Latest / The Accounting Podcast / Is GAAP Inflating the AI Bubble?

Transcript

Is GAAP Inflating the AI Bubble? · The Accounting Podcast

• There may be errors in spelling, grammar, and accuracy in this machine-generated transcript. Blake Oliver: [00:00:04] Investors are applying the wrong lesson of the last boom. The the.com boom. They're thinking that OpenAI is Facebook or that anthropic is Google. And so far they haven't built anything that is defensible. David Leary: [00:00:24] Coming to you weekly from the OnPay Recording Studio. Blake Oliver: [00:00:30] Hello and welcome back to the Accounting Podcast, your weekly roundup of news in the profession. I'm Blake Oliver. David Leary: [00:00:36] And I'm David Leary. I'm Blake. Neither of us are in our normal studios today. Blake Oliver: [00:00:40] No, we're on the road. You're in Palm Springs for the Association of Accounting Marketing Summit. David Leary: [00:00:48] It's the. They actually, it's a m summit, but they pronounce it AME. I never knew that until I came to this conference. They call it AME. It's AME summit. Uh, but breaking news as of like 3 or 4 minutes ago. They've rebranded Blake. They just announced this. It's not going to be called Association for Accountant Marketing anymore. So drum roll it's going to be called Association for Accounting Growth A A G. And it's about elevating careers and advancing firms. So because I noticed when I was here, it wasn't just marketers. You got a lot of like the, the sales bros, guys at firms, the rainmakers are here, you know, those types of people. So it's more than just marketing accounting firms. It's people that are helping accounting firms grow ultimately. So they've, they've evolved their mission and their brand. Blake Oliver: [00:01:29] I guess this week I'm headed to New York. I'm going to Nasdaq Tower in Manhattan for the black or AI summit. And I'm hosting a panel really excited to be talking about AI and see what black is doing there. Another one of those AI tax prep companies, they came out of stealth recently, and this is their first big event, and they've got thought leaders from all over the profession coming reporters as they like. David Leary: [00:02:00] Personal personal returns. Business returns both what is there? Blake Oliver: [00:02:04] They're doing both. They started with the ten 40s like everybody. And I'm really excited to see like what they're doing and hear the discussion from the execs. Accounting Today is going to be there. Apparently a Wall Street Journal is going to be there as well. So I'm excited to go and host a panel. And my panel is all about, uh, how firms are implementing AI today. It's the how to part. Uh, and Jim Bork from Withum is going to be on the panel and he's one of the top guys, uh, in the profession who's actually leading AI inside of a big firm. So I'm really excited to talk to him. We'll talk more about AI today. There's a lot of news. Elon Musk lost his lawsuit against OpenAI on a technicality, clearing the way for an open AI IPO, perhaps this year or next. Uh, I have thoughts about this? Are we in a bubble? And I think we are finally. David Leary: [00:03:00] Caught up to my thinking. You finally caught up to my thinking. Good. Blake Oliver: [00:03:03] So? Well, just because it looks like a bubble doesn't mean it is one. But I think there's an accounting tie in to this. And there's an accounting reason. You have the. David Leary: [00:03:11] You have the numbers to prove it. Blake Oliver: [00:03:12] Now. That's right. But first, David, let's thank our sponsors who are the sponsors for this episode. David Leary: [00:03:18] Sponsors. This week we have digits. We have our cost seg on pay and maxima. Blake Oliver: [00:03:23] Let's be honest, accounting software hasn't changed much in decades. The prices keep going up, but the software still expects you to do all the work. Digits is different. Digits is the world's first AI native general ledger, with built in AI agents trained on your firm's standards across every client. In your book, they code transactions, prepare schedules, reconcile accounts, run quality checks, and even chase clients for open items. So your team moves out of prep and into review advisory and the work that actually grows the firm. And because everything runs inside one platform your ledger, reconciliation schedules, reporting, bill pay, client collaboration. There's no more stitching six tools together just to close one client's books. Firms on digits are reporting 70% gains in workflow efficiency, shipping annual cleanups in days instead of weeks, and running monthly bookkeeping in 1 to 3 hours per client accounting software that actually works for your firm. That's digits. To see why hundreds of firms are making the switch to digits. Head over to The Accounting Podcast dot promo digits. That is accounting podcast dot promo forward slash DIGITS. And welcome to our live stream viewers. Boring accountant is here with five coffees today. I'm caffeine caffeine caffeinating as well. I'm on number two R cola. Great to see you as well and sing one is called has joined us as well. I think you're new to the YouTube live stream, so welcome. David Leary: [00:04:55] And I was just about to say we should thank our other sponsor for making this episode possible. They T-Mobile because I was on my hotspot, but somehow it's not on hotspot anymore. It's back on the conference Wi-Fi, so fingers crossed that we don't get a disconnection here. I don't know how it like it's shift, you know, it just grabs whatever Wi-Fi I can find, apparently. So hopefully we don't lose a connection today. Blake Oliver: [00:05:14] So I mentioned at the top of the show that Elon Musk lost his lawsuit against OpenAI. He was suing OpenAI to prevent them from converting to a for profit structure. Uh, the jury decided that Musk waited too long to file his lawsuit because he knew about the shift as far back as 2017, but didn't file the lawsuit until 2024. Past the statute of limitations, OpenAI argued in the in the in the case that Musk once wanted 90% of the company equity for himself and only turned against them after they refused to give him control? This verdict clears the way for an open AI IPO, which they wouldn't have been able to do if Musk had managed to prevent this shift to a for profit status. Company recently raised 122 billion, which is the largest funding round in Silicon Valley history. And so if and when they do IPO, it will be enormous. And this is one of those things that's making me start to think about whether or not we are in an AI bubble. David Leary: [00:06:20] Before you go to the bubble conversation, I have a question to a story. A long time ago, you kind of covered this when they decided they wanted to make the move from a nonprofit to a profit corporation, and ultimately, they have been paying taxes right up to that point. And I remember you kind of like you created a whole PowerPoint slide, like diagramed it all out. And it's part of these new court proceedings. Is