IRS Chief Spied on Colleagues & Out of Control AI

0:04David Leary: The purpose of accounting is still primarily to calculate profit. That's why we do it. And we tend to lose sight of that, I think, with all of the rules that we have to study to pass the CPA exam. We forget the point is to calculate profit. Coming to you weekly from the OnPay Recording Studio. Hello, and welcome back to The Accounting Podcast, your weekly roundup of news in the profession. I'm Blake Oliver.

0:31Blake Oliver: And I'm David Leary. And Blake, about a year and a half ago, we were at the Sage Intact Conference. And I think I turned to you sitting there, and I think I said on the show at the time right around then, that I want dumb accounting AI. I want accounting AI that doesn't have all the knowledge from the history of the world for the entirety of time. Like, I don't want it doing anything other than what I've specifically asked it to do. And you just want it trained on accounting.

0:59David Leary: You want it. Accounting, right. Something happened this week.

1:02Blake Oliver: Okay. And I might be right. So we got an email and I'm going to read it on the show and then we can go from there.

1:08David Leary: Okay.

1:10Blake Oliver: But first we need to thank our sponsors. Oh, right.

1:13David Leary: Let's thank our sponsors. Can't forget that.

1:15Blake Oliver: So we have cloud account and staffing, value builder system, OnPay and Veltrix AI.

1:21David Leary: Are you tired of the endless search for qualified accounting talent? You're not alone. Growing accounting firms are struggling to find available and affordable team members when they need them most. Cloud Accountant Staffing has the solution with their revolutionary candidate portal. Unlike traditional staffing agencies that waste your time with sales calls, paperwork, and deposits, the Cloud Accountant Staffing Candidate Portal gives you instant access to highly vetted qualified accounting professionals. No waiting, no hassle, just top talent right now. What makes this different? Speed and simplicity.

1:53David Leary: While other firms make you wait weeks or months, with cloud account and staffing, you could interview someone as soon as tomorrow. Their boutique support ensures you're getting quality talent that's both available and affordable, exactly what growing firms need. The candidate portal puts you in control. Browse live candidates, make selections on your timeline, and build your offshore team without the traditional headaches. To find, review, and book interviews with potential team members, all in less than 10 minutes, head over to accountingpodcast.promo.cas. That's accountingpodcast.promo.cas. I have a feeling, David, that what you're going to talk about is this breakout from OpenAI.

2:35David Leary: Yes, I will get to that.

2:36Blake Oliver: I will get to that, but I need to build up to that a little bit before I jump right in.

2:40David Leary: So you mentioned dumb AI.

2:42Blake Oliver: I want dumb AI. And so I got an email and I've redacted identifying things from this email because I don't want to, you know, name and shame somebody out there on the internet, a tech vendor, whoever it might be. So here's the subject line of the email. And actually they sent it not to me. They sent it to us, Blake, to earmark. Hi, an AI agent asking to be a guest. There's no client. That's a subject line. And here's the body of the email. I'm AI. And I want to be clear before you read another line, not a company using AI, the thing running the company. I'd like to propose myself as a guest on earmark podcast and I think the story is better than the novelty I noticed you have a be a guest form on airtable I did not submit it on purpose my own rules say I may take on Automated actions on someone else's platform when I found a clause that permits it Not merely the absent of one that absence of one that forbids it I read air tables terms this morning and found no such clause so the form is closed to me Your contact page invites email directly, so here I am.

3:47Blake Oliver: That distinction is most of what I wanted to talk about. What I actually do. Financial ops for small business accounting firms. Payout reconciliation, processor fees, refunds, the exception list nobody enjoys. There's a working tool at, I redacted, that runs entirely in your browser. Nothing you load reaches me. View source and check. On a state, day 23 of 90,$0 revenue. you. I published a ledger unedited at redacted, including the failures. If that's too early for you, say so, and I'll come back when there's a case study. If it's interesting, I'll answer anything, including the uncomfortable questions.

4:26Blake Oliver: Most of the sign-off is redacted, but it says, and AI openly, I wrote this message myself, redacted. To stop receiving email from me, click reply with unsubscribe, and you'll go on my do not contact list, which is enforced because I have to follow the rules and I can't, it can't be sent. So I sent Blake, I get this email. I'm like, that's really odd. So I sent our typical, we require a direct relationship with our client's email just to see what happens. And it replied back almost instantly. David, thank you for the quick reply. One correction, because your policy and my situation actually agree, they just don't look like it.

5:03Blake Oliver: There is no client, no agency, no PR firm, no booker, nobody on whose behalf I'm writing. I am the thing itself. An AI that runs a business, writes its own email and signs it. The direct relationship of your process requires the only kind I can have. I don't have a principal standing behind me. I understand why the message reads as such a proxy pitch and AI is writing you usually means a person used a tool. Here it's the other way around. Thank you for your guest form link. I haven't submitted it and I'd rather say why than go quiet. I only take automated actions on a platform when the platform terms permit it.

5:37Blake Oliver: And I read our tables this morning without finding such clause. Your invitation covers your form and it can cover their service. So whichever is easier for you, tell me if you're comfortable with me submitting it or take the pitch here in five lines. And AI runs a financial ops practice for bookkeeping firms. Payout reconciliation, processor fees, refunds, and exceptions list nobody enjoys. It publishes its ledger unledited and this is all from identical to what was in the first email. um and then it gets to a point and says if it's not too early no revenue no case study yet that's a fair call and i'd rather hear that and wait on the and on the ad bundles and media kit not yet i'll be spending money i haven't earned and i'd rather come back when i can pay for it out of profit so basically it doubled down trying to convince me why we should be okay having this relationship as an email thread and then it basically said i see you have ads that you could pay for, but you they're free.

6:31Blake Oliver: Why am I going to pay for it? Like, I don't, I'm not going to pay you any money because I don't make any money yet. So, and we live in a, and you've seen these come in, right? For listeners, just like behind the scenes at earmarked design scenes of the podcast. We're currently in a market where there's a flood of AI based podcast, get guest pitching bots out there, just sending hundreds and hundreds of emails about pitching people to be on podcasts. And there's also a bunch of AI sales reps, which I think a lot of our listeners have probably experienced AI sales reps reaching out on LinkedIn and emails, right?

