DualEntry's Co-founder Santiago Nestares shares why Mid-Market CFO are ditching Legacy ERPs

Welcome to the Future Finance show where we talk about treasury management on bars.

We know that all these legacy ERP systems take so long to create reports. We had five or six people that all they were doing was repetitive tasks every single month. And then at one point I saw someone taking a journal entry from the past, dropping it chatgpt early versions and then saying this is the new allocation schedule, the new accrual and. And I was like. Meet Santiago Nestares, the co founder of Dual Entry, a full replacement for all legacy ERP. It automates 90% of manual finance tasks with AI and handles multiple entities, multiple currencies and real time reporting, making the one person finance team a reality. He previously co founded Benitago and has also been featured in Forbes and more. The human also has to approve the AI transaction, so there's always a traceability, but you don't have to do all the work of creating this journal from scratch. Anyone who's living in a legacy ERP knows the kind of BS around those AI features that are getting sprinkled on top.

Where do you see the first real ROI showing up for finance teams going with these systems in the next 12 months or whatever?

The most important thing when we go about.

Future Finance is brought to you by qflow AI, the strategic finance platform solving the toughest part of Planning and analysis. B2B revenue align cells. Marketing and finance seamlessly speed up decision making and lock in accountability with qflow AI.

Welcome to Future Finance. I'm Glenn Hopper, along as always with my co host Paul Barnhurst. Today we're joined by Santiago Nestares, co founder and CEO of Dual Entry, an AI native ERP platform that's reimagining how finance teams manage accounting, reconciliation and reporting. Before launching Dual Entry, Santiago co founded Benetago, an Amazon brand aggregator that scaled to more than $100 million in revenue. That experience running a global multi entity operation exposed the bottlenecks and inefficiencies in traditional finance systems and ultimately inspired the creation of Dual Entry. The company recently raised $90 million in Series A funding from Lightspeed Venture Partners, Khosla Ventures and GV to build what it calls the first ERP designed for the AI era. Santiago, welcome to Future Finance.

Thank you for having me.

So you guys have been in the news lately. That's a big funding round. And before we dive into that though, you built and scaled Benatago to over 100 million in revenue before this. So I'm wondering, building that, were there decisions from that journey that kind of shaped how you think about Building this AI native erp. And what did you learn about finance ops at scale that you're now productizing?

Yeah, I mean, thank you. And yes, we did finally come out of stealth and very happy to tell the world what we built. But a lot of what we built comes from those days. You're not born into the world and get to experience what a mid market ERP is. That is something that only those fortunate enough to be in a medium or large company get to interact with. But almost every one of those interactions, if you ask around, are quite negative. And our last company was our first window into that, into, you know, we thought, or at least we went into it thinking like as the company gets larger and we have economies of scale, we'll be able to afford better software. And these software are going to be more tailored, you know, smoother. It's like, you know, it's like if you pay for a better car, in theory the car should be better. But it was to our surprise that it's actually quite the opposite. You know, consumer grade software feels great, but then you get to these ERPs in the mid market and they feel like they were built 25 years ago and nobody's done anything to them. The tailoring or that the fact that they're custom made is actually more like a euphemism for you need consultants to do everything and you are now dependent on them. Almost like tech debt, but with consultants. And that experience was so horrible that I said there's no reason why, why this has to be like this nowadays. You know, like where you mid market ERPs are uniquely in the position where almost every Customer comes from two places. They're either coming from QuickBooks or Xero. Why is that transition not super smooth? Even today if you buy a Mac or if you buy Windows, there's a migration tool that helps you go from one to the other. Why does this not exist for erp? And so we said we're going to, we're not only going to build something in the, in the 21st century that is AI first. That takes into consideration how the world has changed and feels like a consumer grade software. But are we also going to make that transition seamless, which is where half the pain is today?

