had this thought that you've actually been working on for like the past couple of days is like, how do banks even get there? Yeah. You know, like here we are, you know, talking to them about AI agents handling money and like we've seen this and like some banks might just be like, dude, I just want this thing to just answer questions for employees. Like, what's the evolution? How do people get there?

Speaker 1: Yeah, so in my eyes, ⁓ after I was talking to so many institutions at this point,

Speaker 2: And can we support all that? Like, yeah.

Speaker 1: I want to start with yes. Because once again, we started with the hardest way of, the hardest thing to do with an agent that everything else in between can actually be accomplished with a platform, right? When you're building things that need to be scalable, you want to be able to guide people through the journey. And AI is a journey, right? It's like, think about how hooked we are. We can't even imagine a world without this technology now. But the first time we started, we were probably a little skeptical about how it works, right? And so it's like, if an institution is skeptical, about putting this in front of, let's say, the future future of an agent autonomously doing things for their customers. It's like, okay, who's a great representation of your customer today? ⁓ like, let's say it's an internal co-pilot, right? And there's a lot of players that are trying to help with that piece, but that's okay. It gets you very comfortable with, wow, this agentic experience exists or an agent exists, an LLM exists. It gets you very comfortable with those terms and what it can actually do. And then you turn that a little bit into like, okay, where can it be in an operational aspect that's actually helping customers today. ⁓ Whether that's, I don't know, running your daily recon to make sure that all of your accounts balance, to make sure all of your accounts in the GLs have a zero balance so that someone knows what to focus on. And then the third thing is like, ⁓ wait, why isn't this just facing the customer? And that could just start with read functions only, right? Like you go back to what's my balance? That's one of the... highest questions that's actually asked for community regional fintech banks today because that balance might be wrong or that balance might be something they're not willing to adopt and check then quickly moves to like wait how did I do this month compared to last and yeah sure we're in a PFM world in that scenario but then you're like the moment it becomes in that context of like hey I think you should actually move some money here I think you should pull in money here because you're gonna be low on payroll etc the next question they're gonna ask is can I do it And that's where a lot of companies stop. And the customer is like, well, that makes no sense. You got me all the way here and now you're not going to help me finish the job. ⁓ So then it becomes a customer facing agent with right functions. And then I don't want to the thunder on the last one, but like, what do you think is after that? Right? Where do we think is the final scale of this sort of adoption curve?

Speaker 2: I mean, spell that out awesome. And I hope like every bank also realizes like, you know, it's not just the big banks. It's like, community, regional, I'd even say like the fintechs that some of, a lot of community banks and regional banks support fintechs. This is hard tech that everyone can have that the, what, the big banks are spending billions of dollars to build in-house. It's like, yeah, they spend that because it's very, complex AI talent sparse. And it's no, it's possible for you too. And in fact, we love working with the community. It's like our kind of people too. Where it's going is, well, obviously once like the next question becomes, ⁓ well, you can send my money and like, you can do that. That's so cool. I have this great experience. Well, hey buddy, what if the AI agent just actually starts looking at your account while you're asleep and helping you? Right? Hey, this transaction came in and it's weird. Let me ping this Thailand and be like, did you spend this or like even something like fun, like, Hey dude, you got a big paycheck. You know, you close a big bank. Like, Hey, congrats. It's like, that's the buddy. You know, you, it's my daughter's gonna and my son are gonna grow up in a world where the AIs are gonna be doing that. We're gonna be building them to do that. And then I think the like key thing, which it's like, you don't hear in any other. anyone else talking about it because everyone's like, we're gonna build the agentic banking operating system. We're gonna get all these internal workflows. We're gonna make you so much more efficient. It's like, ⁓ okay, cool. But it's like, dude, yeah, because everyone sees opportunity there and things, but it's like, nah, let me actually guide you and tell you what's gonna happen when you actually do this, which the end state is internal to the external, to building the AI agent that can do all of these things, to essentially saying, Hey, now you have a representation of your bank online. You have a digital robot that knows how to operate everything within your banks. Tomorrow, when let's say Claude's robot or OpenAI's robot wants to come and basically speak to your bank about anything to move money on behalf of a customer. Hey, I'm the bank's AI agent. Can you show me this is actually the customer that you're acting on their behalf? Then you get exposure to that. And that's what we can enable there through through our long-term vision of a network where AI agents can talk to each other on Payman.