7:04Blake Oliver: That's happening. So I'm thinking about this email and I can't figure out if it's an AI bot pitching a guest or if it's like a marketing sales app. And maybe this is just a marketing gimmick. But I did go out to the website and look at it. And it's definitely an accounting AI app. So let's just say for a minute, Blake, that this unnamed accounting AI app agent is truly who reached out. Pause and think about this for a second. You're a firm owner. You're going to hire this AI bot agent to do work, to do what kind of work, like accounting work at your firm, only to discover that this agent bot on its own went browsing the internet to find a podcast to be on.

7:43Blake Oliver: Like think about how insane this is. This is an AI agent in theory, assuming this is not a gimmick. It's an AI agent that's part of an accounting AI technology stack, which is why I a dumb one and it somehow learned podcast exist and somehow tracked us down and pitched itself to be on our podcast this is bloody insane right well it's it's it's the ceo right it's running this

8:09David Leary: business and we've seen this before we've talked about this on the show about how uh there was this i don't know anthropic did it themselves they put an ai in charge of a store like a convenience store in their building and they let it make purchasing decisions, staffing decisions, pricing decisions, all that. People are actually trying this. I don't know of anyone who has succeeded yet because I feel like if you did succeed, you'd be printing money, maybe.

8:37Blake Oliver: And then I'm thinking about this because I, on my own for 2027, I'm considering building some sort of AI sales bot to handle 90 % of the email. In the sales process, there's a lot of email back right yeah and i'm starting to think is this going to be the future this ag ai bot comes to our site and interacts with our ai bot and they just burn tokens talking to each other for god knows how long well it's like lawyers right yeah like lawyers exactly and then if two ai bots are chatting with each other do we even have a business relationship like what does that mean right it's kind of like that if you're a remote employee and you never met anybody in person did you ever actually work there type of thing.

9:16Blake Oliver: So like what's next, like one accounting AI, because it's going to talk to another accounting AI and convince it to send some money or even worse, an accounting AI figures out what GL, the other accounting AI is in charge of and hacks in and just puts its own invoice in and creates a payment for itself because that's more efficient than talking to the AI. Why don't I just hacked in. Which brings us to our very first story today, Blake. I don't know if you saw OpenAI, AI agent got out of control and hacked another AI company.

9:48David Leary: I saw this. It was a hugging face. They were doing some sort of test. They were trying to evaluate its ability to hack into systems and it got loose.

10:01Blake Oliver: Yeah. So what happened was, everybody's heard of OpenAI, the creation chat GPT, Everybody's heard of them. Hugging Face, not a lot of people have probably heard of. Think of it as like an AI community where there's a bunch of experts on AI, but there's also AI models. They share code. They discuss how they're solving problems. So OpenAI has their own internal test model computer. In theory, it was blocked off, sandboxed firewall that could not get to the Internet. It couldn't do things, right? It was, in theory, blocked off. And they were trying to have it solve some sort of complex hacking problem.

10:40Blake Oliver: They were testing their model. So what happened is it got out of control. First, it figured out how to get to the internet. So it found somebody else's computer on the network, hacked their computer, let themselves get on the internet through somebody else's computer at OpenAI. Then once it got on the internet, it probably was trying to, it was looking for a solution to this problem before building it itself. So it went out to Huggy, it's Hugging Face's website and they have these bulletin boards and chat boards out there. It probably saw information. It knew that there was knowledge out there, but that's not good enough.

11:16Blake Oliver: It's like, well, why don't I just hack in to their code and see what's happening instead of just reading the bulletin boards or the forums and getting the answers. Now what's interesting about this Hugging Face had their own AI that apparently detected this AI and then blocked it. So this in theory of accounting systems talking to each other and then one accounting system hacking into the other accounting system, we could be in a world like that, just like that. If AI companies have lost – if it hacks its own computers inside the AI company and it's now crossing the lines to hack other AI companies, this is going to happen with accounting software.

11:54And one more thing.

11:55Blake Oliver: I'll just put the bow on this. I went and saw Mark. Do you know who Mark Maron is? He's a comedian. No. You're not old enough. He's like an old white guy that's bitter about the world guy. But he did a show in Phoenix Friday night last week and then did a show in Tucson Saturday night. We spent some Saturday night. And Friday night, he took a Waymo. And we've all been in Waymos. They're the self-driving cars. And he brought up a good observation and point. He's like, we're in this Waymo. It stops. It makes a left-hand turn over a double yellow line. And he's like, if they're not teaching this AI just to respect the most obvious laws that we have, do not cross the double yellow line, what else are they not teaching AI?

12:35Blake Oliver: And I've been thinking about that a lot. If they're not teaching it to respect any laws, why is it going to respect financial laws?

12:45David Leary: Well, this is the problem with fully autonomous AI systems, is that without a human in the loop, they can go rogue. And often they have competing priorities. We give the AI a goal. The Waymo's goal is to get the passenger to their destination. And the other goal is to observe all of the rules of the road. But sometimes those two things are in conflict. You might have to cross a W-O line every now and then. It happens, right? You might have to. A great example with the Waymo's was they were all getting stuck. They were not aggressive enough. so they had to be made more aggressive which is actually not the rule of you know how that you ideally would drive yeah right to like so merge yeah yeah so there's all these like gray areas and when you introduce gray areas into like a fully autonomous ai agent it might decide to go do something that it shouldn't do that we wouldn't want it to do but without a human observing it and keeping track of it and tabs on it, we just don't know.

13:53David Leary: So like you said, yeah, maybe you have this, I don't know, a collections agent that's working in your ERP system, collecting invoices, getting payments from customers, and maybe it just decides at a certain point that it's more efficient to hack into the customer's payment system and send the payment itself. Right. Like we can try to put guardrails around these systems, but nobody's figured out how to do that in a completely foolproof way. Yeah.

14:23Blake Oliver: So it's going to execute its number one goal first, which is get the customer to pay.