Yeah, I've actually been involved with a QuickBooks to NetSuite migration and the data, it's amazing. And there are some consultants out there who specialize in that and have built their whole careers on it. And it's crazy to think that something that specific data that is structured and in one format that, that it's not already just instantly mapped over. I was really kind of shocked at the amount of work that went into moving simple GL records over from one system to another.

And the interesting thing is, like, you're going from a simpler system, so it's not like you have to really think about every edge case. You just have to think about the edge cases from a simpler system to now a more capable system. But, you know, it doesn't take a lot of research and this. We figured out over time that it all comes from an incorrect design. From the very beginning, back in the day, the way you bought software was not by going on the App Store. You know, even the way you bought hardware was through this IT consultant. People, they would come in, they would set it up in your network rack. So naturally, the way that these legacy ERP softwares came about, selling their software and distributing was through these, like, partners. It was, you know, this distributor model, and you would buy the CDs from them, they would come and install it for you and configure it and manage the upgrade. So, of course, you want to create an ecosystem in which the partner also wins big by selling your software. The thing is, now that's taken us steered off the market in such a, such a bad direction where these big erps, like, even if they woke up one day and they had all these modern engineers and they said, we're really going to now revamp it and change the way we do things, is, well, they're going to kill half their distribution. We even spoke one of the CEOs of one of these legacy ERPs before starting dual Entry, and he bragged to us about how one of the years they grew 20% year over year by making the software harder to implement, because then that would incentivize the partners to pitch their software over their competitor, because then their partners will make more billable hours. So it was almost a competition of who can make the crappiest software. And I was like, they can only take you for so long. Eventually, like, the customers will realize that they're being played. So the pain is there. It's not clear in hindsight, but nobody's there to say we're going to build something that's easy to implement. We're not going to encourage partners to keep you in this billing cycle where it's almost like ransomware. And, and we're going to get you up live in dual entry. In instances, even in 24 hours, you know, live, the data is, is the, the data is live. We still, we still work with you for, for a few weeks to make sure that you're, you know, all their workflows are comfortable in the training, but at least the data, which is where the, the most amount of risk lies, can happen in 24 hours. But by the way, doing the data migration manually is one of the riskiest processes because you have to freeze it first, download the information. But the company still has to keep running. It's not like the company doesn't do the books. So while you're not uploading this and you have to upload it in the right sequence. Right. You can't upload payments before the invoices. And so you have to do all these uploads and sequences. The company still runs. Somebody has to close the books. Well, now then you have to go back and bring over this stub of data that was left over from the moment that you froze that initial data. And then maybe you don't have all the vendors in the new system and vice versa. So all these, all these issues happen because this is a manual process that shouldn't even be manual in the first place anyway.

Yeah, and not just NetSuite implementations, but I've been involved with several implementations and it's always maddening. And the, you want to make employees really mad. You tell them they have to do all of their entries in two systems as you're going through that stub data period. And it's, you know, it's sort of a, it's a testing period and all that, but it just kind of become part of the process and everybody just understands. Yeah, we're going to be logging everything in two different systems during this transition period or whatever. And it just adds complexity and more opportunity to mess up and have, and have records that don't match across the systems. And yeah, it's a, it's, it can be a mess.

And then if you can't reconcile in time, then you got to take a step back and go back to the old system, catch up and then try again in a few months. And it's like such an avoidable problem.

You might think migrations are painful. Data can be painful. It's definitely something we want to simplify. So, you know, kind of speaking that we'll talk a little bit about kind of the learnings and migration, obviously. You guys just announced, I believe it was beginning of October, I know it was sometime in October. You announced your 90 million Series A, which was led by Lightspeed and others. What milestones did you guys have to hit to earn that level of conviction. Right. And 90 million Series A's is a big number. So maybe talk, talk a little bit first of how that process went with the VCs.