there anything addressing like the back taxes they're going to have to pay when they. Blake Oliver: [00:06:45] Well, I mean, they've been losing money, right? They've been burning cash. So there's no profit. David Leary: [00:06:49] I guess. Blake Oliver: [00:06:50] Okay, I guess so. Shifting to this structure will allow them to, um, I mean, like there weren't any profits when they were a nonprofit to, to deal with the issue is the ownership. So the nonprofit is getting a huge chunk of the for profit company, but not all of it. So basically it's give a big chunk of the equity to the nonprofit and the rest goes to the investors. Uh, and there's going to be so much money if they do IPO that nobody cares if the nonprofit gets like hundreds of billions of dollars or whatever it ends up being. And that's how they worked around this. Um, so let's talk about this bubble. Let's talk about whether or not we're in one, just because valuations are soaring and IPOs are forthcoming. And there's this new technology doesn't mean we're necessarily in one. But there's something about the accounting that makes me think we are Because, uh, let's just think about how this is working, right? Companies like OpenAI and anthropic raising all this money, where is it coming from? A lot of this money is coming from big tech, Amazon, Microsoft, Amazon, in particular, investing in anthropic, Microsoft investing huge sums in OpenAI. And often they're not actually giving these companies cash. What they're giving them is cloud compute credits, because what is the biggest expense of these MLM companies? It's cloud compute, compute to train the models and then compute to run the models for their customers. Openai has an 852 billion valuation with an annualized run rate of about 25 billion. But they are burning about 17,000,000,000 in 2026. That's the estimate. So they're burning cash and it's far in excess of their revenue. And they've got this huge valuation. And you know that's a multiple of like 3435 times annualized revenue. By comparison, a mature software as a service company will trade for multiples of 5 to 10 times annual recurring revenue. So even if you believe that open AI can can generate revenue to justify its valuation, you're, you're you're betting a lot, right? This is a very highly valued company based on their recurring revenue. David Leary: [00:09:33] And some of the argument is the recurring revenue is really low right now because they're kind of subsidizing. It's like $6 Uber rides. Remember, you're all taking those. It was very money was cheap and it was getting subsidized like, like. So that's people think eventually they're going to have to raise prices and then the revenue will come up just from that naturally. Blake Oliver: [00:09:49] Right. And these companies are competing with each other on the price at the moment, because if they did raise prices, companies might switch. And like one of our livestream viewers just said, it sounds like a utility, Christopher said, it sounds like a utility. And I'm starting to think that as well. So there's two things we need to discuss here, right? One is what is the moat? What is the potential monopoly that is like defensible for these companies like OpenAI, anthropic? I don't really think there is one right now. And so that is worrisome because in order to justify 35 times IRR valuations, there has to be something that is unique that these companies can sell, that they can charge high prices for. Ultimately, that will then allow them to generate margins, to create profits that justify these really high valuations. I'm not sure there is one. And then there's also just what's going on in the stock market right now. And that's due to round trip accounting. So what is round trip accounting. Pretty simple right. I'll give you an example. So let's say you have two companies Titan Corp and Brain AI. So Titan Corp is your Google or your Amazon. Right. And it's this massive established company that sells subscriptions, write software, that sort of thing, right? Like Amazon's cloud compute and Prime. And then you've got, you know, Microsoft with office 365, right? They're well established, generate great profits. David Leary: [00:11:28] Legacy players. Yeah. Blake Oliver: [00:11:29] Legacy players. Right. And they've got lots of cash because they're just they're just they've got the money tree, right. David Leary: [00:11:34] They have real businesses. Blake Oliver: [00:11:35] Yes. And they have they have moats, right. Like Microsoft and Windows. Right. That is a ecosystem that companies buy into and they do not switch easily. And same thing with like AWS, right? Amazon's cloud compute service. These are things that are hard for competitors to replicate. They are natural monopolies. And therefore these companies are are cash cows. So Titan Corp in our example has all this cash. And they want to get in on AI. And so they invest $1 billion into this startup called brain AI, which is making large language models, but they don't invest cash. They give brain AI, cloud compute credits in the form of $1 billion worth, and they get 10% in brain AI as a result, which that would implicitly value brain AI at $10 billion. So then brain AI happily burns through these cloud compute credits to train its models. And under GAAP revenue recognition rules, as brain AI consumes those server hours, uses up those credits. Titan Corp is allowed to log that as revenue. So Titan Corp gets $1 billion of revenue from brain AI as a customer, $1 billion in cloud revenue. Right? Titan. David Leary: [00:12:58] No, no. No cash is moved. This is no cash barter transaction. They they they got equity essentially. It's all they've got. Blake Oliver: [00:13:05] So Titan Corp stock jumps because they have a boost in revenue. Investors love seeing its cloud business grow. Then brain AI develops this new model and gets all these customers. And even though brain AI is just burning through these credits and the revenue is far lower than its costs, its expenses. Right. It's burning cash billions and billions of dollars a year. Investors want to jump in because they see this massive adoption. I mean, we've got half of us small businesses now using AI subscriptions. This growth has been incredible. So there's lots of money coming in, but it's far less than what they're spending. So VC comes in, venture capitalists come in and they invest more cash into brain AI, which values it in a new round at, say, $50 billion. And this is where the gap magic happens, right? Because under a SC321 public companies holding shares in private companies must adjust the value of their investment whenever there's an observable price change, such as a new funding round. So Titan Corp, they have a 10% stake in brain AI that's worth $1 billion. Because of the new VC round, that 10% stake is suddenly worth $5 billion on paper, and GAAP dictates that Titan Corp must mark up the value of its investment by $4 billion because it's now worth five. And they paid a billion, but