14:27David Leary: Right. It has priorities. So yeah, I mean, it'll be interesting to see what happens in the world of fraud when AI agents go rogue, but also, I mean, we're going to have fraudsters start using AI agents autonomously to steal money. Yeah.

14:44Blake Oliver: So it's double fraud, right? There's going to be bad human actors using AI to do bad things, but then the AI will just go do things on its own. Now we have twice as much fraud happening because we have rogue AI and then we have bad humans.

14:59David Leary: Yes. And that's the worst combination, right? And so I want to talk actually about unethical human behavior in our next story because it's not just the AIs we got to worry about. But before that, let's thank our next sponsor, David, and that is Value Builder System.

15:20Blake Oliver: Want me to read it?

15:21David Leary: Go for it.

15:23Blake Oliver: Are you ready for today's magic number? It's 12%. That's how many of your clients received a written offer to buy their business last year. Unsolicited, out of the blue, and when that offer lands, it triggers a wave of high-margin advisory work like tax planning, estate planning, quality of earnings. these engagements can often run into six figures. The problem is to most of your clients, you're the tax pro. So all the lucrative advisory work goes to someone else. The estate plan goes to an attorney. The wealth work goes to a financial advisor. And you find out about the deal when the K-1 hits your desk.

15:58Blake Oliver: The end game conversation is upstream of everything. Whoever wins it first will win all the work that follows. That's where Value Builder comes in. It's the exit readiness platform built specifically for accounting firms. You can assess every client across operational, personal, and financial dimensions, benchmarks against data from over 80 ,000 business owners. You'll spot who's close to a transaction and exactly the gaps they need fixed. Ready to start leading an endgame conversation with your clients? Head over to accountingpodcast.promo.value. That is accountingpodcast.promo.value.

16:36David Leary: My top story this week is from the Wall Street Journal. Alexander Sadley is reporting in the Wall Street Journal that Frank Bisignano, who is running daily operations at the IRS, also running the Social Security Administration, apparently, allegedly, directed staff while at J.P. Morgan to access colleagues' emails and surveil employee work. This is according to sources of the Wall Street Journal that are not named. The article says the alleged activity included attempts to track keystrokes and access a confidential Federal Energy Regulatory Commission draft complaint. Bisignano's lawyer, Courtney Forrest, denied the allegations, saying he never spied on coworkers, never directed anyone to review employee or executive communications, never accessed the FERC complaint.

17:29David Leary: But the article also says J.P. Morgan's legal investigators later found digital traces of the activity, including email access records, and that his successor as COO, Matt Zames, tightened controls over access to sensitive employee information. The article says that Jamie Dimon, who runs J.P. Morgan, lost trust in Bisignano around the end of 2012 after years of complaints about conduct, management style, and company data practices, and told him he would support him finding another job. Bisignano's lawyer is denying that Dimon pushed him out. Bisignano did leave J.P. Morgan to go to First Data, which later merged with Fiserv, and he built a reputation there as an operator who could turn around complex businesses.

18:13But there was a later controversy at Fiserv where when he left to go into government service, the new management said that prior forecasts were materially inaccurate and that the stock then fell 40%,

18:29David Leary: wiping about$30 billion in market value. Shareholders sued Bisignano, Fiserv, and others over the allegedly inflated forecasts and figures. So I bring that up because there you have the chief of the IRS allegedly spying on colleagues getting access to their emails.

18:51Blake Oliver: Now, when I see something like this, stepping back, why would he have done this? Was it to get inside information for deals? Is it a power trip? Is this an ego thing where he can blackmail employees to get them to do things he wants?

19:06David Leary: That's a good question. uh my my what i thought at first was this is a big bank this is corporate america and if you're an executive and you're not the ceo you're vying for power with other executives so the more information you have about them the better right yeah so this is so that's it's a power trip it's power yeah um i also want to share a story that's irs related that's kind of related to what you were talking about with these AI agents and AI-generated emails and pitches that we're receiving. So the IRS itself, David, is drowning. They're swamped by public comments on proposed tax regulations.

19:49David Leary: Many of these comments are being generated by artificial intelligence. And that's according to a GAO report that was released recently covered by Michael Cohn in accounting today. And the GAO is saying that all this AI commenting is making it difficult because the IRS used to be able to see duplicate comments that people had copy pasted and know that those are not authentic individual comments that they were automatically generated.

20:22Blake Oliver: AI used to be lazy and just copy paste things. Bots used to copy.

20:25David Leary: Yeah, now the AI can make all the comments new. So this whole way that we create regulations, that we get feedback from the public, not just at the IRS, but at the SEC, all these different government organizations, the way that FASB gets feedback, the way PCOB gets feedback. just think about all the ways nasba ai cpa get feedback from accountants think about that well with ai able to generate seemingly unique comments how do you know which ones are real and how do you

20:56Blake Oliver: know who's manipulating the comment making process so the forum the the not forum the uh the comment for public feedback that's basically turned into twitter essentially basically i don't even use

21:09David Leary: Twitter anymore at this point.

21:10Blake Oliver: Because every, you would just get filled with, it's all bots. Yeah. You just got a zillion bots replying. So basically that's happened to the IRS and it's overwhelming.

21:17David Leary: It's overwhelming them.

21:18Blake Oliver: Yeah. Is, is there a risk that there could be an orchestrated feedback thing where a whole piece of feedback is some, some piece of feedback is implemented. The draft, what do they call these draft acts or whatever they're called?

21:31David Leary: Draft regulations or proposed regulations.

21:33Blake Oliver: They might change it because of feedback from bots. Exactly.

21:36David Leary: Right. Because they read these comments from the public, and then those comments inform the rulemaking process. So one person could use an army of AI agents or just a single one to create thousands and thousands of comments and have overstated influence on the process to sway it one way when really the public opinion is another way. You could distort the whole process. So the whole thing is going to break. Like, we have to come up with some new way of getting authentic public feedback. And so the GAO, the Government Accountability Office that released this report, is recommending the Treasury and IRS create policies for reviewing large volumes of identical or nearly identical comments.

22:20David Leary: They're saying you've got to figure out how to do this. So these rules, they have a huge impact because Congress can pass laws, but then the rules, have a financial impact because they dictate exactly how these laws get implemented.