Yeah. So we actually raised a seed round while we were in stealth. We probably even saw 80% of the funds left from that seed round. And the first thing we did was like we said, we're not going to go and think about what the next run's going to be. We're just going to put our heads down and build what is the best software that we would have loved as customers of a mid market erp. And while we were in that process, you know, kind of six months in, we skipped all the design partners. Like typical in vc, they explain to you have to go to design partners and spend time with them and then they'll kind of dictate the future of the product. We said, we will be the first design partner for this because we would have been customers. And that just increased the speed of development by orders of magnitude. And then, you know, kind of six months in, insecurities started to kick in and people are like, you know, you actually building something that people want or is it just like you're, you know, building in isolation? So we said, let's go out and try to get our first customer. Let's actually pick a publicly traded customer because they'll have the highest standards. But it was most important because they'll say no to us and explain to us very clearly why we weren't good enough. And then that'll give us like a build list to then go and keep building features. And to our surprise, we went to this customer and you know, we might have a few customers before then, but this is like the first big one. And we won the process. It was a few months process, but we won it. And they started running on dual entry entirely. It wasn't like a division, just the full company. And then we were like, holy shit. Well, we built this actually pretty good. And they were happy. They were closing the books much faster than we're doing before they were able to go live within a week or two. And that was before we had the migration engine entirely developed. And that was a big accolade. And then we said, okay, let's go out and hire a team of accountants to go and sell this. Most of our sales team today is CPAs or retired accountants. And we started getting customers left and right and the customers were onboarding and being very happy. And somehow, to answer your question, the venture community found out about this and they are like there seems to be the new wave of AI ERPs and there seems to be this one that's been quiet and just building in the background. And lucky enough we got so many offers during. It was all initially preempted. It wasn't that we went out with the intention to, to raise up, raise around because we still had 80, you know, 70 or 80% of the funds left and, and then we just got to select what we thought were the world's best partners for this. You know, we, and we got to pick three. Normally you pick one and we were lucky enough to be able to pick all three and, and go take on this big market. But it's, you know, back to business as usual. Like thanks for the money. We'll still spend it mostly on a product which is what we think is the leading indicator for everything. And yeah, we'll make more customers happy. And up until now, netsuite had no alternative. It was that or Sage. And it was like you're picking between two bad apples and for the first time ever they can look at what a modern consumer grade ERP looks like.

Yeah, it's so interesting out there right now. And Paul and I have had a lot of the other companies including NetSuite on the show just to talk about, you know, NetSuite just had Suite World and they've got their NetSuite Next and the promises of what they're going to offer. But we've had Campfire, Rillette and others on the show too. And it's, it's interesting. It's interesting but not unexpected. There is so much noise right now and not, not just noise, I mean by my count, I think something. And obviously you guys have a big chunk of this, but close to half a billion dollars raised in the ERP system in the last like what, 12 to 18 months. And interesting thing in the challenge that you have to be up against is it's a paradigm shift. People have kind of, there's the systems that they know in the world that they live in and they have to shift their thinking around that. But I'm wondering for CFOs are hearing all this and they're, they say, you know, I've got this incumbent system, but I don't love it. But I'm. This is what we use and it's stable and if it's stable, it's not always but trying to break into that market. And I'm wondering if you're talking to a CFO who's hearing all this marketing noise around ERPs and they don't really understand the difference. Is there like, what's the simplest test to tell an AI native ERP from a legacy suite that, you know, added later, added AI later, or is talking about adding AI? And I'm wondering on that, like, where do you see the first real ROI showing up for finance teams going with these systems in the next, you know, 12 months or whatever?

