Baby, Where The Hell Is My Audit?
There may be errors in spelling, grammar, and accuracy in this machine-generated transcript.
David Leary: [00:00:04] So I have a LinkedIn post as 10,000 views, and I'm asking for people to put the name of one accountant that they know have lost their jobs because a company implemented some AI and now they don't need that account anymore. Zero names, Blake, zero names. It's been weeks zero names in this thread because I don't think it's true. 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: [00:00:35] I'm David Leary.
Blake Oliver: [00:00:37] David. This week I've got an amazing parody video to play for you. The Massachusetts State Auditor performed karaoke at a parody at like, an event on Saint Patrick's Day. It's hilarious. It's called baby, where the Hell is My audit? But before that, let's thank our sponsors who are our sponsors this week.
David Leary: [00:01:00] Our sponsors this week we have on pay and we have uncW Kenan-flagler Business School.
Blake Oliver: [00:01:07] Uh, thank you so much to on pay for sponsoring this episode. 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 partner that accountants and bookkeepers actually love for payroll. 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. Amp 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. They're 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 OMP keeps pricing simple to 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 The Accounting Podcast. That's The Accounting Podcast o p a y. And welcome to our livestream viewers. Great to see you. Our boring accountant.
Blake Oliver: [00:02:24] Happy Friday. Hazardous items. Good to have you with us. All right, let's open up with something a little bit fun. This is a video Massachusetts State Auditor, uh, what's her name? Diana. Diana Dizoglio. And she performed this parody at a Saint Patrick's Day breakfast. And it's aimed at the Massachusetts state legislature, which she says is still blocking an audit that voters already approved back in November of 2024, Massachusetts voters passed question one, giving the auditor authority to audit the legislature. It was a landslide, 72% in favor. And this is a performance audit, not just a financial statement audit. The measure lets the auditor examine the legislature's accounts, programs, activities and functions. And the reason that she's singing this song, where the hell is my audit? Is that the legislature has refused to cooperate, arguing that an executive branch auditor digging into the legislature violates separation of powers under the Massachusetts Constitution. So Dizoglio has sued the legislature, sued them on February 10th to force the lawmakers to turn over records. And on March 19th, a Supreme Judicial Court justice said the threshold legal issue should go to the full court. So this is going to the court and we're going to see what happens. But for right now, we're going to listen.
David Leary: [00:03:52] She rolls up into the executive branch. Right?
Blake Oliver: [00:03:54] That's right.
David Leary: [00:03:55] Okay.
Blake Oliver: [00:03:55] So here we go.
Audit Parody Clip: [00:03:58] Oh, baby, where the hell is my audit? Do they think we forgot it? Oh, tell them if you see them, baby. If you see them, baby. Why is the speaker waiting for me to get over? He's already testing my patience. I only fear he's taking time to destroy all the records and the age he's been taking vacations. Wait till I get my hands on them. Look how long they've kept us waiting, anticipating. And despite my frustrations, we keep saying they must conceive completely how the people want it. Are they far away? Are they okay? The speaker's testing me. Help me, help me, help me. Lord, I need you to tell me, baby, where the hell is my audit? What's the date? It's so long. To find me. Oh, baby, where the hell is my audit? Do they think we forgot it? Oh, tell them if you see them, baby. If you see them, baby. They should holler. I just filed in court, but the AG said no. Leaving the voters with nowhere to go. Access to justice. Just for some. We know where the judge's paycheck comes from. No wonder some are under the speaker's thumb. We already voted to force question one. You already voted for the audit, but the AG and court seem determined to block it. Come, come, come save me. How? The people want it. Are they far away? Are they okay? The speaker's testing me.
Audit Parody Clip: [00:05:37] Ha ha ha ha! Help me, help me, help me! Lord, I need you to tell me, baby, where the hell is my audit? What's taking so long to find me e Eat me. Oh, baby, where the hell is my audit? Do they think we forgot it? Oh, tell them if you see them, baby. If you see them, baby, they should tell him. Tell them all. Tell them, tell them mom. Tell em. Mom. Five, five. Tell em. Michael Brown eyes and a growing fear that if it doesn't find me now. I'm gonna die without it. Hurry up. Sir. Ah ah ah ah ah ah ah ah. I wanted wanted wanted wanted wanted. I would like a ring. I would like a ring. I would like a telephone ring. I would like a bright and shiny on it that I can wave around and talk and talk about. And when the day is here. Forgive me, God, that I could ever doubt it. Until death do I, do I, do I. Is he about testing me? Help me, help me, help me. Lord. I need you to tell me. Baby, where the hell is my audit? What's been taking so long to find me? Oh, baby, where the hell is my audit? Do they think we forgot it? Oh, tell them that my grandma said it. Tell them grandma said it.
Audit Parody Clip: [00:06:52] Your order is coming.
Audit Parody Clip: [00:06:58] I would like a ring. I would like a ring. I would like a telephone ring. Obviously, I would like a shiny on it. On it, on it, on it. On it. Where is our audience? Are.
David Leary: [00:07:16] Super impressive. Super impressive.
Blake Oliver: [00:07:21] I don't know why they're not on their feet.
David Leary: [00:07:24] There's a lot of awkward clapping along the way by awkward non rhythmic white people clapping. Kind of oddly.
Blake Oliver: [00:07:31] The beats on the beat. Yeah.
David Leary: [00:07:32] On the on the ones and threes. Um so somebody put a comment about afroman in the lemon pound cake vibes. And I was going to bring the story that to the show and I talked myself out of it, but now that somebody brought it up in the chat. Yeah. The Afroman case is kind of similar, right? He used social media and a music and a song to get justice.
Blake Oliver: [00:07:55] I have no idea what you're talking about. What is.
David Leary: [00:07:57] The Afroman?
Blake Oliver: [00:07:58] Afroman case?
David Leary: [00:07:59] Let's rewind to the mid 90s. Do you remember that song? Like, I didn't do my dishes today because I got high. Do you remember that song? Yeah, yeah. That's afroman.
Blake Oliver: [00:08:08] Okay.