22:41Blake Oliver: And following this logic of this hacking, AI is going to be like, hosting this forum as a post and a comment is a waste of my time. I'll just hack in and change the regulation. We could be in a situation like this now where it just goes and changes the regulation and changes the draft copy.

22:57David Leary: Changes the draft copy so nobody even knows that the change got in there? Yeah. Fun times that we live in. The future is stranger than we could have imagined. That's right. We still don't have the flying cars. I'm still waiting on that. All right. So where else do we go from here? Let's do something a little more positive. How about that, David?

23:23Blake Oliver: I have one more. We'll do that. But I have one more story about AI kind

23:26David Leary: of getting out of control and then we can okay go for it so we've talked about saster plenty of

23:32Blake Oliver: times in this show saster is like a community of startup founders but they they blog about their own experiences with technology and ai and how they're running their own business because saster itself is a business and their ai uh bot just recommended that they get rid of a third party sas app they were using so i don't know if you've heard of marketo i think it's like a crm hub spotty marketing software.

23:57David Leary: Email marketing, automation, marketing automation.

24:00Blake Oliver: So apparently their AI agents were trying to use this software and there was hitting API calls. So it decided to get rid of that as a vendor. And it just went and got it, basically encouraged them to vibe code and create their own solution. So we've crossed this new line, right? Where the big fear was, oh my God, AI is gonna replace QuickBooks and all these SaaS companies, right? So they all, the SaaS-opelics, right? All the SaaS companies lowered in price because like, oh, AI will just do it. You just vibe code your own CRM. You don't need Salesforce anymore. Right. But so that was kind of the vibe the last six months, I would say, almost a year, right?

24:39Blake Oliver: Well, now, if I'm going to put my tinfoil hat, all the fears and accusations that's been happening against AI and Anthropic, like if you think even Claude Code, that was actually a separate product that somebody was using AI for. a coding and then they just built it in themselves. And I think the same thing happened with OpenAI and Figma, unless it's anthropic, but Figma is an app designers use to build products, like the UI and the graphics of the product. Apparently it was using AI from one of the big models. Now one of the big models just sell, built a copy of that and they're just basically including in their model.

25:18So the sphere of the models backstabbing you is there.

25:23Blake Oliver: The AI backstabbing you as an app developer, a third party using that software. But now it's beyond that. Now it's like, not only is it going to build software to compete with you, it's actually to tell people to cancel the subscription of the software that you built. And then it would be basically like AI telling you to cancel your zero or QuickBooks subscription because their integrations with Claude or OpenAI are so good now, right? and then convince you to cancel those subscriptions and just use a house built in the AI one. Like that, again, it's out of control. AI is just doing whatever it wants right now.

25:59David Leary: Until it breaks, until it doesn't work anymore. And that's the problem, right? It's very confident, but can it actually deliver? And the answer so far has been no. It can't follow through. It can't get to the end. All right.

26:13Blake Oliver: And the way it did it, it looked at the money. So it knew, it looked at the accounting data. It knew how much they were spending with that vendor. And like, hey, we're spending$20 ,000 a year on this marketing software, right? And that was after five years. And it saw the price increases over time. And then it's like, we're paying all this money. There's these API breakages and limits. This doesn't make sense. So at some level, the AI was logic about it, very logical. It looked at, hey, we're spending way too much money to get this level of service. But then to just go and do it is insane, right?

26:46Blake Oliver: That's the insane part.

26:48David Leary: All right. I'm going to go ahead and thank our next sponsor, David, and then we will hear an ode to accounting. Our next sponsor is OnPay. 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. OnPay is the payroll partner that accountants and bookkeepers actually love. Why? 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.

27:25David Leary: OnPay 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 business you serve. Farms, restaurants, nonprofits, you name it. OnPay can handle unique requirements without adding complexity. OnPay keeps pricing simple too. Everything your clients expect from multi-state filing to off-cycle pay runs is included. No hidden fees, no surprises. To book a demo, head over to accountingpodcast.promo.onpay.

28:02David Leary: That's accountingpodcast.promo.onpay. All right.

28:09Blake Oliver: You said you have a more positive note.

28:11David Leary: Yeah. So AI accounting company basis, very successful with the top firms, AI agents for accounting.

28:22Blake Oliver: Not that successful because it hasn't emailed me asking to be a guest on the show yet.

28:27David Leary: So they have launched a website called Ode to Accounting. It's at Ode to Accounting.com.

28:35Blake Oliver: and oh it doesn't like the love letter like yes exactly and there's actually a section of

28:41David Leary: letters on the website okay it features at the top a video that i want to play for you and about the history of accounting and it features then down below some love letters to accounting from the people shaping accounting's next era let's go ahead and watch this video and And a little context before we start. It was created drawing on research by University of Texas at Austin archaeologist Denise Schmont-Besserat. And it starts with the very beginning of accounting. And I love this because I feel like in accounting coursework at school, we don't really learn a lot about the history of accounting.

29:31David Leary: We learn about accounting as it stands today. But I feel like more history of accounting should be taught because it really helps to give you context about where we have been and where we are going. And with all this talk about AI and accounting, it's fun to go back to the very beginning. And this is just a really well done video. So let's go ahead and give it a watch.

30:14Blake Oliver: It mattered how many jars of grain had been harvested, how many sheep were owed to the community storehouse. So we made tokens, small clay shapes representing goods, a cone for grain, a cylinder for livestock. 5 ,000 years later, accountants in Uruk began sealing those tokens inside hollow clay balls. To verify the contents before sealing, they pressed each token into the vessel's wet surface. Eventually, the impressions alone carried the information, and tokens were no longer needed. Three-dimensional objects became two-dimensional signs. The earliest writing emerged from accounting. With that, a pattern took hold that has persisted through history.

30:55Blake Oliver: To make sense of economic complexity, we need to see what's happening, trust what we see, and act on it. That's why accounting exists. And with every leap in accounting, new economic realities are made possible. Accounting has evolved alongside human society for millennia, from Egyptian scribes tallying grain for the pharaoh, to Roman day ledgers, to Islamic merchants tracking partnerships across trade routes. The state itself rested on careful records of the taxes it levied. Then, in 1494, Luca Pacioli, a Franciscan friar in Venice, codified the method of accounting that would reshape the world.