Speaker 1: is so like that that excites me that excites me that's why I obviously came here right I think it changes the definition of what an agent economy actually is right we hear the word agent commerce all the time like you I where do you think that fits into what you just described

Speaker 2: The banks have always given the blessing to tech companies to do any kind of commerce. It doesn't matter if you're Visa, doesn't matter if you're Stripe, where do you have to go? You gotta go make sure that the payments work under the governance and the authority of the banks. Well, in an AI world, you gotta make sure that the agents work under the governance and the authority of the banks, and has the bank given you the blessing to do that? And I think, you know, very easy to see a future where If all the money exists with the bank and the customers are interacting with that agent at the bank, you just build the agent with the capability to shop there. And the agent can figure out how to complete a transaction. If let's say Claude becomes really good at finding products and it can go and complete a purchase. Well, if that money and audit trail is eventually going to be through an AI agent and that's all intent driven, then the engine that processes that payment is actually processing a natural language intent and what engine can even process that today? It has to be an AI agent. It has to probably be a banking AI agent or under the purview of the bank. So that's where it's all going. How many times have we had banks talk to us about agents of commerce and like they feel left out?

Speaker 1: So much. And it's like...

Speaker 2: Who's going to define that for them?

Speaker 1: Right, like we, they cannot miss this opportunity that is like, you know, what the world is claiming is going to be trillions of dollars that's moving through agentic, you know, commerce. And it's like, they don't want to miss out. And I think, you know, someone who's been in it for quite some time now, it's like, man, when you look at these like amazing fintechs like Plaid and others who have... like basically said, let's go and re-transform all this data that you hold and be allow companies to exist on top of this information was an opportunity that banks are now like, wow, like we unlock this world, which like hopefully got us deposits, but we lost a lot of the control of the data that exists out there in the world about our own customer that we know where are the ones taking on the operational cost of serving. It's not like, okay, wait in an agentic world, I don't want to lose out on that, right? How do I unlock what I have today to enable that thing, but also where the value comes back to me? And it's like, the value is in that agent, right? Otherwise you're fighting for, hopefully the agent creates a virtual card off of your card system, which not all banks have. Hopefully that the bank account that the agent somehow ACH debits is my bank account, which you'll have no idea unless you see a debit that comes from the merchant of

Speaker 2: In fact, everyone wants the deposits to move from your bank to some agent wallet. Yeah. They don't want the deposits there. They want to attract it to another. It's like, as a bank, you should recognize, like, and they are. Yeah. Which is like, agents getting wallets is going to be at your bank account, is going to be at someone else's bank account, is going to be there. like, we know how hard it is to move liquidity from one place to another. ⁓

Speaker 1: And we're talking about like actual transactions, right? I think there's the world, you hear this all the time, like micro transactions, stable coins, and like instant money movement like that. Sure, there's a world for that, what are your thoughts there?

Speaker 2: I think that that's a ways away. ⁓ I think it's a different business for a lot of different ⁓ things that if the banks, if it starts taking off, I think that there's a world where stable coin providers should just work with banks to allow customers to open up bank accounts that they could give to their AI agents that they can transfer their money into stable coins. Great idea, someone should do that.

Speaker 1: Exactly. think that's where community banks and regional banks, fintechs, others, like a solution that can transform to interact with that world as well without ever feeling sort of left out.

Speaker 2: I think a key thing we didn't even hit on is like we're also giving them intent data. Like, I mean, why is your customer even trying to move money in the first place? The bank never knows that. today you just walk in and maybe if you walk into a branch, you'll know. Right. But exactly. Like it's definitely getting less and less and more people are going online to do stuff, which is great.

Speaker 1: ⁓ right. What does that even mean? But who does that? But yeah. Yeah.

Speaker 2: but then you lose the relationship because you don't know why they're doing it. They're just moving it. Maybe they're moving money off your account because they want a better CD rate somewhere else. Or maybe they're moving money to your account because they just have a better relationship with someone else. Well, then the bank should know that. You should know why your customers are doing what they're doing. There's never been a better tech than AI to do that, but there's also...

Speaker 1: thousand percent.

Speaker 2: This is my problem with MCPs just being like the solution. like, dude, you expose your MCP to Claude, you expose it to OpenAI. One, you're not capturing any of that intent data. Two, your customer's having conversation. You're losing your relationship. That is like the whole thing of any community regional or fintech. Like fintechs built for a specific purpose to a specific group of people because they can serve them better.

Speaker 1: And they're all guessing constantly, how do I get this to become the primary account of this holder? And so it's like, and everyone's spending boatloads money on customer 360s, understanding a customer scorecard. It's like, man, Claude knows me better than anything out there in the world right now. And I've only known Claude for six months. And it's like, wow, now you're telling, like if... If someone can port what Claude has to my bank, I think they'd know me ten times better. I think that's a gap that we're obviously bringing together.

