The Accounting Podcast

Attention: This is a machine-generated transcript. As such, there may be spelling, grammar, and accuracy errors throughout. Thank you for your understanding!

David Leary: [00:00:04] Pwc is planning to recruit one third less grads in 2028. And it's because of this. Of automation. They truly believe that, um, between AI integrations and their offshoring hubs, there's going to hire less in the US. So one third less grads is a lot. And then this is probably if PwC is doing it, that means all Big Four has a slide like this that they're so across the board. Big Four is not going to recruit as many college grads coming to you weekly from the OnPay Recording Studio.

Blake Oliver: [00:00:39] Hello and welcome back to The Accounting Podcast, your weekly roundup of news in the accounting profession. I'm Blake Oliver.

David Leary: [00:00:46] I'm David Leary and Blake. I've been waiting a whole week. Is your AI agent done categorizing those six transactions from the bank feed yet?

Blake Oliver: [00:00:52] Oh, it did it last week. It just took 40 minutes to do it, and I posted a video taking all the little clips from the episode, putting them all together, showing the whole demo. And I put that on LinkedIn and that has been really popular. And Chad Davis pointed something out in the demo. One of the transactions was a deposit from Smart Agency for $4,500. This is in the Xero demo file, and what you're supposed to do is use Find and Match and select both invoices to match the two invoices to the single deposit. Oh, is.

David Leary: [00:01:27] It one to many?

Blake Oliver: [00:01:29] Yes, and I forgot about that. And so when the agent came back to me and said, hey, what do you want to do with this transaction? I couldn't find a match. I just said, code it to consulting fees or something like that. So we just made it a receive money transaction, but it needs to be matched. And Chad Davis pointed out, hey, uh, a smart bookkeeper would have said, you know, do you want to match these? It would have figured that out. And so I thought, how could I get ChatGPT agent to, like, do that intelligently. And I thought, let me try making a project. So a project in ChatGPT is a way that where you can add custom instructions that the agent uses, or really any prompt uses every single time you run a prompt. So I added some very simple instructions to a project here. I've got this on the screen and it says you are an accountant proficient in zero. How to reconcile in zero. If zero suggests a match and the information about the bank transaction fits the suggestion, such as the vendor an invoice number. Click okay to match the transaction. If zero does not suggest a match, follow this process. Use the find and Match option to search for invoices or payments to match.

Blake Oliver: [00:02:42] You may need to match a single deposit to multiple invoices or record a partial payment. Two if you can't find a transaction to match, consider creating a spend money or receive money transaction to determine how to categorize the transaction. Search for the previous transactions from the same customer or vendor to see how it was done in the past. If it still isn't clear how to reconcile the transaction, ask the user what to do. So now I've given it a process, and I'm realizing as I read these instructions that, um, I want to clarify something. I want to say if zero does not suggest a match or the match, uh, seems like an error, follow this process. So now I'm going to save this set of instructions, and I'm going to select the agent mode and prompt ChatGPT help me reconcile my zero demo company checking account. I want you to start with the smart agency transaction. So I'm going to have it go straight to that transaction and we'll see if it gets it right this time. And while it spins up the virtual desktop and gets going on that, I'm going to log it in. David, why don't we thank our sponsors.

David Leary: [00:03:47] Yeah. Our sponsors this week are relay team up human at scale and digits. Between Blake and myself we now have three, four, maybe five business entities doing 20 or so, 20 or so checking accounts, dozens and dozens of virtual cards. It would be impossible to manage all of this if we weren't using relays. Our small business bank. Really, it's truly part of our tech stack that we use to run our businesses. Really allows Blake and I to each have our own logins. We can grant access to our team and even our accountant without sharing passwords or two factor authentication codes. Relay allows us to grow and scale our banking needs without ever going into a physical branch, and I recently added an account to receive inbound merchant services with just a few clicks, and had to create a payroll checking account. Again, just a few clicks and I instantly had access to my ACH info to give to my payroll provider. With relays virtual cards, we can issue debit cards to our team around the world for needed business expenses. I can instantly spin up a new visa debit card and set both daily and monthly spending limits. And when the team member doesn't need their card, I can freeze it until they need to use it again. Relay also has automation features to sweep money automatically from one account to another based on dates, amounts, or target balances, even percentages. For example, inbound payments could be split to your daily payroll or split daily to your payroll accounts. Sales, tax payable, operating and savings account based on predefined rules. To learn more about using relay for your firm and clients, head over to The Accounting Podcast. That's The Accounting Podcast.