they didn't pay cash, remember? They just gave credits. Yeah. So Titan Corp gets to record this 4 billion markup directly into its quarterly income statement under other income. So what has happened here? No cash has exchanged hands. But at the end of this cycle, let's say within that year, even Titan Corp gets a billion in top line cloud revenue added to its income statement and 4 billion in bottom line net profit in other income. Blake Oliver: [00:15:02] So Titan Corp has manufactured a billion in revenue and 4 billion in profits out of thin air. No customer walked in from the outside world to buy a product. Titan Corp can't use its 4 billion in profit to pay its employees or give investors dividends because it's all paper money, right? It's all paper wealth, and it's tied up in brain AI stock. So if brain AI doesn't go public, that 4 billion profit is going to be reversed in a massive write down. And that's when the bubble pops. And these are not small numbers. In its Q3 2025 earnings, alphabet, the parent company of Google, they reported a over $10 billion gain net gain on equity securities heavily tied to anthropic, according to people familiar with the matter. Amazon same period Q3 2025 reported a 9.5 billion pretax gain under non-operating income, also driven by the paper valuation of its anthropic stake in recent filings this month. In May, Microsoft disclosed a nearly $6 billion net gain over a nine month period, largely originating from a dilution gain tied to its OpenAI OpenAI's recapitalization and restructuring efforts. So you have these public companies reporting these massive gains Now smart investors can see that this is not real, that this is paper profits. But here's the thing about our stock market that's changed over the last few decades, more and more money is now being passively invested. The money in your 401 K, the money in your investment account. If you put that into index funds, which have become really popular, those index funds are simply designed to mirror the stock market as a whole. So when companies report these profits, the the, the passive investing money doesn't look to see, oh, is this other income or is this operating. David Leary: [00:17:20] Income profit number. Blake Oliver: [00:17:21] They see the profit number. So so what happens is that profit number bumps the stock price. That increases the capitalization of the company, which increases its size in proportion to the entire stock market. And the index funds then have to buy automatically those shares in order to stay aligned with the size of the companies, right? The bigger a company gets, the more the index fund needs to hold. And the big tech companies are now something like 30% of the stock market. So you get this feedback loop where. David Leary: [00:17:56] I think I saw something recently, like all the gains in the stock market, like the last 12 months have just been AI related. Yeah. Yes. This, this bubble or loop or whatever you want to call it. Blake Oliver: [00:18:05] So the passive investing trend, which has gotten bigger and bigger since people realized that it's hard to beat the market these days, that's reinforcing this trend. And so here we are all along for the ride. A lot of people's 401 s along for the ride and well, will the bubble burst. And that takes us to the other question, which is what is it that these companies have? What does OpenAI have? What does anthropic have that is truly defensible? In other words, what is their moat? What do they have that prevents anyone else, any other competitor, from coming in and selling a similar product for a lower price and compressing those margins, those margins that they need in order to justify these high valuations? And I'm not sure they have anything because what is their core product? It's these large language models. And yes, they cost a lot of money to make initially, but the cost of creating an LLM has gone down, and we've seen companies come out with open source models like deep seek was the big one that scared everyone. David Leary: [00:19:21] Chinese knockoffs. Right. Blake Oliver: [00:19:23] And Exactly. It's it's it's like the same thing in manufacturing or retail, right? You can create a copy and you can do it more cheaply as the technology improves. And so the LMS themselves are not a moat. Anyone can make one and and sell it for cheaper. It's a utility. And really the cost of the LM to the customer is going to end up being somewhat related, or maybe even directly to the cost of electricity and compute, and we'll get cheaper and cheaper. So like there's, there's no, there's no moat for open AI when it comes to its models, and there's no moat for anthropic when it comes to its models. David Leary: [00:20:06] So it's the moat like, and based on some other news stories I have this week is the moat. Essentially, we just want you in the habit of using us. So you just switching just a headache you don't want to deal with. So you're just honest forever. Like get you to do everything in your life on us. Correct. Their hope. Blake Oliver: [00:20:23] But but the switching costs are actually pretty low because if you are smart as a business, you design your AI workflow so that you can simply swap a model in and out whenever it becomes better or cheaper and for whatever your needs are. And that's exactly what we did at earmark. We built our AI course creation workflows in Zapier, and Zapier is great because it allows us to simply choose what model we want to use. So when Claude started getting better than ChatGPT, I was able to go in and simply swap the model that we're using and the workflow stays entirely the same. So there is one way that these companies can create a moat and make switching costs higher, which is to create the workflow layer and get companies to adopt that. And that's what we're seeing now with Claude and Co-work. Claude is just growing so fast and even overtaken ChatGPT now in terms of business adoption, because Co-work is a great workflow tool, and it allows you to organize your files into projects and share projects with your coworkers and, and to save skills and that sort of thing. But it has nothing to do with the underlying model. That's not where the value is. The value is in the workflow, but that's also not a great moat because anyone can create good workflow software. And so you look at a company like notion. Notion is this really powerful, um, database. Uh, it's almost like a super customizable database and workflow app for small businesses. And they've built AI agents now into notion and you can plug in ChatGPT or clod, or you can allow notion to choose Gemini whatever model is best for the task. And so I think those kind of apps, the workflow apps are also going to proliferate. Blake Oliver: [00:22:37] And if you're smart when you're designing, when you're architecting AI systems for a business, you'll choose something that doesn't lock you into a particular model provider, and you'll be able to just choose whatever it is you want. So, you