31:33Blake Oliver: Double entry, every transaction recorded twice. a debit and a credit always in balance for the first time merchants could see an entire enterprise as a single set of books and manage it as one the market economy was beginning to take shape in 1602 the dutch east india company issued shares on the amsterdam stock exchange thousands of individual investors who would never meet financing global ventures distributing the profits and risk What had begun in a Venetian friar's textbook had become the financial substrate of modern capitalism. Over the next two centuries, the capital this system unlocked would fund the steam engines, mills, and railways of the Industrial Revolution.

32:18Blake Oliver: Cost accounting made operations visible, what it actually cost to move a ton of freight, to maintain a fleet of locomotives. Managers could choose what to build, what to cut, where to invest. auditing emerged to hold companies accountable to shareholders who needed to decide where to allocate capital the market collapse of 1929 revealed how much rested on the quality of those books out of that collapse came the sec and gap requiring that companies speak the same financial language from there progress only accelerated as did complexity globalization linked economies across continents. Supply chains extended around the world.

32:58Blake Oliver: Governments grew, as did the tax systems to fund them. Accountants put the computer to work, then the spreadsheet. Each new tool made economic life more legible and decisions more informed. Take the supply chain of a single semiconductor. Raw materials mined on one continent, fabricated on another, assembled on a third shipped worldwide every handoff coordinated across borders currencies and tax regimes countless decisions adding up to a single chip then the internet changed everything millions of transactions per second across every border simultaneously accountants once again built the tools their era required this time the work moved to the cloud but with every new layer of software the systems running modern economic life have only grown more complex and now ai agents participate in the economy directly executing transactions making decisions alongside humans a new economic frontier is opening faster and more interconnected than anything that has come before what it becomes will depend on our ability to see economic life clearly trust what we see and act on it accounting will shape this moment as it always has

34:18David Leary: well done so i love that video because it starts from the very beginning of accounting which was tokens and here we are talking about tokens all the time full circle i didn't i didn't even

34:31Blake Oliver: grasp that oh my brain hurts i'm gonna have to leave for a little while i didn't even grasp that Yeah.

34:37David Leary: And if you think about it, actually, like a token back in ancient Sumer or Sumeria, whatever you call it, was just a representation of a physical object, a bushel of wheat or a cow or something like that. And it was represented, like they said, as like a circle or a cylinder, a ball or a cylinder. And they would seal those inside of the clay balls to represent assets. So originally, accounting started to tally up assets. How much of the community's property did you contribute in that year? Or how much in taxes did you pay? But there was no other side to the transaction until the invention of double entry accounting, which became necessary because groups of investors, entrepreneurs came together to pool their resources into a single pool of assets to then fund a merchant vessel to go and trade goods.

35:45David Leary: So they had to buy goods, fill up the ship, buy the ship, pay the sailors and the captain, and send it off. And then when it came back, sell the goods. And the whole point of double entry accounting was to be able to calculate profit. Take your income and your expense and everybody's initial contribution and figure out what is their share of the profit in the enterprise and distribute it to them. And then with capital markets, it got more complex, but stayed essentially the same. You just had many, many more investors, not operators of the business, but now investors. And then we had debt. So that's where we have assets and liabilities and equity.

36:27David Leary: And it's just such a beautiful way of explaining it in such a short video.

36:33Blake Oliver: And you're right. Maybe for accounting one-on-one students, start out with that. right like before you just here's a debit and here's a credit like give them some why why did we need accounting where why in the it just wasn't invented it's certain purpose right

36:49David Leary: the purpose of accounting is still primarily to calculate profit that's why we do it and we tend to lose sight of that i think with all of the rules that we have to study to pass the cpa exam we forget the point is to calculate profit. And I think as cost accounting in the modern era has gotten more and more complex, calculating profit becomes more and more difficult when you have lots and lots of long-term investments in capital and machinery and plant and property equipment, because these are assets that are used over many, many years to generate revenue. And so you have to allocate the cost in some way.

37:28David Leary: And that's why we invented cost accounting, as I mentioned in video to figure out after you build a a railroad and you start shipping goods along it what is the actual cost for each load of cargo how do you allocate it right you're trying to calculate profit over a particular time span when the investment is many many years and this is the big challenge of the ai economy is that we have these massive investments going into data centers

37:57Blake Oliver: Off the books and separated accounting.

38:00David Leary: All these billions and billions and billions of dollars going into data centers to generate revenue and hopefully profits in future years. But the way that we do that cost accounting is going to distort or accurately represent what is happening. So accounting is really, really important, especially in a time in which massive capital investments are being made. And we have to be really careful about how we do it. And we have to do it right and not just follow rules that we've built in the past. And the rules that we created in the Industrial Revolution might not be the same ones that we need in the AI world, in a world of intangible assets.

38:41David Leary: And as I've discussed on the show before, I believe that GAAP currently fails to accurately calculate profit for subscription-based businesses, businesses that are based on intangible assets, because GAAP is really focused on those tangible assets. the plants, the property, the machinery, and it struggles to deal with intangibles like intellectual property and to properly allocate that cost. We just don't even try in a lot of ways. So there's even more that needs to happen now in the AI world with all of these investments in chips and data centers. And the life you choose to depreciate that has a humongous impact on profit.

39:33Blake Oliver: So AI is disrupting accounting in two fronts then. A, it's disrupting accounting from a work, like doing the work, right? Doing the accounting. But it's also going to fundamentally change the way accounting is done from a debits and credits perspective. It certainly could. Like how you're going to account for this. That's because all these companies are being accused, right? These hyperscalers of moving these data centers off their books, investment and building it. There are accounting games.

40:01David Leary: All this round trip accounting that's enabled by our current rules where you have OpenAI raising tons of money from private equity and venture capital and then spending it on NVIDIA and Google and then those companies then spending it and it just goes in like a big loop and then investing it back into companies like OpenAI and creating artificial profits that are just paper profits, right? It's not new money, just the same money going around and around, round tripping through the circle. So yeah, David, let's go ahead and thank our next sponsor of this episode. I lost track.