David Leary: [00:08:09] A couple years back, he got his house raided because there was a tip turned in that he had drugs there and a kidnaping victim or something like this. His house gets raided. They wind up filing no charges against him, and he has these video cameras. So he made some songs that went viral on Spotify and TikTok about these police officers to went through his house. Called him out and essentially mocked them so bad that they filed a $4 million lawsuit against him. And he won the lawsuit yesterday against these police officers. But there is an accounting story because one of the. And that's why I almost brought to the show one of the things that.
Blake Oliver: [00:08:49] You are bringing to the show. Now, what is the accounting.
David Leary: [00:08:51] In the show? So one of the things they did is they confiscated about 4000 300,000 or $4300 in cash from his house, put it in an evidence envelope, logged it as a certain amount of money. When he got it back, it was short $400, right when he got it back from the evidence and the accounting story there is they didn't use proper cash accounting. They just counted the cash, right? They didn't. Instead of saying there was two $100 bills, six, $20 bills, uh, 17 fives and 32 ones or something like that. They didn't follow proper cash handling guidance. And that's that's the accounting record keeping. That's a stretch of accounting story, but it's kind of tied to the aftermath. And then and then he actually claimed that they took his money, his earned money that he paid taxes on already. So he, he, he keeps bringing that up. This is money I paid taxes on. So he's very upset about it. But it worked. This viral video type thing. This could basically it resulted in justice by doing these viral videos. And so maybe.
Blake Oliver: [00:09:53] It gets attention.
David Leary: [00:09:53] Maybe this, uh, is she the she's the state auditor.
Blake Oliver: [00:09:58] She's the Massachusetts state auditor.
David Leary: [00:10:00] Yep, yep.
Blake Oliver: [00:10:01] So maybe this will pressure the legislature into complying. Well, I guess it's going to go to the court anyway. So if once the court decides, then maybe the legislature will do it.
David Leary: [00:10:13] Now that you've played it, it's going to have all this exposure now everywhere.
Blake Oliver: [00:10:16] That's right, that's right. It's nationwide now worldwide. All right, David, let's talk about, um, would you take a probabilistic audit opinion? That's the question. Would you take a probabilistic audit opinion.
David Leary: [00:10:35] As in like, like a prediction market opinion on the quality of an audit. What does that mean? Probabilistic.
Blake Oliver: [00:10:41] Well, right now audit opinions are basically pass fail. Yes. No. You either pass the audit or and you get an unqualified opinion or you get a qualified opinion. Or I guess there's another option, which is you get like no opinion at all. So there's kind of three options.
David Leary: [00:11:04] It's black and white, except for they only use a sample to perform the sample of the data to perform an audit.
Blake Oliver: [00:11:10] Well, yeah, the auditor does all their work. Right. But what I'm saying is like what you get as the customer of the audit, right? That actual like opinion, that's the final product, right? Of the audit. All the work that goes into it is either these financial statements or I forget the exact terminology, right? It's like materially correct, right. These these, these, these financial statements are right materially or not.
David Leary: [00:11:38] So it's a skill of.
Blake Oliver: [00:11:39] There's.
David Leary: [00:11:40] No.
Blake Oliver: [00:11:40] 100%.
David Leary: [00:11:41] Correct. Yeah.
Blake Oliver: [00:11:42] There's no.
David Leary: [00:11:43] 0% wrong. And then there's no hey, it's about 65% of yes. Yeah. We don't have any of that.
Blake Oliver: [00:11:52] No. And there's also just like no grade. Like, hey, this is a, this is a five star set of financial statements. This is a, you know, four star, three star, two star, one star, right? There's no gradation there, the quality like ranking or anything. So anyway, the reason I brought that up is because I saw an opinion piece by Wenzel Reyes in accounting today. And he argues that the accounting profession, the audit profession in particular, faces a fundamental challenge from Agentic AI, which is that financial information is being generated, approved, updated by autonomous software, by AI, not humans. And audit standards today were built for human generated financial statements. It might not be sufficient anymore for providing assurance in an AI driven reporting environment. You look puzzled, David.
David Leary: [00:12:50] Well, I'm trying to like, like, should that matter? Like, like the AI should be doing should meeting the same standard that the human used to meet. So the audit, the, the, the how we measure the quality of that work should not change. You don't change the bar.
Blake Oliver: [00:13:05] So one of the examples he gives is like that investors now are taking all of the like financial information of a company and all this data and all this stuff, and they're feeding it into AI systems. And those AI systems are now basically auditing those companies and they're providing like a probabilistic outcome, right? Like a, like they're giving more than the audit opinion is giving now. So basically the audit opinion is just one of those factors and it's becoming less and less important.
David Leary: [00:13:39] Okay.
Blake Oliver: [00:13:39] So because the A pass fail.
David Leary: [00:13:43] Yeah, it's always a pass fail. But now if I'm using AI to do it, AI is going to be able to give me a number. How confident are you in this pass?
Blake Oliver: [00:13:50] So I think like I've always wondered why audit opinions are pass fail. Like why, why is that it? Right? Like it's either you stamp approved or approved with reservations or not not approved. That's it. Why are there only three possibilities? And I guess that's because when we came up with this whole way of doing things, that was all that was possible. That was all that you could do. You could like, because it takes humans to do this, and you have to create all these standards that having more subtlety than that was just too difficult. But now with AI, we could like totally do it. Like there's no reason why we couldn't grade financial statements or internal controls and like objectively rate them against each other, like to compare the quality of the financial information.
David Leary: [00:14:56] And, and that kind of makes sense. So I have an article. Well, two things kind of related, but so and this happened a little while ago, but they're sued. They got sued for negligent audits, specifically on the collapse of a company called Platinum Partners, a hedge fund where the founders of the hedge fund, the executives got charged with a $1 billion fraud. But BDO, obviously, because you said this, audits either yes or no, right? They probably just said yes. But now the investors have sued and they, they, they, uh, they sued about negligence audits for 2012 through 2013, and they won a $9 million award. Award via arbitration panel. Now, in this year, in January, a judge upheld the ruling and expanded investor rights to sue auditors. And it's all based on this concept of near privity. And essentially that means that you can have a relationship that is almost like a direct contract without actually having a contract. And basically what that would mean is this is because the auditor understands that the customer's of their data are the other investors. Essentially, they, in a way, have a contract with investors, which could open the door to lots of people suing auditors.
