# MoonPay acquires AI accounting platform Entendre, founder Kareem Khattab becomes VP of Applied AI Incline, a Stabledash show. Sep 25, 2026. Kareem Khattab. https://live.stabledash.com/shows/incline/episodes/how-moonpays-entendre-turns-accountants-from-doers-into-qLnUdb-Hb_8 Khattab wants AI agents to do the bookkeeping while accountants review it, and he says the company to beat is a frontier lab, not a legacy software vendor. MoonPay acquired Entendre, the AI accounting platform for the stablecoin economy, in June 2026. Founder Kareem Khattab became MoonPay's VP of Applied AI as part of the deal. In the interview he set out his plan for agents to do the accounting work while people review it. ## From doers to reviewers "Our whole goal here is to make it easy for accountants to basically go from doers to reviewers," Kareem Khattab said, describing how Entendre automates whatever workflows are unique to each finance team. He says the company does not measure itself against older software. It measures itself against what he calls a "super AI pilled accountant" working in Claude directly, and asks whether it can be faster, more accurate and more efficient than that person. ## A benchmark built on tax returns Entendre is announcing TaxBench, its first public benchmark. It tests whether AI models can take a CPA's inputs through a full tax return. Kareem Khattab said the workflow covers the whole tax return process, including data entry into tools that have no API, such as ProConnect. He called data "a trillion dollar market" and said much of it sits behind enterprise walls. ## Into money movement Kareem Khattab said Entendre never got into money movement before it joined MoonPay. With MoonPay's infrastructure and teams like Iron, he expects it to move into cash management and transfers between accounts soon. He said his customers have hundreds of wallets, hundreds of bank accounts and a number of legal entities, often only a few years after they launch. ## Headless accounting by 2030 Kareem Khattab predicts that by 2030 accounting will be headless. In his view, founders will ask about P&L, burn and runway over iMessage, talking to a Siri-type assistant and getting answers back by text. Kareem Khattab does not see older vendors as the competition. "The legacy players in your field are not really your competition," he said. "It's really Anthropic that you have to compete again." He added that a founder who can beat OpenAI and Anthropic will build a unicorn. That standard is behind both TaxBench and Entendre's plans inside MoonPay. ## Key takeaways 1. MoonPay acquired Entendre, the AI accounting platform for the stablecoin economy, in June 2026, and founder Kareem Khattab became its VP of Applied AI. 2. Entendre is announcing TaxBench, its first public benchmark, which tests whether AI models can take a CPA's inputs through a full tax return. 3. Kareem Khattab predicts that by 2030 accounting will happen over text messages, and says AI founders have to compete with Anthropic, not legacy vendors. ## Chapters - 0:00 Entendre joins MoonPay as an independent business unit - 1:47 Stablecoins and LLMs shaped the founding thesis - 4:09 Entendre handles accounting for MoonPay's operating system of value - 5:54 Agents do the work and accountants review it - 8:53 Workday experience and crypto complexity revealed the problem - 10:21 TaxBench tests models on full tax returns - 14:15 Pitching Entendre as the harness around LLMs - 19:18 Getting older accountants to use the tool first - 22:31 MoonPay infrastructure moves Entendre into money movement - 24:10 Accounting goes headless with answers by text by 2030 - 26:57 AI founders compete with Anthropic, not legacy vendors ## Transcript **Host** (0:00) Uncharted territory. Where road maps are useless. Conditions don't negotiate. The terrain doesn't wait. Where opportunity abounds, but so does death. The builders who understand what's coming aren't mapping it. They're already **Host** (0:46) Kareem, what's going on, man? **Host** (0:48) Yo. Yo. That's good. So happy to have you, man. We're happy to have you back, and thanks for your patience. I know it took a second. **Kareem Khattab** (0:54) No worries. Yeah. It **Host** (0:56) be confetti time. I feel like there's a lot to sell No. This is a 100% confetti **Host** (1:01) time. **Kareem Khattab** (1:02) Yeah. This is Oh, really? Congratulations, **Host** (1:05) man. How how has the transition been? I mean, did... Was it... Did did you just hop into it right after you met with Drew for Sable-minded, or is it felt like a slow ramp? Or it was like, hey. No, dude. You get you get the weekend, and then now you're on the MoonPay team, and we're rolling. Yeah. MoonPay team, and we're rolling. I... **Kareem Khattab** (1:21) Yeah. I... There was no time off. We just jumped right there. So Yeah. Yeah. Man. Yeah. Todra's staying, obviously, you know, independent business unit. And so, you know, being acquired is amazing. But when you're truly merging with a large organization, **Host** (1:40) it's just another day at the