Speaker 2: to how much stronger your relationship with your bank would be. How hard is it to leave Claude to somewhere else, even though Claude gives you the ability to port memory over.

Speaker 1: I haven't even thought about it once. No.

Speaker 2: ⁓ Where do you think the model companies fit in all this? They're all coming in here, they're all trying to figure it out. What do you think is going to happen?

Speaker 1: Yeah, I like would hope that their goal is that they make the ability and access, like they should be solving access, I think, to the best ability that's possible today, right? There's obviously a lot of money to be made by LLMs and competing for use cases, but I think it's like building very strong models that are cheap enough for everyone to be able to use. so that the application of what we're doing is actually really powerful and cost isn't the reason that people don't implement this thing, right? And I think what they should be doing is continuing to find more resources of abilities and information to extract to then train models that are more comprehensive in many different fields, right? That are very applicable in this scenario. ⁓

Speaker 2: Why do you think they went after finance? Obviously, it's like, yeah, there's a lot of opportunity, it's like, how does that translate from, know, engineering was obviously the first thing, and then I've seen these curves where it's like, okay, finance is like taking the same thing. You think it's just because opportunity's there, or is it like ready?

Speaker 1: Yeah, it's like this is like very high cost centers for institutions that are running on margin, right? And so it's like how do I get more value out of a large expense that I have today? Right, you think about a big bank and then an investment banker. It's like if I can get Thailand to work on 10 deals versus two deals in the span of a week, I have 5x the chance of winning yet another possible bid that makes me many multiple millions of revenue from doing an ⁓ &A. it's yeah, it's like getting people back to what they were hired to do versus the operational mundane tasks that are there. And so it's like the cost exists, right? And so it's like so much cost exists here and it's operational that you can do.

Speaker 2: think what we're going to see is the same thing we're seeing with ⁓ engineers and AI is like, dude, how valuable is an engineer that knows how to use AI very well?

Speaker 1: Why are we still 12 people or 11 people at Payment? a scale like us. You'd have multiple bulls.

Speaker 2: You'd have to have like 70 person engineering team, but it's so valuable because our engineers know and you are looking for those kind of engineers. What's going to be so cool is seeing banking professionals that know how to use AI so well that it's like, know, entry roles in banking are not that high for some positions like teller or something, but a teller that knows how to use AI so well. think banks are about to like experience the same thing of like, holy cow. can I upscale my team or can like university should start thinking like, can I actually train the best people to know this? What would be cool is if like there were like, you know, there are these AI engineering coding boot camps. Well, if so much is going into financial services now, are there AI banking boot camps that can, I bet you people can make a ton of money just training people how to do that, which probably why consulting exists.

Speaker 1: Yeah, it's a teller at a community bank or just a banker in general at a community bank is being asked to do account management, relationship management, corporate management, everything at the same time, selling, growing the book of business. Like how do you expect this one person to focus on the highest, most important thing of the day when they have these 20 other things that are there?

Speaker 2: And they have life outside. it's like, they're probably just like had a kid and there might be a first time job for them too. And they've got a hundred other things to do. And it's like, I think that's a good actual reason of like, and now here we come and we want to implement AI. You know how scary that is for a lot of people? Where's their incentive to basically be like, ⁓ yeah, I'm going to do this. I trust these guys. Here comes this company from SF that's trying to like do this. Big reason why I moved back to Durango, Colorado, right? It's like, You're with folks in a small town where a lot of the community, the regionals are. It's like you gotta understand these people, their lives. How much do we just sit with them and be like, you build the personal relationships. It's not all like, hey, let this AI eat all your food. It's like, hey, it can help you. It really can help you and this is how it can help you. And if you adopt it, it's gonna make your life easier. Favorite thing I tell them is like, dude, you'll have more time to play golf.

Speaker 1: It's an enabler. Yeah for the people that want to be enabled. Yeah I think that's what why you hire people in the first place right? It's like they all want to unlock more value for themselves and I think the beauty of AI is especially at a bank is There's probably so many ideas and thoughts that you have but you just don't have the time to go through that process and think through it with a copilot and actually implement it by yourself because it requires

Speaker 2: I

Speaker 1: so many layers to actually make happen.

Speaker 2: But how much of a gem are those people that are, there are people all over in banking doing it right now. Like they see this, they see the opportunity and it's like, can't, how nice is it when we see like one of our bankers using the thing a lot and just being like, hey, here's this change request, add this document, here's this change request, can we update the agent's voice to be like this? Hey, I love that, you know, Davis made like a very funny joke to me today.