Blake Oliver: [00:16:10] Because currently the situation is that only the, uh, the like indirect investors cannot sue. Only the direct investors. So in this case, it's like there was a hedge fund that invested in this company. Company did fraud. Bdo missed it in the past. Only the hedge fund could sue BDO. But the hedge fund doesn't exist anymore. And so it's the individual investors in that hedge fund that are now suing BDO. And the court said they can. And that is going to open up. If it if that stands right, that could open up a lot of lawsuits.
David Leary: [00:16:49] But this goes back.
Blake Oliver: [00:16:50] To.
David Leary: [00:16:51] Your argument of this probabilistic thing, because BDO is only choice is to say yes or no.
Blake Oliver: [00:16:57] And that's the issue, is that if you only can say yes or no, then some of the yeses are going to have fraud.
David Leary: [00:17:05] Yeah.
Blake Oliver: [00:17:06] It's impossible to, right? It's impossible. And some of the S's are going to have like material misstatements that you missed.
David Leary: [00:17:14] Because all yeses are equal. But in reality, not all audits are equal, but all know.
Blake Oliver: [00:17:19] Yeah, there's greater or lesser risk. And it's sort of like the, the, the auditor Establishes this like floor threshold for risk and says every every set of financial statements above. This is a yes and everything below that is a no. And the problem is then that the risk can actually be like unacceptable to investors in the yes column.
David Leary: [00:17:49] But they don't know unless they dig in and basically audit themselves. You'll never know.
Blake Oliver: [00:17:53] The auditor doesn't tell anyone about all this stuff or like they. So it's like, yeah, if we did gradations in audits or, you know, like if we assign a probability, like, why can't the audit report say there is a 92% chance that these financial statements are correct, right? Like within materiality thresholds or whatever it is, like, why can't we give it a score 100%, 90%, 80%, 70%, whatever it is. It's just a thought. I mean, like, we aren't even thinking about this as a profession because I guarantee you that the AI agents that these investors are using are doing this. I can do this right now. I can put in a set of financial statements and a general ledger, and I can ask my AI to do like a risk analysis and give me like a confidence as a percentage.
David Leary: [00:18:50] And this is already being done. So I don't know if you saw super microcomputer was back in the news this week. And I'm going to rewind. We talked about super microcomputer on episode 400 back in September of 2024. And at that time, the story was this your favorite short seller? Hindenburg was questioning their accounting. And in parallel, around the same time, an AI startup was analyzing 800,000 FCC filings at the time, and they placed super microcomputer in the top 100 companies at the highest risk of fraud. Right? So, so this is basically what you're saying to do. Like companies are using AI to figure this out. And in the meantime, obviously the whole time Deloitte was an auditor and signed off for 20 years on all their stuff. But let's rewind. Long story short, to recap it. So in 2018, they were temporarily delisted from Nasdaq due to failing to file financial statements. August 2020, the. They were charged by the SEC for widespread accounting violations, inflating sales earnings, profit margins. They settled with the SEC for 17.5 million. Three months after that, they rehired all the old executives and brought them back in. Right. And so Hindenburg started this investigation to figure out the short sales. And what they discovered is there's a lot of brothers and family members working for suppliers and customers, you know, channel stuffing, selling things back and forth.
Blake Oliver: [00:20:09] Sounds like Carvana.
David Leary: [00:20:10] Yeah. Moving transactions through Russian, Russian and Iran and Turkish shell companies. Oh, wow. Really, really questionable, right? Yeah. But and during that show, we, we defined them as a recidivist. And essentially, that's a convicted criminal who reoffends especially repeatedly. Well, guess what, Blake. Guess why they're in the news this week.
Blake Oliver: [00:20:28] What happened?
David Leary: [00:20:30] One of the co-founders was named in a lawsuit by the US government. Uh, because they were created some backdoor scheme to divert Nvidia chips because you're not allowed to import Nvidia AI chips to China. And so they were they figured out a way to like import chips illegally into China essentially. So again, another crime. Like they can't stop committing crimes. But this goes like you said, Deloitte could only say yes or no. It would have been nice if Deloitte for 20 years as the writer was like, yeah, it's 68%. Yes, 58%. You can actually trend it over time to see if the company was getting better.
Blake Oliver: [00:21:04] Yeah. Why can't we do that? Right? Why can't we? Why not? Like as a profession, we could if we decided to like create this standard. Um, and another story I hold on. Boring accountant says anyone, any client or business can pay more for a higher level audit beyond the pass fail. But almost no one wants to pay the extra money or pay for the service. Well, maybe they don't need to pay the extra money, right? We can use all this tech to actually provide greater value as a process.
David Leary: [00:21:32] Goes back to who's paying for the audit versus who's the customer of the audit, the customer of the audits, in theory, the investors, but it's paid for.
Blake Oliver: [00:21:39] Theory, but it's the audit committee of the company, right? And then the actual like interface with the auditors is management. So the auditors like their customer. Yeah. Is the company they're auditing. So why would the company that's getting audited want to go from pass fail to a letter grade when they are very confident they're going to pass? They would never take that option. It's a why do that? Why would anyone do that? So, um.
David Leary: [00:22:15] Which is why I'm kind of somewhat okay with. We talked about the story last week where possibly some people. Kpmg employees with inside knowledge are betting on, on the prediction markets on these stocks because they have inside knowledge. Like I've said this before on the show years ago that, you know, company accounting firms should be able to do short sales on companies they audit. They should be able to make some insider trading, because then at least we would know.
Blake Oliver: [00:22:41] It's a crazy idea. But you're saying that like, it would give them incentive to find the dirt.
David Leary: [00:22:47] It would give them incentive to find the dirt, but then everybody else would be able to judge the quality of that audit, right? Like, because they're putting their money where their mouth is to some extent.