know, I, I don't, I don't see the moat here. I don't see what what is open AI have that's special that will allow them to create this kind of monopoly. And, and if you're not sure what I mean by this, think about what happened in the last tech boom. Think about Facebook. Think about Google. Think about let's talk about Uber. They all have moats. And when it came to social media, the moat for Facebook was they managed to get everybody on Facebook. And then it creates this network effect where, well, everybody's on Facebook. My friends, my family, they're all on Facebook. So I'm going to be on Facebook too. And nobody wants to go switch to a small, tiny social media platform. And so we ended up with a handful, right? We ended up with Facebook and Twitter and LinkedIn, and then Instagram came along and Facebook bought them. Oh, and TikTok. Right? Like, so basically a handful. And so there's a few winners. Same. Google is a great example of a natural monopoly because once everyone started searching on Google, um, you know, with their PageRank algorithm, like that just became the place where everyone went. And so every, every advertiser also had to be on Google. And very few people have any reason to switch to another search provider because none of the other providers were any better. Um, and so you have to think like, what is the monopoly? What is the reason these companies are going to be sticky? And I don't see it. David Leary: [00:24:29] So I could dig into some articles of that were announcements that were made this week that I think what I think their play is going to be, if you want to talk about that, we have two plays they came out with. One is personal finances. So ChatGPT has announced they're going to add personal finances. Blake Oliver: [00:24:47] So so David, I want to hear about this. But first let's thank our next two sponsors. All right. So our next sponsor is rhe cost seg. If you have real estate clients, you already know cost segregation can save them serious money. But actually delivering studies in-house is a headache you don't need. And sending clients to an outside firm means losing control of the relationship. That's where rhe cost seg comes in. They're the partner CPAs trust to deliver engineer backed cost segregation studies start to finish. You keep the client relationship, they bring the engineers and the deliverables. Your client saves 6 to 7 figures and you're the hero. They've completed over 15,000 studies and the pricing is actually reasonable. We're talking $1,000 for rapid reports and 3000 for fully engineered studies. That's way below the typical five K minimums you'll see elsewhere. And they never compete with your firm with the One big beautiful bill act reinstating permanent 100% bonus depreciation cost. Seg studies are more valuable than they've ever been. To schedule a strategy call with Ari Keszeg and get a special bonus perk by mentioning the Cloudaccountingpodcast.com. Head over to The Accounting Podcast dot com slash Ari cost seg. That's accounting podcast.com/rcostseg. And our next sponsor is on pay. Blake Oliver: [00:26:10] Are you tired of payroll headaches getting in the way of the client experience you want to deliver. Manual workflows creating bottlenecks, compliance, nightmares, and endless support calls that go nowhere. There's a better way for your team and your clients on pay. Is the payroll partner that accountants and bookkeepers actually love? Why? Because it's easy to use, packed with value, and backed by support that actually supports you. Their team gets rave reviews for being fast, expert, and actually reachable when you need them on pay. Handles the heavy lifting. You get a dedicated onboarding coordinator who sets up worker profiles and transfers year to date data from previous providers, all at no extra cost. Their seamless QuickBooks and Xero integrations eliminate manual journal entries, and they support any type of businesses you serve farms, restaurants, nonprofits, you name it. On pay can handle unique requirements without adding complexity. And on pay keeps pricing simple, too. Everything your clients expect from multi-state filing to off cycle pay runs is included. No hidden fees and no surprises. To book a demo. Head over to The Accounting Podcast dot promo slash on pay. That's accounting podcast dot promo forward slash OPAY. All right. David. Apologies for interrupting you. David Leary: [00:27:20] Yeah. No. So yeah, two two, two moves. I think they're making to try to not so much create the moat, but um, I think let me explain what the two moves were. So one move they did is you now can sync your personal bank feeds to ChatGPT. So ChatGPT as of last Friday, their pro customers can now use Plaid. So a lot of us have experienced with Plaid because you connect your bank feeds through the accounting GL, some of them use Plaid. Other apps use Plaid to connect the bank feeds, and it's going to give ChatGPT read only access to your balances, transactions, investments, subscriptions, and upcoming bills. And essentially, it's making ChatGPT now become your AI financial advisor so you can. How much did I spend on shopping? How much I spend on vacation? You can. All the stuff. If you think like all the things you would use quicken for back in the day for personal finances. It's just going to be built in the ChatGPT. And that because they partnered with Plaid. I mean, that's giving them access to 12,000 banks, financial institutions. And so I have kind of two thoughts on this. One is, I think Plaid had no choice but to partner with them on this because the risk of ChatGPT or Claude just building their own knockoff bank scraper is probably very high. David Leary: [00:28:29] Like this. Is this like these these companies have so much money and they're so aggressive that you have no choice but to partner with them because there's a chance they could eliminate you if you don't. So. Plaid. Plaid Plaid to do this. So now you know, all your bank information can totally be accessed in there. But it just makes me think about the goal and vision, right? And this ties back to your moat is the vision of all these top AI plays is to be like an Iron Man Jarvis type bot that you just talked to, or like a C three, C three Po like, but digital butler that you just talked to and they handle everything for you. So or if the GPT, the chat interface or voice interface becomes the UI for everything. There's no more register. There's no more UI for QuickBooks or UI for quicken. You just talk and it tells you about your bank. How much I spend on coffees last month. Well, it connects to your bank feed and reports how much you spent on coffees. Is that kind of the long term vision of all these? They want you to just do everything in the one platform, and that's their moat. You just don't leave because you're too invested in that one platform. Blake Oliver: [00:29:30] It's the best they've got, if you ask me. And that's kind of where we're at with Claude. It was easy for us to switch