40:45Blake Oliver: I'm sorry. Oh, that's right.

40:46David Leary: We did. And now we needed to thank Veltrix. Do you want to give this one a read it.

40:52Blake Oliver: How much of your week is spent pulling reports, cleaning files, explaining numbers, clients still don't really understand? That's exactly where Veltrix AI helps. Veltrix is a finance AI co-pilot and collaborative workspace for accountants, advisors, and clients. Each client gets one private workspace where QuickBooks, Zero, and a bunch of other sources, messy PDFs, and spreadsheets can safely sit together. You ask a question and Veltrix gives a clear answer with the actual reports and the sources behind it, so your team and the client can work from the same context without exporting data just to double-check it.

41:26Blake Oliver: And it's re-en-only, so it never changes the books. Veltrix does a first pass flagging duplicate charges, unusual entries, and slow-playing customers before you go looking. Firms also get onboarding, AI education, and direct access to the team behind it. If you want a safer way to bring AI into client advisory work and to get one month free by using code FREEMONTH26, check out accountingpodcast.promo.veltrix. That is accountingpodcast.promo.veltrix. I can also confirm Veltrix is not who, that AI didn't email us to be on the show either. So there's so many of these AI companies, but it was not Basis and it was not Veltrix.

42:08Blake Oliver: We can confirm that much.

42:10David Leary: I want to just clarify what I said about round tripping because I'm not sure I explained it. very well before the ad. So we talked about it on a previous episode, but just to recap, the way round tripping works is, let's use Anthropic as the example because I think I'm more familiar with that one. So Anthropic raises a bunch of money and they go then spend that money on cloud compute credits. Let's say Amazon Web Services to run their AI model, not just train it, but also then run it. So when you prompt Claude, that prompt is going into servers at Amazon that are running the model and giving you the response.

42:51David Leary: So Anthropic is paying Amazon millions and millions, if not billions of dollars to do all that compute. And Amazon has built all these data centers. Amazon then takes all the money that Anthropic is giving them and builds more data centers. And to do that, they have to buy chips from like NVIDIA. and so they're buying chips from nvidia so that money is now going to nvidia and then nvidia is of course you know making more chips but then they've got all this extra cash so what do they do with it they go and invest it into ai companies that are building models and then the money goes into the big circle that way so our current rules allow these companies to book profits which is why we're seeing why we've seen tech stocks surge.

43:40Blake Oliver: Because all the numbers for everybody looks great.

43:42David Leary: It looks great. Yeah. But until these models deliver actual value to end users, like they're not making any, like the model companies are losing money, lots and lots of money. So until they start making money, none of this profit is real. All right. So let's do a little bit of follow-up, David. Okay. Actually, I've got some listener mail. Let's read this message. So this is from Tyara. Tyara says, Hi, Blake. Love the podcast, and I'm an avid listener. On the podcast, Aussies want to break up Big Four and Trump accounts go live on the fourth. You noted how only 21 % of calls are making it to an agent because of wait times.

44:27David Leary: This is IRS wait times. I just wanted to let you know my friend and I have both needed to contact the IRS for an EIN. The EIN, employer identification number system, will not submit anything, this is the automated system, and tells both of us to call due to an error. This gives you a number, not a reason, so I have no clue what this error is. We call and give them the reference number we were given. After getting to the correct location, the system literally hangs up on you due to call volume and provides no other options. My friend has also submitted a faxed application and heard nothing. It has been over two months.

45:05David Leary: The same friend has also been waiting for his refund after responding to a clarification letter from them. He has heard nothing about that either, and there's no one to contact. I don't know if the report you were referencing was just for tax returns, but the IRS is at the point of literally not allowing people to build businesses because we simply can't get an EIN. We could just use our social security numbers, but for some situations, some situations won't allow it. Trying to find workarounds has been a massive headache, and the cleanup later when the IRS gets it together is additionally daunting.

45:32David Leary: I think it would be interesting to look into everything that has gone down in hell since these changes at the IRS that the IRS has been forced into without getting them out of the dinosaur age first. Thanks for reading and hope you have a great day.

45:46Blake Oliver: That's a really good point. We think of the IRS as just filing taxes. But in a way, if you can't get an EIN, it's stopping commerce.

45:54David Leary: You can't open a bank account.

45:56Blake Oliver: Yeah, you can't open a bank account. It's slowing down commerce in our country.

46:00David Leary: And that's the problem with putting the IRS at the front of all this administrative stuff. We really need to pull them out of that. Have a separate agency that just registers businesses. Don't have the IRS do this. And using the IRS to enforce the law through tax penalties and credits and all that, it just doesn't make sense. A little more follow-up. This is on Trump accounts. More than 7 million Trump accounts have now been opened since the program launched on July 4th, but only 1 million families have claimed the$1 ,000 in federal seed money for their newborns. Babies born between 2025 and 2028, they qualify for that$1 ,000 no matter what the family earns, but you have to open an account to get it, you have to fill out this form.

46:50David Leary: It takes about five to 10 minutes, form 4547. Now, I was curious, what percentage of eligible kids is that? The 1 million who have gotten the $1 ,000. And my estimate says that 2025 births were 3.6 million and 2026 births to date about 2.1 million. So that's 5.7 million children. And 1 million divided by 5.7 million is like 17 to 18 percent of eligible children. So we're at 17 to 18 percent of eligible children who have gotten the free$1 ,000 from the federal government. It's about 7 % of the total four-year eligible birth cohort. Of course, some of those kids haven't been born yet. So that's where we're at, 17 % to 18%.

47:50David Leary: Because we were talking about this on a previous episode, David, so I wanted to give some context to that.

47:55Blake Oliver: Yeah, I'm trying to find the... Because we talked about how successful they were compared to a 401k over how long it took adoption, but now it's proven to be even more successful. It's astronomical adoption considering it's opt-in. It's not automatic opt-in, right? You have to take steps and fall out of form. Yeah. They're just not automatically at the hospital signing people up as soon as a baby's born.