Blake Oliver: [00:22:56] I like that, I like that idea, auditors putting their money where their mouth is. All right. I want to continue on this thread, David. But first, let's thank our next sponsor and that is UNK. Let's face it, the job market, it's especially tough right now, but every industry needs. Accountants and accountants are always in demand. In fact, employment for accountants is projected to grow 10% through 2026 faster than most other professions. That's where uncW Kenan-flagler Master of Accounting program comes in. It's one of the top ranked Macc programs in the country, with 98% of students accepting a job offer within three months of graduation and earning more than those with just a bachelor's degree. If you're currently working full time, raising children, serving in the armed forces, or living halfway around the world, they're highly flexible. Mac program can also fit your lifestyle. You can choose their 12 month on campus program or their online only option, where you have up to 36 months to complete your degree. Plus, you'll join the powerful 46,000 strong UNC Kenan-flagler alumni network connections that will serve you throughout your career. If you want to set yourself up for a lifelong career, pick the Mac program with proven ROI. To see why you should get your Master's in Accounting at the UNC Kenan-flagler Business School, head over to The Accounting Podcast dot promo one. That's Accounting Today dot promo forward slash. You see. I did a live stream David this week with puzzle CEO Sasha Orloff, and he taught me something about AI that I did not know before that I really want to share with you. Yes. And it has to do with this concept of probabilistic versus deterministic systems. So we were just talking about.
David Leary: [00:24:42] This teeny example of what a deterministic system would be. Would that be bank rules.
Blake Oliver: [00:24:46] Let's let's use an example that Sasha used. Let's use the analogy that he used on the live stream, which I love. So I go to my 11 year old son and I say, clean your room. The outcome of that request is probabilistic. There is some chance that he will clean his room. There is some chance that he will not. And there is some chance that he will clean his room. To a certain like specification, but not what I was expecting. There's a probability that he will perform the task anywhere from 0 to 100%. It's not yes or no, and that is what it's like when you prompt a large language model, a chatbot like ChatGPT, that's what it is. You get, you have a probability of success. And what changes that probability of success? Well, actually, before we talk about that, let's talk about the problem with that. Right? So the problem with this in accounting and finance is that we need deterministic outcomes. We need to know when we make a request that it was either completed. 100% or not. So how do you get that deterministic outcome, the yes no out of a system that is inherently probabilistic? Because the nature of LLMs is that you don't always get the same response, right? Put in the exact same question and you will get a different response over and over and over again.
David Leary: [00:26:30] Always been my biggest issue with them.
Blake Oliver: [00:26:32] Yeah, yeah. And that's because it's, it's essentially statistical. It's based on.
David Leary: [00:26:37] Yahtzee. We're playing.
Blake Oliver: [00:26:38] Probability. We're rolling dice, right? And so how do you get deterministic results, which is what we need to be confident in the outcome. We need to be able to say, yes, this was done. No, this was not yes, this was done. And we have to check all those boxes in order to do our job and provide assurance. So at puzzle they've figured out how to do this. So he to go back to the analogy, right, I asked my son, uh, clean your room. And then, you know, he goes and does it, and then I go inspect the room and then it's like, not clean, right? Maybe he got it like halfway or 20% or whatever. Uh, okay, let's imagine the other possibility. Let's say, let's say my son is a, is a, an AI agent that puzzle is working on. So I ask him, uh, I tell him, go clean your room. Then my son says, before I clean the room, may I ask you a few clarifying questions to ensure that I know exactly what you want me to do? And I say, okay, great. Go ahead. Like, this would be fantastic, wouldn't it? Like, if only we could get our kids to do that. And then he says, all right, so you know, when you say clean your room, like, what is that? Does that mean like, what does that mean? I said, well, okay, it means like, I want everything off the floor I want your bed made. I want your clothes folded and put away in the dresser. And maybe there's some more, right. But those are like three things. Like if you do those three things, then your room is clean. And so then my little, you know, child AI says, okay, great. Um, now like when you say fold the clothes and put them away, like, do you have any like particular way you want that done?
David Leary: [00:28:31] Yeah, it's one drawer. Okay. Or should I separate the shirts in a drawer and pants in a drawer?
Blake Oliver: [00:28:35] Yeah. And, you know, you could go into more and more detail, right? At some point, it stops asking the child stops your child. At some point, my son stops asking questions and goes and does the work. But what he's doing is he is making a deterministic checklist, which is pass fail. So removing all the clothes from the floor, that is a yes no. An AI can actually look at a picture of the floor and decide, is this floor you know, clean or not? With high confidence. Now it can't determine that with 100% confidence but can do a really high right. So it's taking this this like vague request and it's creating basically a checklist of smaller items that can be that it can decide yes, no with high confidence. So closer in the, in the folded. Yes. That's, you know, we can see if they're folded in the drawer. Uh, bed made. Right. Maybe we break it down further. Right. What does it mean for the bed to be made? Is it enough for just like the the blanket to be thrown over the bed, or does the sheet have to be pulled up? Right. Do the pillows have to be arranged nicely parallel.
David Leary: [00:29:52] You've brought up all the time where it's like you have to have the processes first and you think about every process. Every step in a process is a. Did you do this step? Yes. Did you not do this step? No. Yeah.
Blake Oliver: [00:30:04] So it's yeah. It's amazing you notice that. David. Right. Because like what, what, what you need for the AI agent to be successful is exactly the same thing that you need for a human to be successful.
David Leary: [00:30:16] And you've been on this kick for six months, a year. Yeah. Well, process.
Blake Oliver: [00:30:20] For years, right? Because what I learned when I built my firm was that if I wanted, say, a bookkeeper I hired to be successful, I couldn't go just tell them, go to the books. I had to give them a detailed checklist with instructions for dozens and dozens of tasks, with subtasks, with instructions that they could follow to do this according to my specification. And so what puzzles done and what I think all these AI developers are doing, Claude has been doing this with Cowork, right, is they're teaching the AI to go and make that checklist before it does the task. And then it asks you questions to make sure it has the checklist right. And then it works through the checklist.