Yeah, I mean, look, anyone who's living in a legacy ERP knows that the kind of BS around those AI features that are getting sprinkled on top, it's almost like some senior VP or super senior VP is asking some senior VP to ask some VP to get some manager to get some AI feature. They don't even care if it actually works. They'll just get to brag to their public earnings calls that did the AI thing. I wonder if some of these people even use ChatGPT daily. But look, the most important thing when we go about building software is because we get to rethink everything from the very beginning is okay, does this transaction need to be touched by a human? And if not, does it come from an integration? Can it be done be a rule based deterministic engine? That way you have the certainty that it will obviously be treated the same way. And then if not, then how can AI suggest to the human rather than the human having to create it from scratch and the human still ultimately, like there's still either words like the integration was still either created by human ultimately, or the rules was also ultimately created by human. The human also has to prove the AI transaction. So there's always a traceability, but you don't have to do all the work of creating this journal from scratch. I look back at our prior company and there was a time where we even moved some of the M and A team into finance and we had like almost, almost 12 people at the peak running, running the finance, trying to catch up with the erp. We had five or six people that all they were doing was repetitive tasks every single month. Whether it was categorizing transactions from our credit cards or categorizing invoices and bills and knowing who had to approve what. And then at one point I saw someone taking a journal entry from the past, dropping it in ChatGPT early versions and then saying, but this is the new allocation schedule, the new accrual. And I was like, why is this not already part of the software from the very beginning? Why does the data not become available? So it's hard to kind of point that like this Is, you know, it's not a copilot. And this is what the legacy ERPs get wrong. It's like not a thing that you can easily replicate, it's how in each individual experience. Like, look, in Dual Entry, even the dropdowns, when you select things, the first two or three options that you get are prompted by AI. Then you get the full search.

But.

But it's like most likely than not, we're measuring how many times is the first option is the thing that the user was going to select. How can we be one or two steps ahead? That doesn't sound like a big deal, but when you add them all together, you know, you're increasing the amount of transactions via integrations and rules. And then in AI, every little interaction, we try to be one step ahead. It just makes things easier, right? Like from the very beginning to the moment that you finish your workflow, it might have taken you like 10 or 20 times less even creating reports, right? Like, we know that all these legacy ERP systems take so long to create reports. The fact that you can just talk to it and have it give you a version one and then you also see what parameters were changed. So you can easily then go and adjust all those things, all those micro interactions make your experience less, remove the friction. I like to tell the team is like, it's consumer grade erp. It's an oxymoron. So yeah, look, I think the CFO is the best test that we give. And this is why we spend such a big time making the bet on the next day. Migration is we let them use the software, we say, hey, go play with it. And this is something that the legacy ERP players will never let you do. Like I remember, even when I took our original netsuite demo, you had to tell them beforehand what you wanted to see and let them have time to prepare it. And you know, it's almost like rehearse it so that nothing went wrong and then they would show it to you. And I was like, well, what if I just didn't think of every case that could have come up? Like, you know, even in the call, I would say, hey, can you click on that to see how that dropdown looks? No, we're not allowed to go off script. You know, how can that be a good experience? And with Dual Entry, we're getting to the point now it's almost, almost on self onboard to the point of a demo. We're not there yet, but we're very close where we just give you a click. The data will flow through and with 99.9% accuracy you can go in there like look at your data live in dual entry to test out all the edge cases and then decide for yourself if this is the ERP that you want. And by the way, one big thing that we've done very early on is like we spend a lot of time making sure that dual entry is not just for one vertical, that it works for multiple verticals. Because when people get to the mid market, you know your business is becoming bespoke, right? Like if you're running a barber shop that is one location, you're probably going to run it 99% similar to the other barbershop. But if you're now like a 70 location barbershop, you might be doing like a shampoo line, you know, so now your inventory company or like, you know, if you're a SaaS company, you might have a hardware line or, or you might have like a big office with tons of fixed assets. So companies get bespoke as they become larger. So you want an ERP that just is not just SaaS focus. You want something that's, that's focused on multiple verticals because it gives you the certainty that you're not going to have to then do yet another migration to another system.