office. So Okay. So we love we love the founder journey here at Stable Dash. And I know we we dove deep with you. Your episode came out. You're looking good. It fit in Miami and wandering around and all that. But I want us to rewind for a second and go back to you had all these startup experiences. You kind of observed the field, saw what worked and what didn't. Then you start on Tundra. When you're starting, are you, like, beginning with the exit in mind back then? **Kareem Khattab** (2:13) No. I mean, I think I think when you get the itch to start something, you don't really... You're kind of, like, agnostic of the outcome. You're more excited about just getting started. And because it was my first first go at this, it it it was really a lot of just learning. Like, I didn't have the foresight to know how this would would ultimately conclude. But what's interesting is, like, always go for the biggest problems that you can find and, ideally, the biggest problems that you're good at. And what was the... When you... What was the founding thesis? When... What was, like, the problem you looked at? And you said, this is a big problem. I should really go for this. Yeah. Just like we mentioned on on the show with with with you guys in Miami, financial operations, accounting is just this kinda unsung hero of of prosperity. A lot of large businesses, a lot of wealth transformation happens because of the people on the back office side of things. And that's... And because of the promise of stablecoins and this new way of moving money and storing value globally has has has begun to exponentially increase. The thesis was just simply that money and value transfer is is changing, and we need tools for tracking digital money. And luckily, it was also at the the the the rise of LLMs. And and and at the time, they weren't called agents. They weren't really easy to use, but we stumbled our way to ultimately a very exciting product that customers of ours really love. So **Host** (4:09) Where where does Entendra start to fit in, like, to the greater MoonPay family now? Because MoonPay, you know, honestly, have been on just an acquisition spree, it feels like, in the last twelve months across all these different sectors. And so somebody looking at it from the outside and be like, I don't really understand what MoonPay is doing. Help us understand more of this this thesis and where you guys fit into it. Yeah. No. We're incredibly excited to work with all of the MoonPay **Kareem Khattab** (4:37) teams. MoonPay is really just building the new operating system of value, and that kind of crosses tokenization, trade, spend, and every financial operation that happens on chain, including things like on and off ramps, of course, that that that on chain movement that that... Pardon me. The the fiat aspects and the intersections of fiat as well, all of those have a financial impact, whether you're a consumer or a business. And because of MoonPay's, like, deep integrations and partnerships and moving institutional financial institutions and and enterprises, that's a huge opportunity for Entadrip to provide this kind of paternity solution that is compatible with MoonPay products so that if you're a CFO on the board of directors, a finance manager, an accounting controller, you have that peace of mind when you're, you know, when your teams are adopting and enabling stablecoin adjacent or native products that that, you know,. **Host** (5:54) Yeah. What... When you think about... You know, it's funny. We're we're having... He's actually my CPA. He just, you know, launched Solteris Labs. I don't know if you know Cameron Brown. He'll be coming on the show. He says, hey. Where do you fit into this conversation where we've just met? You know, we just met Kareem, and he goes, two completely different sectors. And I think people start to really misunderstand AI and accounting and even, like, what you guys may doing... Be doing for accounting teams. And it's not that you have agents that are... You know, like, you see Coinbase going out and say, hey. They're registered financial advisers. Your agents aren't registered CPAs. You know, maybe what's what's the difference there, **Kareem Khattab** (6:31) and how do you even think that should help inform people for how they should think about working with agents instead of agents just replacing them? Yeah. And I and I absolutely love that you guys pulled up the website. So, yeah, I think a few things. One, on Tonjas, really, our core focus is agents, you know... And this is the moon pay kind of motto here. It's, you know, agents supported by humans. And our whole goal here is to make it easy for accountants to basically go from doers to reviewers and being able to essentially automate whatever workflows that are unique to them. But with the with the visibility, traceability, auditability that they expect from any financial system. And, also, we're not a rip and replace. We are simply going after the parts of your business on the back office side that, you know, you have to determine whether