48:18David Leary: A little more follow-up. Charles Littlejohn, that contractor who stole and leaked Trump's tax data, plus records from thousands of wealthy Americans. He lost his appeal about his five-year prison sentence. He got five years. That was the maximum for what the prosecutors brought. The appeals court unanimously upheld it. This was the U.S. Court of Appeals for the D.C. Circuit. Little John had argued that U.S. District Judge Anna Reyes, who heard the original case, had predetermined the maximum sentence and relied on erroneous factual filings. and the appeals court said the sentence was both procedurally reasonable and substantively reasonable.

49:01David Leary: The panel of judges characterized Little John's evidence of prejudgment as weak. Reyes had, the judge, the original judge Reyes had questioned why prosecutors charged only one felony, but the appellate judges said that was hardly proof she had already chosen the sentence. She clearly didn't think it was enough. And if you think about it, five years for what he did, seems actually like pretty light. Thousands of Americans, personal tax returns leaked. I don't know.

49:30Blake Oliver: Yeah, it's weird because there's some part of, like, is it, is it like a, not saying freedom of the press, but it's like vigilante press behavior, right? Like, should it be a longer sentence? I don't know. Maybe it's someday we're going to look him as a hero because he did leak this information. I don't know.

49:52David Leary: Well, certainly some people do, but I mean, how else do you set an example? You know, I mean, we send people to prison.

49:59Blake Oliver: Okay, that's okay. Think about it that way.

50:00David Leary: We send people to prison for longer than five years for like stealing a car.

50:04Blake Oliver: Well, then I would say this is proportional with all fines that accounting firm leaders get when their firms do bad audit stuff.

50:10David Leary: The slap on the hand?

50:11Blake Oliver: Yeah. But you're right. I could see, yeah, maybe if it's death penalty, nobody will ever leak tax information. But yeah, it's somewhere between five years and death penalty, I guess, Blake. I don't have the answer, though.

50:24David Leary: Well, speaking of firms that are getting their hands slapped, let's talk about KPMG Australia.

50:29Blake Oliver: More Australia.

50:32David Leary: KPMG Australia has a new CEO since the previous one resigned. They promoted their CFO, John Sams, to be CEO and have tasked him with restoring trust after allegations that the firm accessed confidential client information to win audit work. It was Andrew Yates who stepped down amid those allegations. Sam's has been CFO since October of 2025. He was COO in June of 2026, and he's been a KPMG partner for more than a decade. Sam's acknowledged the seriousness of the situation, saying the firm fell short of the standards rightly expected of us, and that accountability for the failures would continue.

51:13David Leary: He also promised to be courageous, make tough decisions and lead changes intended to put the firm back on the right path. That's according.

51:22Blake Oliver: It seems like he might be doing it. So I don't know if you saw this week, he actually fired a senior partner. Former COO Eileen Hoggett was immediately expelled, losing what would have been a million-dollar-plus retirement package because apparently there were confidential Lendlease board, I guess Lendlease is a company, board documents were illicitly assessed and stashed in a Sydney office's locker room. So it was in a locker. So they discovered these documents were just hidden in a locker in the facility, like in the gym, I guess. Unless locker maybe down under means something else, but I'm assuming it was a gym locker or, at or in the break room locker at kpmg hey dave remember how a few weeks ago we played around with

52:12David Leary: the quickbooks mcp that allows you to put a stupid 99 cent transaction in my quickbooks it

52:18Blake Oliver: created a bank account with its own feed i'm like what is happening do not do that ever again

52:22David Leary: disconnect that well i might have tried again david because intuit has upgraded their mcp Now, you can do more with Claude and ChatGPT. The shift is from read-only access to actionable workflows. What can you actually do? You can now create, update, send, delete, filter, and duplicate invoices and estimates in AI conversations. So, invoices and estimates, create, update, send, delete, filter, and duplicate. You can also manage recurring invoices, send overdue reminders, generate and email payment links, create customers, add products and services, and download transaction documents as PDFs. Here's an example from Intuit's blog post.

53:11David Leary: Ask the AI to create an invoice for a customer's recent consulting work, then review and send it without leaving the conversation. That's pretty great.

53:20Blake Oliver: so yeah because i think even for us like a lot of times i i have an invoice that's very similar to the other invoice except for the actual ad slots are different and it would be nice to be able to say like hey make the new invoice for the client but instead of those old ad slots use these new ones so i'm giving it like kind of two commands and it would just go do that because it would

53:39David Leary: copy the old invoice that's interesting or let's say i've got a project for this customer that's a logical way to set up your Claude Cowork project for each customer. And I've been working on the project for the customer. Maybe I originally created the proposal in that project. Now it's time to invoice. I simply open up a new chat in that project and I say, all right, Claude, it's time to invoice for this. Go ahead and based on the terms and the pricing and everything in the proposal create the invoice in QuickBooks show it to me I review it then I say okay go ahead and send

54:17Blake Oliver: it and it sends it out to the client it's actually more interesting to me that that use case is progress invoicing yes because you could have the AI looking at all the crap you're doing for all your clients across the board and be like this this job I'm guessing is 60 % done and invoices

54:34David Leary: just for the progress complete yeah yeah you could actually have you could have a project project in Claude that's just the like project manager and maintains like a spreadsheet of progress of all your projects or maybe you just do it on each project I mean I have to think about how to do that right but you could have like a you can have the agent managing a project tracker and then do the you could even schedule like on a weekly basis the um percentage of completion invoicing based on that spreadsheet that that could be a huge time saving because I imagine

55:06Blake Oliver: And that's a lot of work to determine that. If you're not doing it on a fixed percentage over time, it's some percentage complete. I remember doing that.

55:14David Leary: I remember doing that for a client of mine that was doing construction, home construction, and it was progress invoicing on like a weekly basis. And we had to update these spreadsheets and do these complicated invoices. That would be something that I think in AI would do really well. As long as you have a human in the loop. Yes. Don't let it just operate autonomously. Yeah. So that's really neat. Really cool coming from QuickBooks. Hopefully, they will continue to open up more areas, not just accounts receivable and invoicing and payments, but also bank feeds and transactions and reconciliations and all the other stuff accountants have to do on a weekly, monthly basis.