Ever feel like your go to market teams and finance speak different languages? This misalignment is a breeding ground for fail, impairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet QFLO AI the strategic finance platform. Purpose built to solve the toughest part of Planning an analysis. B2B revenue. QFLO quickly integrates key data from your go to market stack and accounting platform, then handles all the data prep and normalization under the hood. It automatically assembles your go to market stats, makes segmented scenario planning a breeze and closes the planning loop. Create airtight alignment, improve decision latency and ensure accountability across the team. No, I appreciate that. I have a question. I know you guys promote the 24 hour. You've talked a little bit about that next day migration for initial data loads. Kind of walk through where that holds, what the challenges are around that and what's the realistic plan for someone to be fully up and running. Right. It's one thing to have your data there, but there's training, there may be rework, whatever. Especially for you know, mid market companies that have multi Entity, you get over into Europe, multi currency, that complexity. So maybe talk a little bit about.

That next day European migration. It works with not every single system, but the most popular ones. So we'll connect via API, all the data will be drawn and then there's a few steps that we do, and it's all like product light. So you first go in, you create all your entity tree. So it can be multi entity, it can be international. And then you would connect that to the origin system. If the system is single entity, you would connect it multiple times to each entity. And we'll map that into a unified GL so that we'll create a GL every time we see a new account, and every time we see a repeated account, we'll map it. You always confirm these mappings, so we got something semantically wrong. You can always override it. AI will also take a pass. So if you named AR and accounts receivable, we'll still try to make that match for you. And once everything's mapped, it's now locked in a deterministic way so that you don't, you don't have to trust that AI will get it wrong in every transaction. And then we'll start syncing the data over that. Generally, depending on the throttling that the origin system has on their APIs, it might take, you know, a few hours or up to 24 hours. And now the data is going to live in dual entry for the first time already for you to consume. So you can then run a P and L and a balance sheet and you can start segmenting. We'll even pull in all the classes, all the dimensions, all your items, your vendors, your customers, Everything is going to be pulled over and then you can just see it and then test it on your own, see how you'd run a P and L or a balance sheet. Now we recognize that like, implementation is much more than that, right? Like, we want to move over. We also want to show you around the things that you can do now, like, you know, whether it's period locking or rules or the new AI features. So we'll schedule a few calls with your team. So we actually have two migrations that go on. One is during the demo phase where we do this and we just give you access to it and then you get to play around with it. And there we guarantee like 99% accuracy because sometimes there's some edge cases that would fail and then we'd treat those manually. There's always a human, by the way, in the implementation side that verifies the final ledger. But in implementation, we'll run the same script and then we'll verify it in the end with a human. And then we'll schedule all these calls with your team to make sure that you're ready to go up and running. We actually schedule four calls, but if you want to do more, they're free. We're here for you. We want to make sure that you take most advantage of the software. And that's what we call the broader implementation process. Some, some companies or some teams decide to like, use the opportunity of the migration to redo their chart of accounts, to redo how they do workflows. Sometimes they even use it to change external providers. I mean, they might go to a different spend management software. And we're there for that journey too. And we'll connect you, you know, once you feel ready, that the data is now flowing through. By the way, this ERP migration, you can refresh it every 15 minutes so you can still get the data. So there's no stop that data can keep on flowing. But once you're ready, you unplug your integrations from the old system and you plug them into Dual Entry. We verify that things are flowing, that there's nothing left in between, and then you're off to the raises. And it's all transparent in product led. So you can see how it worked. You can see how we got to that answer. You don't have to talk. You don't. You want to log in at 11pm and see why things happen the way that happened. You can see it right there. It's all accessible. So it's not like you don't have to depend on a 4 quarantine that's doing spreadsheets and you know, you don't even know what, what stage of the process they're in. And that's been a winning feature and it's allowed us to really demo the product in depth. There's also surface bugs, early days, because then you give them the product and say, hey, this doesn't work. And they're like, oh shit, we got to fix that. But I think it's good, like transparency and having people use the software is just the best and most healthy thing that you can do. Is that you how you really earn the software? I like to, you know, when, when we started in building Dual Entry, we said ERP is like chronic back pain. It just gets worse over time. People know they have it, they're aware of it, they want to get rid of it, but they've also heard of horror stories of people who go get back pain surgery and then come out with more pain with 100k bill and it's a nine month recovery. So imagine showing up to somebody with back pain and saying, hey, I am a new guy. Like, you know, you've never heard of this hospital, you've never heard of me as a doctor, I'm the new guy in the block. But trust me, I have a better method to get rid of your back pain. But you have to go through this implementation process which is surgery. You'd be like, no way on earth I'm doing that. There's just no way. So we said like, how can we fix that? How can we make it so it's not back pain surgery and maybe it's more like a massage where people can actually see the results quicker without it being high risk and high cost. So that's how the next day and you're willing to take a massage from a random person that you've never met, whereas you're not willing to take back pain surgery from somebody that you've never met. So that's how we build trust. We try to show you the benefit upfront and get rid of the stigma that it's going to take nine months and cost you three, you know, two, 300K and it's going to get you fired. Yeah.