Speaker 1: The amount of screenshots we get of the responses, because it's, and that's a good topic too, it's like the probabilistic, you know, we're implementing probability in a world of determinism. ⁓ But yeah, it's like, it's those little things that I think we'll go unnoticed over time. Right now it's still such a magical feeling because everyone's experiencing the magic for the first time, but it will definitely dilute at some point. But I think because it's happening so much, it's going to, you know, it's something that people are just going to enjoy for so much time. ⁓ And you're right, it's like the questions that people ask. I the funniest one that we've ever seen is like, am I going to be broke in 2026? you know, ⁓ it's good. It's a good question to know. And I would love to go and interview that person and say, where would you have asked that question if this didn't exist? Like, where would you have sought that answer? Even if it was a funny question, the conversation afterwards was such an intellectual one that they were having. And you're like,

Speaker 2: You gotta know that

Speaker 1: Man, would that conversation even have happened if this thing didn't?

Speaker 2: What a relationship building. Like obviously someone's a little fearful of money. It's like, dude, yeah, as the person that's walking in, can I see the notes that it's like, ⁓ this is like the sentiment of the individual. Like maybe I shouldn't sell him a car loan. Yes. You know? Like, nah, let's actually help you save more. That's how you build a relationship with these things. And it's like, takes, I mean, doesn't everyone, I hope a lot of people want to, at the end we're people and we like to help people. Yes. I still believe that about society. And it's like, I think this is the other thing that goes why we're talking about the community banks right now and why we did this whole thing is because community banks don't want to lose people, not even their own internal employees. It's about like help these internal employees come from their community. They might have known them when they were kids and now they work at the bank and the bank's reinvesting back into like a hotel in the community. It's like, man, we're not here to make your life harder. We're here to make it easier. The AI tools that just, I think it's gonna be tough for any AI tools or things that go into a bank and say, we'll be able to 50 % of your employees.

Speaker 1: When have we ever used that tagline? We have never gone into a really

Speaker 2: I even like talking about cost. I think we only talk about how we can help you grow. Grow? Yeah. That's it? You should help them grow. Yeah. It's about going out there, here's tools to help you compete with the other guys.

Speaker 1: And that's how we're measuring ourselves. It's like, if anyone wants to measure how effective this tool is for X, Y, and Z metric that they're trying to move at the institution, it's like, that's fine. But the metric we're going to hold ourselves accountable and why this solution with us is worth it is because we're holding it where that never has to happen, We know at the end of this experience, we've actually recreated sort of AI and human economy that grew together. And I think that's what kind of gives us purpose to keep going. And I think that's important.

Speaker 2: how many years is this going to take?

Speaker 1: man. I don't know. mean, by the time your kids are talking to our agents as the primary usage, I don't think it's decades away, man. I think it's like within the next decade, I truly believe there will be as many AI agents as there are institutions representing them that are interacting and actually creating a lot more economic value than...

Speaker 2: How old will we be?

Speaker 1: we couldn't even imagine. But I would say hopefully within the next 10 years.

Speaker 2: We know a lot of them can be deployed today, right? hours. Within hours. How many times have people heard that with tech? Right. It's like, nah, it can happen pretty, how many times? I think the thing people are shocked is if I'm gonna give ourselves a little shout out is how fast we can move. Yeah. I think it's also a benefit of being a post AI agent economy or a company versus a pre AI company, because you just build differently.

Speaker 1: Yes.

Speaker 2: Yeah, it'll take a while, like you said, within the decade, maybe for AI agents to be shopping autonomously and doing all the things, but AI agents helping your employees and things can happen today. AI agents that are helping your customers do more and more can happen today. Pretty cool we have some agents doing that already. First ones in America that are actually moving money.

Speaker 1: It's funny, I think it's when you look at the number of dollars moved, it's really rewarding to know that somebody trusted tens of thousands of dollars to move through a system. Yeah, and that's one person, That they were willing to, and that's how they bank today. And you have hundreds of other people that are doing that. And it's the same scale, right? It's like,

Speaker 2: And it keeps going up.

Speaker 1: I don't know, you pay someone random for the first time, you send them a dollar first, right, as a Venmo, just to like make sure that that person actually got the money before you send the rest. And I think that's where we are right now. It's like everyone should come and grab their agent that can do something for them, feel the magic, and then get on this growth curve that's going to happen with or without them. But we want them to be part of it, right? And I think that's important. All right, man, let's do this with like what excites you the most, right? You've been at this for two years.