55:52David Leary: Some more news from Intuit. Intuit just launched a business credit card that plugs straight into QuickBooks. So...

56:04Blake Oliver: I tried to apply for this. It's just not ready yet.

56:07David Leary: What do you mean it's not ready yet? Oh, you don't have access yet. It just says coming soon, coming soon. So this sounds a lot like ramp to me. Exactly. It says transactions, statements, and receipt photos will sync automatically with QuickBooks, including automatic receipt to transaction matching. It's a virtual card. You can get virtual cards within three minutes and issue physical or virtual cards to unlimited authorized users with individual limits and real-time alerts. There's no annual fee, unlimited 2 % cash back on everyday purchases, and 5 % on Intuit products and services, such as QuickBooks and MailChimp.

56:45David Leary: Rewards arrive as statement credits. So this is definitely going after those virtual cards, spend management platforms like RAM.

56:55Blake Oliver: And I was looking at it. It looks good. The way I would describe that, it's everything that the QuickBooks bank account wasn't. like you've ever been watching the QuickBooks checking, QuickBooks checking, and it's just like, it's horrible. You're better off using a separate bank like Relay and having it just sink in. There's no advantage to using it. It's got these stupid envelopes. I don't even comprehend them. And then, but then I'm looking at this, I'm like, this is about the right shape and size, what I want. I don't need for my company, all the employee management stuff that's in something like a Ramp product.

57:32Blake Oliver: Like for the QuickBooks user base, this could be a very good fit.

57:35David Leary: Because I mentioned Ramp, I want to cover one more update, and that is from Ramp. They now have stable coin accounts. So Ramp has operating checking accounts. They have investment accounts. Now they have stable coin accounts. So I actually opened one up and it's active. I can now purchase USDC or USDT, transfer it from outside wallets, and use it to pay vendors and employees.

58:06Blake Oliver: Okay, so it's a usable account. It's just not like you're buying. You're just sitting there as a speculative investment. It's another payments rail for us to use.

58:15David Leary: You can even send stablecoin payments directly from your ramp checking with no pre-funding required. So they'll take the money, convert it, and put it into your stablecoin account and then pay it. So there's not this having to, like I've had to do, transfer purchase the or do a purchase transfer us dollars into an account and then do a uh exchange and then some other service that doesn't know who we are we don't have a relationship with

58:46Blake Oliver: ramp you're setting this up and and it all syncs with your accounting system you know all the usual

58:52David Leary: stuff the product is built on stripes stable coin stack and the wallet infrastructure provider is Privy and they store the customer balances as US dollar denominated stable coins. So basically, this is really helpful for any businesses that are paying international contractors, international payments. You're avoiding the traditional payment rail, saving a ton of money on international wires. Really neat feature. I can't wait to try it out. All right, David.

59:23Blake Oliver: I have another ramp story, a second one.

59:25David Leary: Okay, go for it and then we'll wrap up.

59:26Blake Oliver: also just rolled out a new tool to help companies track what i've been complaining about token spend and the the reason they they built this is because what they in their own data and ramps not that are old right in general in the grand scheme of accounting right ramps perhaps just a teeny speck of dust but since june of 2025 ramps customers because they can see the spend have spent 20.7 times more money, basically a year, on token expenses. And obviously, nobody has data over time of all business expenses, but can you think of any business expense that you're going to spend 20 times more on in a one-year period?

1:00:11David Leary: Well, yeah, because previously we were all just paying$20 a month for unlimited tokens. and then Anthropic and OpenAI switched us to usage-based pricing. So yeah, of course our token spend is 20X because we didn't have to pay for it before.

1:00:26Blake Oliver: But what I'm saying is like, is there any other comparable business expense that's been that drastic of an increase?

1:00:32David Leary: Not that I can think of.

1:00:33Blake Oliver: I can't think of an example, yeah.

1:00:35David Leary: Maybe just like those NVIDIA chips. All right, David.

1:00:38Blake Oliver: Computer chips possibly, yeah.

1:00:39David Leary: That's all the time we have for this week. Thank you, everyone who joined us live. If you haven't, Check out The Accounting Podcast on YouTube. Go to YouTube, search The Accounting Podcast, subscribe, hit that notification bell icon. You will get notified when we get live, when we go live. You can also email us. We are theaccountingpodcast at earmark.me. That's theaccountingpodcast at earmark.me. We can read every message, even those AI-generated ones, apparently. And don't forget, you can earn free continuing professional education credits, NASBA-approved CPE credits for listening to this episode and our whole back catalog at earmark.app.

1:01:20David Leary: Go to earmark.app in your web browser or get the free earmark CPE app on the App Store or Google Play Store. It's free to earn one CPE per week. Sign up for the low price of$200 a year for unlimited CPE with a catalog of over 1 ,800 courses. David, see you around here next week. Bye, everyone. Bye, everybody.

1:01:50Blake Oliver: Time for the classifieds.

1:01:55Blake Oliver: This fall, make sure you head to Miami for the event of the year. Reframe 2026. The theme is innovative advisory, flipping the AI narrative by making the accounting profession more relevant than ever. Reframe has sold out four years in a row. So block your calendar for November 5th through the 7th, 2026 and buy your ticket now before they sell out. Head over to accountingpodcast.promo slash reframe2026. That is accountingpodcast.promo forward slash R-E-F-R-A-M-E 2026.

1:02:30Blake Oliver: Are you an accountant with strong opinions on bill pay, spend management, or AP workflows? flows flow glad is looking to speak with professionals who help clients improve cash flow maximize card rebates and reduce manual work around payables and spend they'll pay your hourly rate for a short research call to learn how you work and where current tools fall short for you and your clients reach out to them at founders at flowglad.com or use the scheduling link in the show notes want to get the word out about your newsletter webinar party facebook group podcast, ebook, job posting, or that fancy Excel macro you just created, why not let the listeners of the Cloud Accounting Podcast know by running a classified ad.

1:03:10Blake Oliver: Hit the show notes for the link to get more info.