from ChatGPT to Claude, because ChatGPT wasn't really connected to anything, but Claude now connects to all of our Google Drive and to our emails and to notion and our CRM, like everything. But at the same time, I could just go and connect another model to all of those tools, and the model could go and search and find and get everything. So if you ask me, the the best thing for them is the agent layer. So if you've been paying attention to how people are using Agentic AI to build automated systems, there's these apps like Nate N Zapier has an agent builder. And these are, these are layers. David Leary: [00:30:35] Like an orchestration of your agents, right? Blake Oliver: [00:30:37] Exactly. And so once you build a business on one of those apps, you're not going to want to switch it out. It's sort of like once you get on Google Workspace or on Microsoft 365, the switching cost is really, really high. David Leary: [00:30:51] Yeah. You never switch off a QuickBooks if you're on it, it's too much work. Blake Oliver: [00:30:54] So that's what, um, Claude is or anthropic is really investing in. Um, similar apps are like Microsoft's copilot studio, Azure AI studio. Um, and what these do is they connect the models with memory, with files, with saved prompts, how you want it to do things. And then they prompt the models to get the results and then pass them around and orchestrate all these multiple different little agents doing the work. So it's still though, it's still not like patented, right? It's, there's no, there's nothing that prevents anyone else from building something like this, but it's definitely better than just the LLMs, which are like going to be like a utility. David Leary: [00:31:44] I mean, one thing that's an interesting number with this personal finance announcement is, is the sheer volume of queries they get actually their moat. So they know what to build. They have a competitive advantage because they know what to build. According to OpenAI, they say they get 200 million plus users a the month as financial personal financial questions in ChatGPT. So they have. So they know what people are using the Gpts for. So they have the competitive advantage to make a move and beat everybody else to market. On what a solution should exist in the GPT. And it's just if if 200 million people are asking you about personal finance questions, how many million are asking about some health issue? I got this dot on my skin. What is this? Right? Like they're going to be able to beat every industry to the punch because they know what people are lack of word, searching for or chatting about. Google basically took your search and that was a competitive advantage for them to turn that into ads. What are people searching with Blake searching for? Hockey skates, hockey ads. Right? They know what this. But this is different. This is you're asking problems that have to be solved. And now they can go solve these problems. And that's what they did. We have 200 million users a month asking about personal finances. Let's build a personal finance app. And now you can talk about Claude Clouds basically just done the same thing. I don't know if you saw that Claude launched their small business portfolio, if you want to call it that. Blake Oliver: [00:33:06] Yeah. So it's like, um, I've seen them roll out these collections of skills. That's what they, that's what it looks like to me. It's like install this, what are they calling it? It's like a package of connectors, automations and pre-built workflows. David Leary: [00:33:22] Screenshot. And I lost it. Um. Blake Oliver: [00:33:24] But I installed this for like, uh, finance, right? You can, you can install Claude for finance finance. David Leary: [00:33:29] Exactly. Blake Oliver: [00:33:30] Inside the Claude desktop app. It brings in all these skills that have already been pre-created for you, like month end close as a skill and then you can customize them, uh, and it can help you with your workflows and that sort of thing. Is that what this is? David Leary: [00:33:44] It's kind of, so what this is, is it's the Claude for small business. So they're taking their integrations, kind of wrapping them all up under one umbrella. So your integration with QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, Microsoft 365. And then you combine that with the skills, like you said, payroll forecasting, month end, close reconciliation, invoice collection reminders. So the the play is they're trying to be the interface between all the small business apps you use and all the workflows and problems you might have in your small business. Now, cloud business is going to be your, your again, it's exactly like ChatGPT play with the personal finances. They probably know this. Many people are trying to reconcile or use small business use cases inside a cloud. And now they just built. That's their mode. They're going to be able to build because they have the knowledge of what people are trying to use AI for. That's their mode. Blake Oliver: [00:34:35] Well, I guess it's not really a mode, though. It's like competitive intelligence, but then they still have to build something that has a moat that's going to be sticky, that people are going to be willing to pay a lot of money for. That gives them great margins. And I'm just not convinced that, I mean, they have to do that. Right. And the problem right now is these valuations are based on this idea that the LLM is a moat And it's not? David Leary: [00:34:57] No. Yeah. Blake Oliver: [00:34:59] Investors are applying the wrong lesson of the last boom the the.com boom. They're thinking that OpenAI is Facebook or that anthropic is Google. And so far they haven't built anything that is defensible. Here's an example going back to this small business collection from anthropic that I thought was interesting. One of the examples of the workflows they've built is payroll planning. So you can compare your QuickBooks cash position with incoming PayPal settlements and forecast 30 days ahead. So you can basically do like a short term cashflow forecast if you connect these, uh, these connectors. The connectors are Intuit QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365. David Leary: [00:35:55] I just don't buy into this cash flow forecasting stuff. I just don't because especially for small businesses, they many of them, yes, you're going to have some stuff on a PayPal or people. Oh, you have an open invoice that you created in QuickBooks, but I guarantee you you have ten other invoices you haven't entered yet. That's actually your cash flow. And that'll never be in the reports. Like, like it's never going to be accurate to do a cash flow future thing because I think some of it's back of napkin at the small business level, they're never going to have the data. Blake Oliver: [00:36:24] Well, if they can somehow get the information about those bills that you need, like probably in your email. David Leary: [00:36:30] Yeah, that's true. Blake Oliver: [00:36:31] Or in your file