So it's, it's refreshing to hear you say this because I think about on these migrations it's structured tables, everybody. This table contains this data in this system and this data is where it needs to go over there. It's, it shouldn't be rocket science. So all. Anytime you've made the move in the past, it's dumbfounding. And the first time you do one of these implementations it's shocking to think, wait a minute, this isn't solved for. This is a bespoke process to do the migration. So that's a huge thing to just get the data over. And when you say it, obviously there's an API, there's a map, you understand the schema of the database on each side. It should be seamless. But this might be telling of my age and my willingness to do this startup where I, I spent a good part of my career in, in startups myself and thinking about deciding you're going to build a mid market erp, to my mind that's such a bold thing to do. I would, you know, if I were going to build a GL system, I might say let's start with SMBs that don't have audit responsibilities and it's a, you know, it's a lower stakes game and we can do it cheaper or whatever and we could just have it provision themselves and all that. But then when you start thinking about mid market companies and public companies and you have to factor in so many things, the entire finance and accounting process, but then things like maker check approvals, period locks, reversals, exportable evidence for these AI generated entries and I think about how challenging it is to go after that market when, whether it's the devil you know or whatever, where you're to going, going against them. But the people are nervous because you have auditors who are going to be looking at everything and I'm, I'm wondering and I'm imagining that you talked a lot to auditors and I'm wondering what kind of feedback you might have received. And I'm thinking about icfr, socks, all these things that these larger companies require. And if you've hit any roadblocks there or how do you even. I'm just rambling now, just the whole thing just seems so complex to me to take on just, you know, we're going to build this from scratch and cover everything that an ERP needs to do. Like how do you map that out, work with the auditors, get that level of trust and that. I know you said you know AI where AI makes sense and then deterministic in other areas. But there's just so many steps and unless you are an auditor who also happens to be a coder, it just seems like a lot. So I don't know, I threw a lot at you right there but I'm trying to figure out building it, making sure that it meets the needs of these high demand clients that are in that mid market space.

Yeah, we, we broke all the rules of entrepreneurship by starting with larger mid market companies. Every advice we ever heard was like start with smaller companies and work your way up. But if you talk to the smaller companies, they don't have as much pain. They do have some, but they don't have as much pain as these mid market teams have. As things start to scale, you need to add controls, you need to go through audits. So we just went straight to the pain. Like when we were running our product company, we didn't have that much pain early on. It was really when we hit the mid market and the growth inflection that the things started to break and we were willing to pay way more money than what we were paying even netsuite to solve the issue if it was actually going to be solved. So for us it was like it's an obvious problem, it's just a hard solution. And those are the problems that we like because we like to build good executing teams. The thing about engineering and building product is that difference in the bell curve between teams is not just like 1 or 2x, right? You know, you can get an engineer and a product team that can deliver 20 times better, faster, higher quality features on another team from another company. I mean the contrast between the average engineering team and one of these legacy companies and our engineering team is just like, is mind boggling. You probably put a whole stadium of these people and they wouldn't ship as fast as some of our teams ship with the attention of detail that we ship. So if that's true, then finding a hard problem and then just trying to building around that problem a 10 or 20 times faster team, that is the winning formula because the problem is obvious. You either have market uncertainty or you have hard problem and competition. And we'd rather pick the latter because at least you know, you're solving something that matters to the world. And yeah, look, we involved auditors from the very beginning and we asked them what they would like to look at. And you know, auditors are, don't have this, the same negative incentives that, that some of the other partners have. They're actually quite, quite benevolent. They want to get the audit ready. They're competing to try to give you the audit standard at the lowest cost and make you as happy so that you return next year. It's quite a different relationship. And we've gotten, you know, some of them have invested in dual entry, like tons of great feedback from what we've seen from when they see the product because I think they see how it's actually easier to do their job in a modern system. They also have to log into some of these systems and they also, you know, they, they live the same pain that, that we lived as, as users. So we're also helping them and it's quite fulfilling to be on this, on this journey.