you're going to hire or you're gonna, you know, burn the midnight oil to do yourself. And, you know, agents and AI is a option on the table for you when traditionally, you know, you might not have thought you had the technical capability or know how. So Antandia's real goal here is automating financial knowledge work, and we do a lot of things on the back end that we abstract from these users to make sure we're routing to the right model, that we're evaluating model performance, that we're building the right tools so that agents are, you know, connected to all of your core systems. And you go from becoming, you know, top 1% of finance teams to the top point 1% of finance teams, and that's really the gap. And you can almost... You know, for me, like, our benchmark is not... You know, we don't compare ourselves to, like, Intuit or NetSuite. Right? Yeah. They... You know, they're... These legacy tools are deeply entrenched in in in millions of businesses around the globe. We're actually measuring our performance against a super AI pilled accountant. That's the measure that we hold ourselves to. And can we be faster, more accurate, and more efficient than some expert level AI accountant, you know, doing it in Claude directly. Right? And that's that's kind of the bar we hold ourselves to. Yeah. Yeah. It's it's interesting. I'm thinking about sort of, like, **Host** (8:53) be the tourniquet, not the aspirin, like, the extreme pain. I don't... When I think of, like... What's accounting pain? I think of, like, busy season. Think about your friend at or something, you know, up all night for six months or something or three months in busy season. Were you experiencing... You're talking about, like, all this financial knowledge work and the just slog of it. Is that what you were experiencing? You're like, wow. This is a big problem. We gotta fix this. I mean, it's funny. It it... It's just like when you've seen it **Kareem Khattab** (9:24) a few times, it it kind of just becomes the same for every business. And, you know, I Mhmm. Had started in, you know, building ERP solutions at Workday as a software engineer. And then, you know, from there, was a product manager on the payment side of the house. So a lot of accounts receivable, quote to cash, financial reporting, that just was super fragmented, manual. These were all things that just were recurring themes regardless of the type of business you operated under, whether it was fiat or crypto. And the fact that it was crypto was even more paramount to get this right because a lot of our customers were heavily regulated, had to have multiple subsidiaries. We're dealing across multiple chains. So the complexity kinda multiplied than a traditional business just operating in The US and accepting, you know, ACH and credit cards. And if you actually go to our resources tab, you'll see our research. And this is kind of the the the point I was making about financial knowledge work. And if you click our tax bench one, which is the first public benchmark that Entendre released, and we're actually announcing it in the next few hours, you'll see how, you know, this is the type of this is the type of level you have to be at to be, you know, competitive in all of this kind of... This this really crowded market of AI vertical SaaS products? **Host** (10:58) Wow. Yeah. I was actually gonna... Z, can you stay on that for a second where we're seeing these these dataset scores? So kinda walk us through how we should be interpreting, you know, the... This scoring. I think we... You open X and you see, oh, I compared all of these models. It's obviously a relevant conversation in our business and our our engineer, you know, is is also on top of it going between OPUS 4.8, 4.7, Fable five. That's actually interesting to see Fable five in between four point eight and four point seven. So help us understand, like, how should we be evaluating these these models and building with them and using them inside our businesses or our business and everybody else's business? **Kareem Khattab** (11:37) Yeah. So look. Entendre is one of those rare companies that has thousands of hours of CPA labeled data, a CPA approved workflows, you know, CPA approved outputs. And because we have this kind of unique data... Proprietary dataset, we're able to then take that data, anonymize it, change it completely so it doesn't actually represent our customers' data, but then figure out, hey. For those workflows... And this workflow's really, really interesting because it's actually the entire tax return process, which includes actual data entry into tools that don't even have APIs like, like ProConnect. So can the agent actually, given all the inputs from the CPA or your tax accountant, produce the work paper and then enter it into these super archaic tax tools that then, you know, you fire off into, you know, the IRS. And so, like, we kinda did the whole kind of suite for this workflow. And based on multiple attempts per model, we get kinda this mean average score, and we have our partners reviewing