storage. And so it's all about these tools, having the context and knowing where to go and get all this information. But if, if they do have the context, it's really powerful. Um. David Leary: [00:36:43] Yeah, because we could check the emails and be like, hey, these deals are close to being a yes. Let's put that in the cash flow forecast, right? These deals that are about to land, which historically would not be in the accounting GR or B anywhere. Blake Oliver: [00:36:56] And, you know, once it has that like the, the ability to build these spreadsheets is just spectacular. Like, um, I just want to share an example of something I did recently. I, uh, I wanted to, I wanted to do a workbook to figure out if I've been, you know, paying enough in estimated taxes and payroll taxes and, you know, a few different income streams. Income has increased a lot over last year. I'm not sure if my withholdings need to change on my S Corp payroll. I'm not sure what I need to do about estimated taxes. And so I just started a project in cloud work called Withholding Analysis. And I dumped in my year to date payroll register and my last year's tax return and w-2s and, you know, gave it all the information I'd give to like my tax accountant. And I said, hey, help me figure out if I've got enough withholding if I need to make some estimated payments, what I need to do. And it created this entire workbook showing. Here's where you're at and here's based on your income growth, where I expect you to be. And did it like a few different scenarios and calculated my year end tax liability, calculated the safe harbor amount based on my tax return. And then compared that to the projected liability and said, all right, if you want to meet the safe harbor, here's the bonus payroll that you should run at the end of the year. If you want to completely clear out your liability, here's what you should do. Gave me recommendations. I went back and I said, hey, you know, I want to run a bonus payroll now just to like get some of this paid, right? Catch up for what I owe at this point. And it gave me the numbers to enter into on pay for my zero net check payroll. And then I gave it the pay stub after I did that and it updated the analysis and it said, okay, now, you know, check in with me in a few months and we'll do another bonus payroll, that sort of thing. David Leary: [00:39:02] And that would have been five emails and possibly a Zoom call with a CPA. Blake Oliver: [00:39:08] Yeah. And I could have done it myself, right? I could have figured out this all myself, but I was able to do it in like, I don't know, 15 minutes. And then when I went into on pay, you know, and I couldn't figure out where to change to manually override the withholdings, I just opened up Gemini in a sidebar in Chrome and I asked it, where can I do the bonus payroll off cycle? And then when I couldn't find where to change the withholdings, it pointed me to a little tab that I was missing. David Leary: [00:39:34] And this is, I think, where the why the investors are so hyped about this because even though they may not have a competitive moat against another AI company from doing the exact same thing you just did, but they're just going to steal that work from an accountant, they're going to basically roll this out as a skill. It'll be part of the ChatGPT personal finances or Claude personal finances, whatever you want to call it. And this, that'll just be something common in out of the model, out of the box that people just get. And now they'll never have to ever ask an accountant that for that help ever again for the rest of history. Blake Oliver: [00:40:07] Well, but the thing is, like, the only reason I was able to, you know, do this confidently is because I know enough to check the work. Oh, sure. Let's look at this for five minutes. David Leary: [00:40:16] Let's wait five. Blake Oliver: [00:40:17] Minutes. David Leary: [00:40:17] The numbers waiting five months. You know, that's the problem, right? Blake Oliver: [00:40:21] But also like the, the AI, like can't confidently do everything or see the whole big picture. And so it's not like I'm going to fire my tax CPA because I can do this one thing now, or I can at least gut check it. And a lot of people, I mean, it's still effort to do this and like to open up that workbook and understand what the AI had done and, you know, check the numbers that it had put in and like gut check this, you know, like it created a table of, uh, of, with the latest, you know, with, uh, tax rates, right. And it like calculated my tax liability based on my projected income. I mean, there's a lot of things that could go wrong. David Leary: [00:41:03] And the thing that they're all missing that they have no clue about because of tech hubris, what they don't. There's a reason Intuit sold quicken. And the root root cause is personal. Finances is a hobby. You could argue every single person should do them. Every citizen of the United States should do, but only like 12 to 16 million people do their personal finances in any meaningful way. And quicken didn't grow for 30 years and never grow new users because the hobbyists got quicken. And that was it. And that's so, so like you look at ChatGPT, they rolled out this whole product, all this announcement for basically for some personal finance hobbyists. There's no they're never going to grow that pie. There's a there's a limited amount of people that track their personal finances and have any personal responsibility for their personal finances. It's a very small percentage of the population. Blake Oliver: [00:41:55] And what are those people going to pay for an individual cloud subscription or ChatGPT? David Leary: [00:42:02] They want quicken for $29 a year. They're still using quicken desktop versions they have installed. They still have the floppy disk. Those people like they are, you know, meticulous, meticulous about every check getting typed in. They love their data. They're hobbyists. It's personal finance hobbyists is really what they think about them. Blake Oliver: [00:42:18] So David, um, let's thank our final sponsor and then let's talk a little bit about some other companies that have released new features. We've got some news from digits. They're one of the sponsors of this episode. We should cover that. Zero's got some new stuff going on, and I want to talk a little bit about what ramp is doing. David Leary: [00:42:36] I have a something from, uh, the puzzle founder, Sasha Orlov. I want to talk about Klarna in AI. And then I also want to talk about the boomerang here. Bench accounting is back. Well, kind of. We'll do the ad and then I'll talk about that. Blake Oliver: [00:42:51] All right. Our final sponsor of this episode is maxima. It's the 15th day of the month, and your accounting team is still scrambling to close the books. Sound familiar? Controllers and accountants everywhere are stuck in this cycle. They're pulling data from dozens of systems, reconciling transactions line by line and racing