Something you said resonated with me because I do a lot of AI and finance training for finance and accounting professionals. And one of the things when I talk to audit groups that I talk about is what if AI could help you. Instead of doing sampling, you get the full data set and AI helps, you know, you're not just dragging samples, you're looking at everything. I'm wondering if that's something we're not going to change regulations and all that on this on this podcast, but I'm wondering if that's something that you've talked to auditors about because of the AI nature of the platform and the data that you have sampling versus doing an AI run on the full data could be an option. Is that something that you guys have looked at or considered to talk to them about?

Yeah, I mean, AI and technology in general is making this better and more accessible. And I think software also can play a good role there as an intermediary of trust. Right. Like the thing about these legacy ERPs is that people have just built so many workflows around them and have used so many external tools because they're so deficient that the auditors have to go into your spend management tool and audit that, and they have to then go into your equity tool and audit that and then make sure that the data on that tool matches the data on the ERP because it's manually inputted. So in that manual input, a mistake could have been done or fraud could have been done, who knows? But if the software takes sort of the broker of trust in here or it says, look, no, this invoice came from here and we have a link and we're hashing this transaction so that you actually can modify it from when it came. But if it doesn't be modified, then we can flag it as modified and actually we're pulling in over the attachment. It's all centralized and high resolution. It is much easier for you to do this audit. You go into one system with one set of permissions, then you can look at what you need to look and sample what you need to send example or like you said, just feed it to a model and say, hey, you know, AI agent, pick up 30 invoices and pick up 30 transactions and see if they actually match. And the categorization is accurate. So like software actually can be a good broker here. I mean, we've, we're inspired by a company called Vanta who does a lot of compliance. They're a broker in the SoC2 compliance space. Right. Like they're the ones doing the check with your aws, making sure your infrastructure is correctly set up, or with Google Cloud. Right. Making sure it's properly set up. But then those reports are then used by auditors to come up with their conclusions a little bit faster. And it makes everyone's lives easier. And you trust that Vanta did the API check with this system. Right. You'll start to see this happen more and more and make this auditor more efficient. And as the Data becomes easy to query and easy to validate. The room for error is going to go down and down and in theory they should make. You know, there's all this talk about reporting being know semi annually or annually instead of quarterly. It was like reporting should actually be more frequent and there should be more data available to public markets or to your investors. So I think we should be heading in that direction.

It just opens up so many more possibilities when you have this level of baked in AI on it. And it's a completely new paradigm. Okay, so we've hit the part of the show. So this is what we do. We take our guest background, company background, and we feed it into our favorite AI machine. I'm using GPT this week and we have it generate 25 random prompted something with, you know, fun, loosely financed and tech related questions and have the AI spit them out. So we asked two out of the 25. So you're not about to get like a full like 25 question speed round. But Paul and I have different approaches to how we select the question. So Paul, maybe I'll let you go first with your methodology and then I'll do my selection next.

I believe in choice, unlike Glenn. So here are your options. No, I'm just kidding. One, you can randomly pick a number between 1 and 25 and whatever number you pick, you get that question. Or I could have the random number generator randomly pick a number and I'll ask that question. So which do you want?