David Leary: [00:31:05] I've noticed that you say something and it pauses. It's like it tries to digest what is he really asking me? And then ask some clarifying questions and it goes and does.
Blake Oliver: [00:31:12] And so this is how you can get a deterministic answer or outcome out of a probabilistic system. It's the same way you get a deterministic outcome from humans, which humans are probabilistic. This is the thing that's so mind bending about AI is that LLMs, they fundamentally work the same way. Our brains work with one key difference. But this part is the same. So and that one key difference, which is kind of like, well, it is relevant. The one key difference in my view is, uh, LLMs get trained once and they are fixed. So it's like, imagine if if like you took a brain and you just froze it in time, you took like an image of it all the neurons and everything, and you just froze it and you could like, talk to it, right? But it doesn't change. You could like give it stimulus and it gives something back, but it doesn't change the difference between. The difference with us is that our brains are constantly updating as we receive stimuli, as we receive, as we hear words and see images and all that. Right? So our model is updating in real time and LLMs don't. That's the difference. That's the only difference. So the good news for us is that actually, as far as I know, nobody has figured out how to get these LLMs to update in real time because they're like these giant data sets of numbers, right? They're not flexible, like neurons are like neural pathways can like, uh, remap biologically, but that's why you can't, you.
David Leary: [00:32:53] Can't.
Blake Oliver: [00:32:54] Change this giant spreadsheet, which is essentially that's what it is, like a giant data set. That's what the NLM is. You can't it can't change. Yeah.
David Leary: [00:33:01] So, so, so I guess in summary, here is the way you're saying puzzles attack. This is it's really just checklists, which essentially you could argue a checklist is a set of rules.
Blake Oliver: [00:33:12] It's a checklist. And then they created, they're creating functions. They call them functions. And the functions are all the different things that the AI can do in the app. So like to modify a transaction or to post a transaction. They're like, he didn't say it this way, but it seems to me sort of like API access, right? For the AI. Okay. So like you can basically say, like in the example he used in the, in the live stream was, um, a prompt that was like, go through last month's transactions and identify anything over $2,000 that looks like it should be capitalized as computer hardware on the fixed assets, and then also create the depreciation, like put it into service and do the create the depreciation schedule and book it. And the AI did that. It went and found the transactions and it did it in steps, right? It asked what we wanted it to do. It went and found the transactions, and it would pause each time it was going to do something and ask clarifying questions until it was confident it knew what the tasks were supposed to be. And so then it identified the transactions. Uh, asked about what the depreciation schedule should be, it investigated what it should be based on past transactions.
Blake Oliver: [00:34:29] Then with our approval, booked it into the fixed assets register and made all these entries right. And like it did that all like away a human would asking questions. So anyway, they've got this new like AI close checklist product that you can, they're like, it's in alpha and it's like, that is to me, that is what all the successful apps are going to do. When it comes to AI agents is you have to like give them the tools in the software so it can do this stuff. And then you have to build the guardrails around how it uses those tools. And for AI in accounting, it needs to like be a really high threshold of confidence, like a like very close, like close to 100%. And if it does that and it breaks down everything into smaller tasks, I don't see why you can't have AI agents running around all the time doing this work. And then we get 100%. We get to near 100% confidence. Like we don't have to be 100%, but it could be like 99.999999%, just like with uptime for like software, right? To the point where it is 100% effectively.
David Leary: [00:35:37] Yeah. Or it's, it's not 100% effective, but because it's running 24 hours a day, we're doing work that's, that's the, the or.
Blake Oliver: [00:35:46] Actually, and the standard should be more reliable than humans because humans are not 100% reliable right now.
David Leary: [00:35:53] Yeah.
Blake Oliver: [00:35:55] And for certain tasks, that's like really easy to achieve already. Okay. Uh, I've taken up a lot of time talking about this, but I hope that was helpful for our listeners because I just, I find that fascinating. It's also a way, I think, to think about how to use AI in the work you're doing outside of all these apps. Like, um, you can, you could actually like create a prompt where you ask, like actually one of my prompts that I love using is I think I shared it on the show before, right? It's where I just add it to the end of every prompt. Uh, when I want it to like, really think and I say, ask me questions one at a time until you have enough context to do this job, like at 100% confidence and like, that's what these newer agents are doing on their own without having to tell them that. But you can already get that result by putting that into like ChatGPT or copilot or whatever. Okay, wait, before we move on, let's thank our next sponsor. Uh.
David Leary: [00:36:53] It's UNC again.
Blake Oliver: [00:36:54] It's UNC again. Okay, great. Uh, earlier we talked about how strong the demand is for accountants. And actually, I think it will be even with all this AI stuff. We'll talk about that. So let's talk about what actually separates you in that market. A master's degree or even better, a master's in accounting, or even better, a master's in accounting from UNC. Unc Kenan-flagler. Master of accounting program isn't just highly ranked, it delivers results. 98% of students land a job within three months of graduation, and they typically earn more than candidates who only have a bachelor's degree. That is a real return on investment. And here's what makes it practical flexibility. You can complete the program in as little as 12 months on campus or if life is busy. And let's be honest, it usually is. You can go fully online and take up to 36 months to finish. It's designed for working professionals, parents, military service members and students around the world. You're also stepping into a 46,000 strong alumni network from UNC Kenan-flagler connections that can open doors long after graduation. If accounting demand is strong and it is, the question becomes, how do you position yourself to take advantage of it? Uncw Kenan-flagler is master of accounting can help you do exactly that. To learn more about the Macc program at UNC Kenan-flagler, head over to The Accounting Podcast. That's The Accounting Podcast forward slash u n. Cwc. Oh, I talked over you, David. Go for.
David Leary: [00:38:20] It. I was going to pivot into AI taking jobs off your AI stuff, but do the PUC because I think we might get an argument or differing opinions on these next couple stories. So do the PWI.