against deadlines just to get accurate financials. While AI is transforming every other part of business, accounting teams are still drowning in spreadsheets and manual work. That's where maxima comes in. Maxima is an Agentic platform built specifically for record to report operations. Their AI agents actually understand how accounting works. They pull data from your entire finance stack, apply your accounting policies, draft journal entries, reconcile accounts, and get everything prepped for human review. Companies like Rippling and Zendesk already use maxima to cut their clothes time by up to 80%, while automating 95% of repetitive work. Which, honestly, it's kind of hard to believe until you see it. Imagine closing your books in days, not weeks, and actually having time to analyze the numbers instead of just assembling them. That's what it looks like when AI handles the prep work and your team handles the judgment calls. To see why accounting leaders are making the switch to maxima, head over to The Accounting Podcast dot com slash maxima. That's The Accounting Podcast dot com forward slash MAX. I'm a. David Leary: [00:44:14] Yeah. So synthetic, a new startup from former bench accounting co-founder Ian Crosby, raised $10 million from Coastal Ventures, um, to build an AI powered bookkeeping platform aimed at SaaS and AI startups. Now rewind to bench accounting flopped, right? That was the, hey, we're going to have bookkeepers and then we're going to have engineers automating everything. And this is this is kind of the pre AI boom, right? And that was the business model, which we're just going to have engineers build it. Bookkeepers on the front end and the model failed right. But bench went bankrupt. They wound up they raised $113 million. They wound up going bankrupt. And they had 65 million in liabilities, 2.8 million in cash. Did the model did not work. We raised new money. Now, there's two things that's kind of interesting. This there's three, I guess. So one, he wants to keep this, uh, not launched publicly. He does not want to launch it until he believes the AI is more reliable than a human bookkeeper. So basically he's raised money just to go build this now, right now. He also said that he admits this process could take anywhere from six months to six years. Blake. So do any of us know what we're doing? Do any of these people know what they're doing? Like, like it's six months to six years. It tells me he nobody he doesn't know what he can build or not build. It's kind of ridiculous. And then the other thing that caught my eye on this, and this is you've just been your argument forever. The startup is going to focus on SaaS and AI businesses and software businesses first, and SaaS startups. And what have you always said about startups? It's the easiest accounting to do. There's no revenue accounting, payroll. Yeah. Blake Oliver: [00:45:49] And they're there. They go to business all the time. So they're not they're not great accounting clients. David Leary: [00:45:54] No. But they're easy to do bookkeeping on because there's no revenue and it's just payroll expense and that's it. Yeah. Blake Oliver: [00:46:00] You're just calculating burn and it's putting burn into like four different buckets. Yeah. It's not too complicated. David Leary: [00:46:05] It's marketing expense that you can pull from what you spend on your Google and Facebook ads. You have payroll expense and you have no revenue in many cases, these startups. So it's very easy to automate, quote unquote, that type of a business. And AI is probably really good at it. But you don't see these people like I raised $10 million to solve accounting for plumbers. Like that's the hard problems to solve. Nobody's doing that. They're, they're using AI to solve arguably the easiest accounting in the small business industry. Blake Oliver: [00:46:31] I have to say, I'm also skeptical, David, because I think all of this bookkeeping, automation, AI bookkeeper stuff is going to be done by the general ledgers themselves, by digits, by puzzle, by QuickBooks, by Xero. Perhaps if Intuit and Zero can get their act together when it comes to these AI agents. Um, Kik is another one. Uh, um, lots of them. And so like an app that just plugs into your GL again, there's no moat for that, right? Claud and OpenAI can create skills to do the month end close if they can plug in to your GL anyway. David Leary: [00:47:16] Somewhere there has to be a source of truth, right? The data. You still have to put the. In theory. Hopefully we have a world where data lives in a table and it's structured somewhere. You still have to have a GL. You still have to have a place to record the data. We I don't want I don't like this world, this point of view. Google had this point of view. Stop using folders, just save it anywhere you want and just search for the things you want to find. Yeah. And like, I don't know if I like that for accounting and business stuff. I want things to be in a field that makes sense, logical postings. Blake Oliver: [00:47:45] And that's why I also think that the SaaS apocalypse predictions are totally false. They're wrong. All those people out there that are saying that small businesses are going to build their own SaaS applications and replace QuickBooks. It's insane. And the reason is. David Leary: [00:48:06] That they're going to try. Blake Oliver: [00:48:08] Sure. But there's so many problems with it, right? Yes, you can clone QuickBooks. You could clone a GL. Debits and credits are not that complicated for an AI to manage. Like you can, you can go into cloud Co-work right now and ask it to build you a double entry accounting system, and it will do it for the database. David Leary: [00:48:28] You could do this before. Blake Oliver: [00:48:29] We're done it in five minutes, right? And that's not the value of QuickBooks. And there was a recent podcast interview, uh, CFO brew did an interview with, um, Alex. Um, is it, how do you say his last name? Balazs Ballas. He's a CEO of Intuit. And it was about how investors are predicting a SaaS apocalypse. And I've been punishing Intuit in Zero for this. But this interview was about into it. And you know, the CTO makes the correct argument that AI can commoditize user interfaces and workflows. They can copy apps, but the data storage that and the trust in the data is not something that they can easily commoditize. So if like you built your own accounting system, David, and you bring it to me as your CPA. I don't know what this thing is. Or you bring it to the bank and you say, I'm producing financial statements out of my own home brewed accounting system myself. David Leary: [00:49:39] Yes. Blake Oliver: [00:49:40] Yeah. Like no chance. Right. So lenders, tax pros. Um, anyone else who needs to like use this system, they're going to want your data in a system that's trusted and, and, and audited or auditable. Right. And not in something that you made yourself. Uh, and then there's all just the connections, right? Yeah. You can build a