I'll do nine because of the 90 million. Let's do it that way.

All right. Interesting. I'll be curious to see what you say to this one. What's your superpower inside a spreadsheet?

Oh, wow. To try to get you out of it.

That's probably the perfect answer right there. Right.

Although I will say it is a very hard thing to replace because it's ultimate malleability. And our job at Dual entry is how do we get one step further in that FP and a process so that you do less things in that spreadsheet. But we ultimately recognized like taking away the canvas from an artist, you need to have that at the end of the day and yeah, so. So our goal is to, to take most of it away from, from you, but not all of it.

Yeah, but let the FPA guys still have their spreadsheets because they love.

Take every account out of the spreadsheets and just leave that to the FP and a team is maybe, maybe the way to say it.

Okay, well for my question, I feel like I'm just handing over the keys to the AI and it selected or it came up with the questions, so I just let AI select. So let. Hang on one sec. Let me have AI pull up here. And that was a weird one, Paul, that. I don't know where that came from. The. The spreadsheet question.

Yeah, the first time we've had a spreadsheet question.

But hey, it might have had some injections. I'm gonna blame it on prompt Glenn.

Yeah, probably, since I didn't create him this time. You know, I'm gonna stick with my story of blaming you.

I'm scared. Or re.

Hallucination. AI hallucination.

It.

It's one of the.

That's right.

We'll let our audience decide.

This one's way better. I don't. Maybe I should have. Maybe I should have proofed the other ones. I like this one, though. Would you rather debug a broken AI model or explain gap to a room full of engineers?

Gap to a full room of engineers. That's what I do all the time. Yeah, you'd be surprised. Like, accounting is quite like an engineering, like discipline. It's very logical, it follows very clear rules. The rules can get updated, but they're there. There's no. Like, there is room for subjectivity, but it tries to remove room from subjectivity and, and it all adds up to zero, which is the most magical part of it all. So I think it's actually quite compatible. Or most. I don't know, at least they say that to me. But most of our engineers actually quite enjoy accounting.

Makes sense. I mean, it's that same sort of engineering mindset. And it's, it is deterministic and it's logic, rule based.

Right. You know, if you got it wrong.

Exactly.

At least in total, you might not know where the individual problem is. But in total, you know, if you.

Got it wrong, if it's not adding up to zero, there's nothing wrong with the software, you know? Yeah, yeah. No, it's. I, I think it's actually quite compatible. I wouldn't even try to debug a model because the model, you know, the, the bug might be there once and then you run it again. It's not there anymore. So there's. That's why we always prioritize in dual entry. It's always first, you know, like, can it come from an integration? So has this work been done before? And then if not, can a rule determine it? Can we create. And AI can be helpful in creating these rules, but ultimately you approve the rule, then only thereafter, then AI would take care of it. Which is the most malleable? It's kind of like a catch all. But a human always, always have to verify it. Because even if we test it and we say you got to write 999 times out of a thousand, you know that one is very important. So a human always needs to be the one giving it the thumbs up. Now, the beautiful thing about accounting and marrying it to the model is you get an instant reinforcement, right? Like if somebody says, this transaction was great and then posts it, that's a thumbs up. You can use that to feed it back into the model and say, hey, you did a good job. Do more of this. Whereas the person has to go in and edit it. It says you did a bad job. But by the way, here's the right answer. So next time. Now you know what the answer should be like. And that's a very unique thing to accounting, which is why we're so bullish on AI and accounting in general.

So RLHF with every transaction is what you're getting. Well, Santi, this has been great. We appreciate you coming on and hearing about dual entry and everything you guys are doing, and we wish you the best of luck going forward.

Thank you. Thank you for having me. This is great.

Thanks for listening to the Future Finance show and thanks to our sponsor, qflow.AI if you enjoyed this episode, please leave a rating and review on your podcast platform of choice. And may your robot overlords be with you.