Blake Oliver: [00:38:31] You know, we haven't talked about the big four in a little while. According to The Guardian, PwC US CEO Paul Griggs said that partners and senior staff who do not fully adopt artificial intelligence risk being pushed out of the firm. Anyone not being paranoid about being AI first is likely to be replaced by people who are more willing to embrace the technology. So yeah, don't use AI. You are going to get replaced. And what's the context for this? Well, PwC is still hiring overall, but the talent that they're hiring has shifted. So this may be of interest to those of you who are trying to figure out what to do with your careers in accounting. They're not recruiting the same proportion of traditional accountants and consultants. They are increasing their hiring of data specialists and engineers. Uh, PwC actually reduced headcount by 5600 employees last year, bringing in its global workforce to under 365,000. That's a lot. This is something that's interesting that Greg said. So he said that some tax and consulting services are going to be converted into AI enabled automated tools, and that these offerings may be sold through annual subscriptions, rather than the traditional consulting model of billing by hours worked. Some services could even be delivered without a PwC professional directly involved. An example of this is called PwC one. Pwc one is the name of it. It's an AI platform with six automated client services, and one example cited in the article is an anomaly detector that is designed to identify problems in a company's sustainability data. So they're building these AI tools for stuff they've traditionally done with consultants and auditors and accountants, and they're going to sell it directly to clients without people doing it. And I wonder about this, like, Is this really what the professional services firms, the biggest ones, want to be doing because they're not software companies. That's not their skill set, right? They're not. That's not what people hire them for. They're not very good at it. Right.
David Leary: [00:40:59] Well, why would you pay the big four to use AI to give you some advice, when you could just use the AI yourself to get the advice, doesn't it? Like, I don't think there's a business model of there being a premium. Oh, it's because you put your logo on this AI, it's all the same shit.
Blake Oliver: [00:41:14] Exactly. So like this could actually be the exact wrong thing to do. I would be if I'm PwC, right? If I'm the big four and I have that brand I'm leaning into, not that this is automated because I can go get automated from anyone. I can do it myself. Like you said, I can just use Claude coworker, whatever enterprise thing they want to sell me to do this. What I'm paying for is your assurance that it's right.
David Leary: [00:41:43] So I don't know if you saw this article came out. So E seems to be taking possibly a different path than the other big three, the other big four. So E is now doubling the CPA exam bonus. Now, when you got your CPA, were you on your own firm then or were you working for another firm?
Blake Oliver: [00:41:59] I was, I was, I was at Armandino.
David Leary: [00:42:03] Did you get a bonus for passing it?
Blake Oliver: [00:42:05] I, I did, but I had to fight for it. They didn't want how much was.
David Leary: [00:42:10] It just for our listeners? So we understand.
Blake Oliver: [00:42:12] Oh, I think it was like a few thousand bucks.
David Leary: [00:42:13] Okay. Well, E is doubling its CPA exam bonus to $10,000 now 10,000.
Blake Oliver: [00:42:19] Wow.
David Leary: [00:42:19] If you pass the CPA exam and they're doing this is because they're trying to attract more talent. So people they're trying to hire more bodies. And what I thought was interesting about this, we've covered, you know, how all the big four are committing billions to spend on AI. Deloitte. In 2023, they announced they're going to spend 1.4 billion on AI. Pwc said they were going to spend $1 billion over three years on AI. Right. Kpmg, we just talked about they're going to commit $2 billion to AI and cloud, right? And it's all 2023 announcements. E AI announced in 2024 they were going to invest in tech and talent. Not really specifically AI. So E might be taking a slightly different strategy than the other three.
Blake Oliver: [00:43:02] All right.
David Leary: [00:43:03] And proof of that is they're putting money into people hiring.
Blake Oliver: [00:43:06] Mhm.
David Leary: [00:43:08] And that's my thing is like, what do you believe? There's two paths. You either believe AI is taking accounting jobs or not. I think there's a big division.
Blake Oliver: [00:43:16] I think it's complicated. I think it's complicated. And David, you have a chart here that you're going to share that I think shows this why this is complicated.
David Leary: [00:43:24] Yeah. Let me open.
Blake Oliver: [00:43:24] It's not so simple.
David Leary: [00:43:26] So so.
Blake Oliver: [00:43:27] This chart from.
David Leary: [00:43:28] Claude.
Blake Oliver: [00:43:29] Here. I'll put this on the stage for you. Oh, there we go. Oh, sorry. It's on the stage now. Okay, so. So. Yeah. You. It's a study from Plaid. I saw this, too. And there's this amazing chart here on the screen. But tell us about the study first.
David Leary: [00:43:42] Yeah. So, um, anthropic, who makes Claude, right? This is their AI exposure index. And we have on the screen, it's one of those, uh, circle graphs with 0% chance in the middle, out to 100% on the outsides of the circle. And if you follow it around, it has all these different professions. And I forgot the exact name of what kind of chart this is.
Blake Oliver: [00:44:00] But it's like a surface area chart, right? So it's showing like the surface area of exposure for each job.
David Leary: [00:44:07] And it.
Blake Oliver: [00:44:07] Has.
David Leary: [00:44:07] A theoretical AI coverage. So, you know, people are projecting all business and finance jobs and computer and math jobs are all going away. They have the hugest, uh.
Blake Oliver: [00:44:16] Hold on, hold on though. It's not all though, right? It goes, it goes out to almost to the edge.
David Leary: [00:44:21] Almost to the edge, right.
Blake Oliver: [00:44:22] But there's an empty space.
David Leary: [00:44:23] Theoretical coverage versus and then they have the observed AI coverage of where they think they're at now. And to put a. Uh, like business finance is almost at the edge, maybe 98% to the edge in the theoretical, um, grounds maintenance. So basically landscapers is at zero. So, and then other industries like, um, education and libraries at about that's the 60% one um, sales is about 60%. Construction is about 25%, 30%. So these, yeah, all the industries have different projections.
Blake Oliver: [00:44:58] And they've kind of grouped the profession. So you see management, business and finance, computer and math, architecture and engineering, life and social sciences. Those areas are like really covered by the potential automation and then protective service, like police food and serving grounds, maintenance, personal care, right? All that is like very low coverage. And then the observed AI coverage is the red surface area, which is actually very low for all the professions Still, but like you can see, it's like it's moving up in business and finance towards the theoretical. It's, it's moving in that direction. Yeah. Yes. Right.