copy of QuickBooks desktop, but how do you connect it to your bank? How do you connect it to your payroll? David Leary: [00:50:06] Built in now to the, it's like, that's the thing. Like, like these little, all these pieces are starting to show up. But for you to unify that and build it all together and get all to work is work. Like. Yeah. And don't you have a small business? Blake Oliver: [00:50:18] Yeah, exactly. No small business is going to say, oh yeah, it's worth building my own software to save 100 bucks a month on accounting software. That doesn't make sense. And any app that tries to go do that is not going to be able to survive because you're building a like a QuickBooks competitor. You know, like the marginal cost is not enough to get people to switch. David Leary: [00:50:41] I have a graph you can bring up if you want to bring up the thing. So not all jobs actually make sense to be represented by AI. So this was from a LinkedIn post from Sasha Orloff. He's the puzzle founder and he covered basically we covered this a while back. Klarna. Klarna is one of those buy now, pay later companies. And they made, they they got rid of like 700 customer service reps back in the day. And they said, we're going to have AI do all our customer service. And what they've done recently, they've started rehiring people because they've discovered that quality suffered. It's the cost for customer service versus the tokens just doesn't make sense. So this graph you pulled up essentially, and this was published on the A16 newsletter. If you compare the cost of like the AI tokens to a human. So a data entry agent, a human, you might pay $80 an hour, right? Or $80, $80 per day. And the agent, an AI agents maybe like 60 bucks a day human for a call center agent is like 90 bucks a day. Ai is probably going to cost you 92. So it exceeds the cost. But where it really makes the most sense is code engineers, right? A human engineer is like $300 per day, but the AI coding agents like 14 bucks a day. David Leary: [00:51:51] Yes. And, and I really think that makes the most sense because if you think about the call center agent data center agents, they're I wouldn't say winging it, but they are they're just making up the answers every time and starting over every time. But the coding agent will code something you can use day after day after day. It's reusable code. It's a reusable solution. Yes. And it makes so much more sense. So not all things you're going to are going to replace. And he ties this back to like with accounting. So if you think about accounting is more like customer service. There's a high quality bar. You can't have mistakes because that means reinstatements, audit failures, compliance risks, etc.. And so we have to think about like this, this whole like AI is going to replace accounting isn't going to be just, oh, it's the data entry, right? There's just too many other risks involved. It just doesn't make sense. Right? To straight up, we're going to have agents replace accountants and bookkeepers and data entry people. It's just it's expensive work. Blake Oliver: [00:52:45] Remember, these costs right now are based on highly subsidized AI tokens. So OpenAI and anthropic are not going to be able to burn through tokens. When they start running out of money, they're going to have to generate profits. And so the price per token is going to go up. David Leary: [00:53:03] It's going to go up 5 to 10 x. I think honestly. Blake Oliver: [00:53:05] Like, I mean, yeah, if we look at like, I don't know, Uber at the beginning, right, you could get a ride for under $10. Now granted, that was like 20 years ago. You can't go. David Leary: [00:53:14] Six miles for under 40 bucks now in an Uber. Blake Oliver: [00:53:17] It's crazy. Yeah. Let's say with inflation, you know, maybe it was like inflation adjusted. I'm just ballparking it like 15 bucks for a ride that's now 30 bucks. So the prices have basically doubled. Once everybody started to do this thing. And so I think we'll see like token prices, they have to like AI prices have to go up. It's just incredibly cheap right now. I mean, you know, our entire company is spending like a few hundred dollars on tokens at this point per month. But we would happily spend more. Shh. David Leary: [00:53:46] But. Blake Oliver: [00:53:47] Well, but here's the thing. David Leary: [00:53:47] Is. Blake Oliver: [00:53:48] That here's this this goes back to the whole moat, right? It's like if one of the providers raises their price, then we'll just switch to a different model, or we'll switch to a different company. David Leary: [00:53:58] Until they all raise prices universally because they're going to be like, well, electricity's the same everywhere. They're going to be able to they're going to be in cahoots together. Yeah. Blake Oliver: [00:54:07] Anyway, David, I have to go. This was so much fun. Thank you. Everyone who joined us live. Great to see you. Boring accountant and Fluffy Hammer and Chris Karn and can't pay attention. And more and more that I can't thank. Great to have you with us on the live stream. Sorry if we couldn't get to your comments today. We had a lot to go through. Um, we didn't get to the news about like digits, new features. We'll talk about that next time, zero and black or I'll be reporting from the black or event. I'll have some takeaways from you for you. We got some news from ramp with AI agents for procurement, so we'll save that for next time. And don't forget, you can earn free continuing professional education for listening to this episode of the Accounting Podcast and many other fine accounting, tax and audit podcasts. Get the free earmark app. Go to the App Store, download it, sign up for free, earn one free CPE per week or go to earmark.app in your web browser. It's the easiest way to earn CPE. We have now issued over 200 000 CPE certificates that are usable by CPAs, SCMAS, enrolled agents and even chartered accountants around the world. If your jurisdiction accepts Nasba CPE credits, they will work for you. David Leary: [00:55:28] And because if you notice this episode, not one story was not AI related, we're probably going to get some bad one star reviews. So if you liked our coverage, please go leave to go to your app. The, uh, not Google Play, the Apple Store. I always say Apple App Store, the Apple podcast. Blake Oliver: [00:55:43] Podcast. David Leary: [00:55:43] App, leave us, leave us a five star review. If you're getting value from listening to this show, like we, we're not obsessed with AI. It's just every news article is about AI in this industry right now. You can't get around it. Blake Oliver: [00:55:54] I just realized I had a story in the title of this episode, PCO be considering cuts and I totally didn't get to that. So I guess next time we'll we'll lead with what's going on with PCO and SEC. David Leary: [00:56:06] All right. I'll let you go. Blake Oliver: [00:56:07] Have a great time at the conference and I'll see you around here next week. Bye, everyone.