these outputs themselves and telling us where the model went wrong. And they're incentivized to do this because if they're going to use Entendre for, you know, 600, 800 clients, They wanna know that they're using the absolute most accurate model. Wow. What... Did this surprise you to see where Fable five landed? It was funny because we... As soon as Fable launched the first time, I know, like, it kinda... We got rug pulled by The US Oh, this crazy. Yeah. Yeah. We we had, a term internally, like, we were fable maxing for a week straight. Like... And we actually got this benchmarked in before the rug pull. So, yeah. And this is really cool because a lot of the AI labs themselves are extremely interested in this data. I think data is a trillion dollar market, and a lot of it's on... Well, it's kinda behind enterprise data. And so you could think about the accounting firms that we work with, the finance teams that we work with. All of that is so unique and so valuable to improving the models. And on... Entendre just gets to be kind of this... It could be this thought leader in **Host** (14:15) in how these models should perform and how they should be built and where they fall short. And it could be a huge revenue driver for for Entendre and MoonPay. So when you're having discussions, Kareem, and you're talking to enterprise businesses that are considering adopting this, and like you said, they're sort of incentivized to look at other things that stack up. Are you using these terms like agentic finance? Or, like, how are you talking about this? Are you sort of emphasizing the novelty and the buzzwordiness? Are you kind of talking about speed, talking about efficiency, and talking about **Kareem Khattab** (14:50) more sort of common language. Yeah. I think we wanna we wanna, like, make it accessible to everyone to understand. And the way that we sell it is if you're not the model, if you're not the LLM, you're the harness. And you need to build a harness around these LLMs that is unique to your business. Right? Obviously, if you think about AI adoption within enterprises, you know, engineering is at the tail... Like, the far end of that ladder. They're able to produce a lot more than nontechnical departments within a business. And so we're not really worried about engineering typically. And a lot of the last few years, the narrative has been about coding benchmarks and evals around coding and all the models competing for for coding. But the kinda next iteration is, well, if if you wanna be a AI native organization and you wanna be one of the top performing AI organizations in your industry, you need to get everyone else up the ladder. And so we kinda talk about it in that context, how Entendre can enable people to be... Go from, you know, the l l zero to l two, l three, l four. The goal is everyone gets to the top, and now they're audit... Automating more than a 100% of their job. So now they're actually working on strategic things and things that no one had time for or would require a new hire. **Host** (16:27) And so that's the sell from from our It makes perfect sense, and I like the kind of moving people up the ladder, even in the sense of doing more human work, doing more creative or valuable work, because there is... Man, there's a lot of cells that have been moved around manually in a spreadsheet just when it's like, hey. You know, there's there's so many things where I can imagine people kinda knocking their head against their computer screen. I remember years ago... It's funny because this is five years ago. It feels like, wow. It's really the past. Having a conversation with a financial executive at Brookfield that was a consultant for Deloitte at the time. We're talking about, like, going through the accounting organization at that that massive organization in the same way that you are now. It was early AI adoption. It was talking about automation and efficiency, but they were they were focused on... We're not trying to cut all our people and cut our headcount and eliminate the accounting division. We're more thinking we want Cheryl and Tom in accounting to actually be able to go home at six and, you know, to have done something valuable and to have actually found a way to save our company money. Sometimes you get stuck just keeping the lights on in a financial organization, and you don't go home at six. So it's like Yeah. What if we could empower a future like that? That's what I'm hearing a little bit. Yeah. Well, also, like, **Kareem Khattab** (17:54) one thing to to note is, like, if you're an individual and you have anxiety around AI and your own kind of career security, like, you are incentivized to move up that ladder because if if you're a management at a at a big organization and and you're, you know, able to... With with the telemetry that, like, Entendre provides, understand who's using AI in the most valuable way. If you're going up that ladder, why would anyone wanna get rid of their top performers on that ladder? You want... If if you're an l five and on the ladder and