David Leary: [00:45:37] But still pretty low in the grand scheme.
Blake Oliver: [00:45:39] It's still very low. It's still like, you know, well, it's like 40% in computers, like so. And that makes sense. You see all these memes of like computer programmers are just using cloud code to like generate all their code now. So it's like half of their 40% of their job is now automated, but they've still got to do 60% like the code review and cleaning it up and all that stuff. Right? Um.
David Leary: [00:46:01] And so this is like some real data and some piece of it. They're saying that there's little evidence of widespread job loss so far. And I don't know if you saw there was another article that came out. The Microsoft's AI CEO is predicting that within 18 months, uh, all white collar jobs could be all could be eliminated. Right?
Blake Oliver: [00:46:23] Who said.
David Leary: [00:46:23] That? He's the AI CEO at Microsoft. And he warned.
Blake Oliver: [00:46:28] Within how many months?
David Leary: [00:46:29] 18 months? 12 months to 18 months.
Blake Oliver: [00:46:31] The guy's out of his mind. To me, this guy is out of his mind.
David Leary: [00:46:33] Yeah.
Blake Oliver: [00:46:34] He will tell you why.
David Leary: [00:46:35] Yeah. Warned that jobs involving sitting down at a computer, including accounting, legal, work, marketing and project management are especially vulnerable. Right? And then anthropic CEO in 2025 said that half of entry level white collar jobs will be eliminated. Ford CEO he predicted in 2025 that 50% of what US white collar roles would be eliminated. So there's this this fear that's happening, right? These these headlines come out. Yeah. I saw another article about how a survey of 250, 250 UK mid-market mid-market finance leaders in the UK found that only 14% believe traditional accounting skills will be critical for the next generation. Okay, so but these headlines have real results because now there's a survey of AI, of AI, a survey of finance professionals, and half think that AI is now hurting their job prospects and it's headlines like Accounting Today.
Blake Oliver: [00:47:28] Stuff like that. Right? Because you've got these CEOs out there, like what was the headline in Accounting Today?
David Leary: [00:47:33] Accounting Today made the headline accounting and tax staff worry AI threatens jobs. Yeah, because there's just like this scary thing that's happening everywhere.
Blake Oliver: [00:47:40] I understand why you look at this chart and you and you look at what these CEOs are saying, and it sounds really bad, but the CEOs, the AI people are missing something, right? Which is that just because a certain percentage of the job can be automated doesn't mean the entire job can be automated. The part that can't be automated still has to get done by somebody. And you can't necessarily just have somebody do only that percentage of the job.
David Leary: [00:48:12] So my argument is if half of all these jobs are gonna be gone in a year, shouldn't at least one accountant at this point have lost their job so far because of this? At least one and I. So I have a LinkedIn post that's 10,000 views, and I'm asking for people to put the name of one accountant that they know have lost their jobs because a company implemented some AI and now they don't need that account anymore. Zero names, Blake, zero names. It's been weeks zero names in this thread because I don't think it's true.
Blake Oliver: [00:48:41] Well, well, okay, so I'm thinking about these like AI agents, right? The puzzle is building and like that the general purpose ones like Claude Cowork and I'm thinking, okay, like, you know, updating depreciation schedules, rolling forward work papers, reconciling bank accounts, like all that stuff will be done by AI in fairly short order. And that is the work of a lot of like staff accountants and even yeah, I'm not going to say managers because managers have other responsibilities, which is like managing work, right? Like if that's, if that's all you're doing, Then yeah, the agent will do it faster than you and better than you and cheaper than you.
David Leary: [00:49:26] So there was a you know, I talk about.
Blake Oliver: [00:49:30] But I'm not done. This is the good part, right? The good part is that it's not going to do the manager job. It's not going to do the the CFO job because that maybe. Well, that's the part that's the piece of the chart, David, that it can't cover. Do you understand? Yeah. So like, let's say it does 90%. Right. Think about an accounting firm. A traditional accounting firm might have a, you know, 10 to 1 staff to partner ratio, like a small firm that's well run. So 10% of the work in that unit of people is the partner that can't be automated. The AI is not going to do partner level work.
David Leary: [00:50:18] Ever so full circle here. Are you familiar with the Cohen Paradox? C o w a n.
Blake Oliver: [00:50:28] C o w a n.
David Leary: [00:50:29] Cohen Cohen Paradox essentially came out in 1983. Labor saving technology, let's say, like vacuum cleaners, didn't actually reduce chores or housework. Essentially, housework stated like 51 hours a week, but the standards increased. So your, your, your carpet.
Blake Oliver: [00:50:46] Cleaner got.
David Leary: [00:50:46] Cleaner. You still were doing all this work, right? And what's happening is, and this is an article and a lot of it was from a talk from I talked about SaaStr all the time because they're building all these AI agents and they had some examples of this. So for their sales team research, time dropped from two hours to about 15 minutes. So maybe you're doing research before a sales call, right? So basically they're not working less. They're just going to handle more pipeline. They're going to take more calls. Nobody's actually going to work less. Right?
Blake Oliver: [00:51:15] Well I am.
David Leary: [00:51:17] You hope right. And because essentially what's happening is as people become more productive, the baseline increases, right? So let's get rid of AI. And if I don't know if you ever you work at a big company corporation, if you're an employee, you're expected to be an, A employee every time at bat over and over, just like you're a baseball player, you're expected to hit the home run every single time. It creates this pressure because the bar just keeps going up and up and up. So the secret to life, kids that are listening B a C employee forever at a corporation and you will do perfect. That's what you want. You want to hang in the C area because nobody expects you to do more. You get you get a paycheck. That's probably 90% of what the A person gets, like B, a C employee at a corporation. That's a secret to this.
Blake Oliver: [00:51:56] David. That is the worst advice I have ever heard in my entire life. I'm ashamed.
David Leary: [00:52:05] You're ashamed, but.
Blake Oliver: [00:52:06] I'm ashamed of you.