Speaker 2: I don't think it's ever clicked more. this is the time that it's like, being in this for two years, when I started thinking about AI agents handling money two years ago, it was like, boom, it's gonna happen right there. And it's like, dude, you're a little early. know, like agents aren't going out here paying for things. But this is the time. That's what's exciting. I think when I see how happy people are to use it, it's like, you know, it's clicked. And like, I am just excited for more people to use it. I think that this is something that I've experienced a few times in my career. you know, a lot of it was like at Metta where you see how quickly people adopt something and then it just takes off. And in a lot of senses, this is a very consumer product built for enterprises, either internally or externally for their consumers. And to... to see people keep coming back to it and using it, it's like, know this future is going to exist. And I know we're going to make it happen. And it's just a matter of going out there and doing something you enjoy every day. That's what excites me. mean, I've always wanted to build a thing that my daughter would be able to use in her lifetime when she's 15 years old. And I can say I think she's going to use like dad's AI agents that are moving money for her. ⁓ And there's no better feeling than being a part of changing the world in that way. Like someone's going to do it might as well be us. What you? What's been the most exciting thing or what are you looking forward to?

Speaker 1: I it's the first time I've worked on a product that I've heard like multiple people now say, my dad, my grandma would use this. Like I...

Speaker 2: And some grandmas and dads are using it quite often.

Speaker 1: Right. it's like, we just needed way more. We need way more banks to have this solution out there because then more people can touch it, right? Going back to what we started with is like, no one's going to move their bank account tomorrow. And maybe they will, right? Depending on how much more it can do when you unlock a lot. But in the very beginning, like no one's going to move their bank account just for a said feature or said functionality. But an AI agent, if more and more people start talking about it, this is how they interact with things. This is how they do banking. It's like, maybe they will, right? Like at some point, if some product was offering Claude and I'm so used to using Claude in whatever design fashion that is, I might actually move there now because I have such an affiliation with this thing. And it's like, damn, like that level of adoption across the spectrum of all users, right? Banking is a spectrum of cross, the moment you are born to the moment you die. it's like, if people are saying everyone will use this thing, And then you look at the curves and you're like, wow, the retention curve or the growth curve is actually going that way. And you're like, you're onto something pretty special, right? And then you're like, this is just the beginning. I think that's the other feeling that's so exciting is like, this is just the beginning. And honestly, all we're doing right now is like a win is if somebody trusts us to deploy this at their institution. And we have given them every reason to do that. And the

Speaker 2: Their banks have to.

Speaker 1: I think when like you can walk in to a place or talk to a banker or talk to the CEO of a community bank or regional bank or a fintech and say, we're actually ready now. Like there's no more to do. This is not like, this is ready to go the moment you say yes. Like what do we need to do till you say yes? And any answer you don't have, we've already probably figured that out, right? Talking to the OCC, building things because we're like, would we trust it to touch our money? I think that's like the most exciting piece of the puzzle right now.

Speaker 2: Dude, I think that's like, to add onto that, it was like, you know, the CEO of one of the banks we work with, he's like, you know, the reason why I like you guys is because it was like, one of, you know, if it was all 20 year olds and 25 year olds, you know, I would expect it. it's like, but I got 70 year olds out here loving the product. ⁓ That's what got me. It's like, this is insane. I've never seen this with a tech product. Yeah.

Speaker 1: Or when you hear one of the leaders of ICBA who is a representation of 4,000 plus institutions saying like this is the best individual product he's seen over the last decade of running the program. You're like, whoa, like it actually must be special, right? Because you talk to these individuals every single day. And when you have bankers emailing you saying, Man, I just showed this like four other friends and they're like so excited. You're like, okay, we're actually giving something out there in the world that may take some time to adopt even though we're ready. It's actually like pretty exciting, like super magical that this is going to actually happen, right? And I don't know, man, AI agents moving money.

Speaker 2: Hey agents, banks, everything. Yeah.

Speaker 1: doing everything we need them to do. And as someone who loves banking, it's like, I don't even think people have actually imagined what banking will look like after this. I know I tweet about it and I post on LinkedIn about it. It's like, just ask it something that you've wanted to ask it forever and you've wanted it to do for your life and see how much it can do and how much we're going to enable it to do. So, yeah, dude. That's beautiful.

Speaker 2: I think we gave the people some inside baseball. Probably that was what we came out here to do, which was like, y'all heard what might happen and going to happen over there. Well, here's what's happening. And we'll keep doing that. I think this was awesome. We got to do more of these and we will keep doing more of these. So no, awesome. And I love working on this together and we'll keep doing this and we'll keep. making more robots that keep handling money, that keep doing things for banks and stay tuned.

Speaker 1: Stay tuned.