that's the highest ladder and you're automating more than a 100% of your job, you're pretty secure. I I, as a leader and founder of a business, like, I would 100% keep everyone who's progressing up that ladder. If you're a nonadopter, that's a different story. You might not be around much longer, but that's where, you know, when we go into it, we we talk in those terms. We want to move everyone up the ladder. We want everyone that, you know, is part of the team, part of the organization to feel empowered and to be, you know, again, from doer to reviewer and and then going after... Not just low hanging fruit, going after the more strategic and more complex tasks, which then you'll see the models start to underperform as the, you know, workflows get more complex and are... Require more and more human in the loop **Host** (19:18) per se. So So how would you Go ahead. Go ahead, Zach. No. Was gonna say, so how would you approach that accountant that maybe they're 50 years old, you know, they've been in the industry for twenty years. They are obviously gonna be in it probably another fifteen years. You know? They wanna retire at 65. They wanna move up to the ladder. Where do they start? Is this a... You get them in front of... You build out the workflow. You say, hey. Right now, all you gotta do is review the work, but also we want you to do this on the side. What does it look like to bring them up to speed? Yeah. I think it's... I mean, maybe it's a coincidence, but I I think, like, we've been working with a lot of super gung ho young accountants, and and **Kareem Khattab** (19:56) maybe that's just, like, the nature of the customers that This is where the space is going. Yeah. Just... But, you know, obviously, we we wanna sell into the banks and the financial institutions and the asset managers. So maybe that's something I'll have to get good at, Sell it to the 50 year old accountant. But I would just say, like, again, what's beautiful about this is because of all the applied AI engineering that we do at Entendre and our, like, attention to detail in crafting something that's, again, accessible, easy to use, reduces friction. Right? Like, ultimately, we want to give the account a frictionless experience when it comes to actually getting work done. We don't wanna have an opinion on what the work should be, So it goes back to routing all their requests in a centralized place that we can evaluate, allowing all of their core systems to be connected so that if they have a query or an instruction or they need to, you know, reference... The LLM needs to reference something that they're mentioning, like, is unified. Then we start to build intelligence around what are the things that they're doing the most, and what are the things that they recurringly ask the LLM to do. And that's the Entendre harness. And so they're they're going into our harness, and we're the ones from there compounding their efficiency as they, you know, just start to use it. And that's kinda my number one priority is just get them using the tool, and it becomes kind of second nature after they see... That kind of like that spark **Host** (21:39) in their in their brain goes, okay. I see what Oh, yeah. To do. Yeah. I love the the framing of of a harness, and I can even think about that moment with my my own, you know, my own dad where he's like, dude, have you seen this thing, like, Copilot from Microsoft? You know, he he works for, you know, Kaiser in the hospital in the hospital. He's like, man, this thing helps me write my emails. It's awesome. And it's like, this is the guy that has never had Facebook and has never had Instagram. You know? And it's just Yeah. But, again, when when you reach that use case of just, like, helping them with some of the most basic things they do on a day to day, it just completely blows their mind and helps them go zero to one. Alright. Here's something I'm thinking about with this... These examples screen. It's like, you're talking about human in the loop. You're also talking about automating a lot of the sort of down rung, down the ladder tasks. Is there anything that you say, we should never really automate that? A human should always be in the loop. Or **Kareem Khattab** (22:31) it would freak me out if we gave all this to agents or something like that. Well, it's funny because before joining MoonPay, we never really got into money movement. But now with MoonPay's infrastructure and working with teams like Iron, I actually think you'll see very soon, Antundra starting to jump into the the the the cash management game and the transfer between accounts as at least a a starting point for us to kinda test people's trust with... Within our system. And and it's fascinating because our customers have hundreds of wallets, hundreds of bank accounts, a bunch of legal entities. These are very complexly structured businesses, and they're young businesses. Right? These are businesses that have been around for only a few years, and they're already structured like large enterprise organizations that have