David Leary: [00:52:07] Like it. Just.
Blake Oliver: [00:52:08] And you know what's funny is that, like, you could never do that David person. Never do.
David Leary: [00:52:12] I don't have the personality to do it. I I can't do it. Yeah.
Blake Oliver: [00:52:14] Yeah. You wish you could, but I wish I could.
David Leary: [00:52:17] That would be, I wish I.
Blake Oliver: [00:52:18] I would love to collect a big fat paycheck at a big corporation and get below.
David Leary: [00:52:22] The.
Blake Oliver: [00:52:22] Radar and just like, never, never accomplish anything or, you know, generate anything of value in my entire life and just enjoy it. Why not?
David Leary: [00:52:31] Yeah. So, so basically AI is going to expand work, right? Um.
Blake Oliver: [00:52:38] So, so okay, on that point, AI is going to expand work. I have another way of thinking about it. So for me personally, AI has, it has given me a promotion. So I left public accounting as a manager. I believe if I went back in time and gave my manager self access to all these tools, I could be doing partner work instantly. Like well, after I learned how to use the tools, right? So. It is A for those who use it, for those who learn to use it, it it up skills you. And I think it does it as much as like two levels. So like in a small firm, right, you're just going to go manager to partner. But even in a big firm, I think it could like level you up from manager to senior manager to director. Like like that is the it's not just a productivity boost, it's a knowledge boost and like capability boost and skill boost because it gives you all this knowledge and skill that you would have had to accumulate over the course of like ten years to go from manager to partner or whatever it is.
David Leary: [00:53:52] Article that's going to completely reinforce your point on this. Like, um, there's an article about two weeks ago on yahoo, a record number of companies are replacing older CEOs for younger ones with less experience. So how do you think that younger CEO is making up the gap.
Blake Oliver: [00:54:09] Ai.
David Leary: [00:54:10] They're using AI to supplement their skill set because what companies want. They kind of want they don't want a long tenured executive who hasn't, you know, been on the front lines in a long time and is out of touch. They want somebody that still has that frontline experience, talk to customers. They felt the customer pain, but now they need to lead the company. And so they're hiring these younger executives. Um, and it's a trend that we're seeing. So 1 in 9 CEOs at the largest 1500 US public companies were replaced in 2025. It's the highest replacement rate since 2010. And the average age of new CEOs fell to 54, down from 56 a year earlier. So you're just seeing younger CEOs because they're supplementing their skill set with the tools that we have available to us.
Blake Oliver: [00:54:52] So here's my takeaway from all this. It's a great time to be in accounting. If you use the tools, if you learn how to use the AI and you upskill yourself fast enough to outpace the automation, that is going to happen over the next few years at the staff level. If you can outpace it. If you can surf ahead of the wave, then the opportunity is almost infinite. Like it's crazy good.
David Leary: [00:55:18] Yeah, I do have data on who is actually losing their jobs because of AI. And so this is data from ramp. So ramp has because.
Blake Oliver: [00:55:27] You said no accountants are losing their jobs.
David Leary: [00:55:29] No accountants, no accountants. Right. Okay. But so ramp has their economics lab where, you know, because ramp sees everybody's credit card spends across thousands of companies, right. And thousands of spends and what their and watching tracks firm spending level between 2021 to 2025 companies are shifting dollars away from freelance marketplaces. So you think your upwork's your fiverr's towards AI providers so people are spending less on freelance workers and more on AI solutions.
Blake Oliver: [00:55:59] That makes sense.
David Leary: [00:56:00] Freelancers are going to get impacted, especially since overseas ones that are like 20 bucks. Yeah. For a job.
Blake Oliver: [00:56:06] You think about what is Fiverr really? It's like I put in a prompt with a bunch of like, contacts and I get a result back. It's just it's a human powered like chatbot.
David Leary: [00:56:16] That's a good that's a really good analogy. Think about it that way. Yeah. So there's real data. Like that's who's losing their jobs is people aren't spending money on freelancers.
Blake Oliver: [00:56:24] David. So great talking with you as always. I have so much fun. Thank you to all our livestream viewers. Heather, great to see you. Heather said, I'm looking at your chats here. News from OZY last week was the Accounting and Business Expo in Sydney. It'll be the biggest accounting conference in Australia this year. I enjoyed myself 30 to 40% of vendors were based on or attached to an LLM to work. It seemed to me, with the speed of AI development, that many of the exhibitors section solution would be replaced by the very AI that they were using, for instance, the tax research LMS. Yeah, we talked about that, David, on the show, like these proprietary tax research solutions that cost like a thousand or more dollars a year. I mean, eventually we'll just the AI.
David Leary: [00:57:12] It's just out of the box. You just get it off the shelf.
Blake Oliver: [00:57:14] Yeah. Because like, they can just ingest all that information.
David Leary: [00:57:19] And Heather, I saw you were putting all these great pictures on social media. I saw all the other accounting celebrities that are down under all at the show with you look like a lot of fun. Next time, you gotta get Blake and I invited.
Blake Oliver: [00:57:30] Giles says, hi, Blake and David, this is Giles Pearson. Just been to the Accounting and Business Expo in Sydney. The words private equity not heard even once know private equity there. Um, Alison says AI CEO feels like a pizza CEO saying burgers will be wiped out of restaurants in 24 months as people expected to move to calzones. I'm sure there will be jobless, but not what they project. Yeah, they're trying to keep the hype cycle going, right, to keep their investor investments coming right. They need to. They need the cash to buy from investors to fund all this, like enormous spending they're doing on data centers. Okay. Don't forget, you can earn free continuing professional education for listening to this episode and our past episodes and many, many fine accounting and tax podcasts. Get the free earmark app, earmark app, and your web browser, or download earmark CPE on the App Store or Google Play Store. It is totally free to download, free to create your account, and free to earn one CPE every week. If you're a CPA, get your Naspa approved CPE. If you're an IRS enrolled agent, get your IRS CE thousands of courses available on the app. What are you waiting for? Earn CPE anytime, anywhere. David, great seeing you. Have a good week. Bye, everyone.