been around for decades. And so this is a real pain point, and I think having a partner like MoonPay with the money movement and all the infrastructure they've built is going to give us peace of mind internally and then also give our customers peace of mind. **Host** (23:39) Exciting, man. What keeps you up at night? To answer your **Kareem Khattab** (23:42) question directly, it was always money movement was the was the black box. **Host** (23:47) Money movement is the... Well, I'm curious as you're talking about what's being enabled with MoonPay. I know that, you know, an acquisition, we we threw the confetti your way. We're really excited for It goes both ways, though. Right? When you're negotiating that and you're deciding, is this good for Entendra? MoonPay's deciding, is this good for MoonPay? If you fast forward and say, like, now that we are together, what's possible? Take us to the the MoonPay Entendre 2030. What's happening? Yeah. What are we gonna hear about in a few years? **Kareem Khattab** (24:21) Man, I think there's... What I love about MoonPay is their, like, push to going headless. Like, you should be able to move money, spend money, spin up cards, trade, perks, meme coins, 1 to 1 stablecoins, whatever it ends up being, and you don't even know it's MoonPay. And I think for us, like, that resonates with us because we believe that this move from the old way of doing things around RPA based automation and legacy software, everything is going headless, even accounting. So if we can make accounting a headless experience where someone's like, what's my p and l? What's my largest expenses? How... What's my cash balance? What's my burn? What's my runway? And all you have to do is type, text, or speak into your phone and get the answers. I think we're just moving into a headless world. I think by 2030, man, you'll just be doing accounting in your phone... On on iMessage, and you'll just be getting a text. You'll be, you know, talking to, like, a Siri type AI assistant. Man, that's a great great point. Can I name my AI accounting assistant, Karim, in your honor? Hey. Of course. We'll add a setting, and you can give it a name. I think naming your agent thing is, like, archaic. Whenever a company, like, launches an ape... Like, an agent with a name, I always get a little... Just like, that was very 2024. **Host** (25:50) So 2020 I You know, Polygon, and she... They shared a post the other day talking about, you know, their agent is Poly. But there is this interesting paradigm shift, and we're even experimenting with it with with Claw tag and using Clawd within Slack versus, you know, in the Clawd interface and trying to unify those operations across the organization so you can start to see more and learn from what other people are doing it because sometimes we really start to operate in these silos. So I I'm thinking about that a lot too, candidly, which is like, how are we going to be engaging with AI? Where is it gonna take place? Is it the terminal? Is it a chat? Is it your text message? Is it a little bit of both for now? I don't know. **Host** (26:37) Whether it's so 2024 or not, I'm thinking about these these crazy predicted... Predictions of, like, one man unicorns, one woman unicorns. And I'm thinking, like, whatever lonely founder is trying to build a billion dollar business as themselves, they're definitely naming their agents because they want they want some friends along the way. They want some friends. Holly, **Kareem Khattab** (26:57) Cream, all these people helping me. Say this. I... I'll say this. If you're the model, you can name it whatever you want. Fable, whatever, you know, Opus. But it's just funny because, again, ultimately, like I said at the top of the show, like, we're competing against an AI pilled account who's just using clock co work. Like, that's the model. Right? Like, we need to be better than someone just using the model directly. Mhmm. And I think that's so so we're we're competing on inches, you know. That... That's like the the the level of, like, veracity and obsession you have to have. It's no longer... You cannot be, as a AI founder, like, benchmarking yourself against the old guys. Right? Because those are... Though... Like, the the legacy players in your field are not really your competition. They're they're they're kind of out of the picture at this point. It's really anthropic that you have to compete again. **Host** (27:53) Yeah. Right. And if you can have so much focus on this use case that you can beat Anthropic **Kareem Khattab** (27:59) plus the You're a unicorn. You can be... Yeah. If you can beat OpenAI and Anthropic, you're a unicorn. Yeah. I love it for sure. Well, it's This is awesome, Kareem. It's great to sit. **Host** (28:12) To you to you, congratulations. We're pumped for all that's coming. As Zach was saying, Mike, this is a... It's a really cool glimpse into another layer of what the agentic economy broadly will look like. Absolutely. **Kareem Khattab** (28:25) Thank you guys for having me. Thank you. Alright. Thanks, Kareem.