Blake Oliver: [00:05:16] Checking in on ChatGPT. Here it has. I have logged it in to zero, and I'm going to say finish controlling and allow the agent to take over. And we're going to see what it does next. So while it's chugging away there, we'll come back to it if anything happens. I want to talk about AI agents. I have been on the hype train as somebody reading all the news about AI agents and so far, to be honest, like until recently, I have not found them to be that useful. And like in the last few weeks, something has changed where now I'm able to use Zapier agents and I'm able to build stuff that actually works. And so I'm going to show you also after this. Something that I built with Zapier agents, I basically built in about an hour. I built Hubdoc, the core feature of of Hubdoc or auto entry, where you forward a bill into an email address and then it enters it into the accounting system. Like you can now build that with an agent very quickly. It's kind of crazy. Okay. So we've got the AI agent is working away. It's got the smart agency transaction on the screen.

Blake Oliver: [00:06:37] And I'm a little worried because it looks like it's trying to code it as an advertising expense. But don't do that. I tried this earlier and it actually worked properly. It didn't do that. I'm going to interrupt. I'm going to say wait. Check to see if there are multiple invoices to match to, and we'll see if it can figure it out. So we're still a little ways away from this working. It worked. It worked before I tried this once before and it worked. And now it seems to be struggling. So I guess now that I've told it to open, find and match, it's going to do that. But I'm not sure why it didn't do that given that I'd given it custom instructions. Okay. Oh is it doing it now? Yes it is. It's opening up the match tab and let's see if it will like filter for smart agency. Yes. It says I'm searching for invoices related to smart agency and it misclicked. It's struggling. We'll see if it gets there. David, what's top of mind for you in the world of apps and technology.

David Leary: [00:07:49] I though there's a big, uh, headline that I thought was interesting that you're not reinforcing the truth of this headline, but S&P 500 companies could save $1 trillion a year. Basically, they'll have $1 trillion a year in less cost because of AI. So Morgan Stanley's projecting that artificial intelligence could reshape corporate America by cutting nearly 1 trillion a year in cost for S&P 500 companies. The bank estimates that estimates that 90% of jobs will be impacted, either automated or augmented, primarily through Agentic AI software and humanoid robot robotics. Um, and the savings essentially would come from reduced payroll. So the S&P 500 is going to reduce those companies are going to reduce payroll because of AI that's doing this data entry like you're trying to show us right now.

Blake Oliver: [00:08:35] But we haven't seen that yet. And we saw the market decline significantly on Tuesday after an MIT study found that 95% of companies investing in generative AI report no return on their investment, 95% of all that money that is getting poured into developing AI in companies is having no impact on the bottom line. And we saw Nvidia fall 3.5%, Palantir dumped 10%, Nasdaq dropped over 1.2%. That's the steepest decline since August. So what is going on here?

David Leary: [00:09:14] The big four that these big each have invested $1 billion each. That's pretty.

Blake Oliver: [00:09:19] Scary.

David Leary: [00:09:19] News.

Blake Oliver: [00:09:20] That's very scary news for $1 billion each. Yeah I mean well so but that's good news if you're a worker, right. Is like, um, if you are like, I would not be worried at this point about AI taking my job. And I don't believe any of these reports that say that AI is automating like 80% of jobs. Like what? Jobs are getting automated by AI. At this point, it's just not good enough. We're seeing that right now on the screen with this zero, you know, agent, which is struggling still to match the deposit to two invoices. Right. It wants to create a spend money transaction. So it's not even as good as like I mean, it's probably better than your worst bookkeeper who doesn't know how to match to an invoice, but that's it, right? It can it can do the, the very, very, very easiest task. So unless your job is something that's just completely rote, like data entry, it's not getting eliminated anytime soon. And it failed. It can't do it. It didn't do it. Even with my custom instructions in the prompt, it's not reliable enough. So I found this MIT study really fascinating. And I'm curious, like, why are these AI initiatives failing? If you own an accounting firm or you run an accounting firm like you want to know if your investment is going to pay off. So maybe the study can tell us what is happening where they're not seeing a return. And MIT did some analysis of this. And there are some patterns as to why these companies aren't getting a return. There's a learning gap in poor integration. These generic tools, these out of the box chatbots, they don't learn from or adapt to company specific workflows. So you can get them through the pilot stage, but actually implementing them in a systematic way doesn't happen because everybody is using them individually. So the chatbots just rolling out ChatGPT to everybody at your company, you're not necessarily going to be able to attribute more ROI or revenue from that.

David Leary: [00:11:30] So it's interesting. So you can pilot the chatbot and kind of train it. And it's good enough to prove proof of concept. So you're like, all right, we're going to move forward. But then maybe it doesn't work. But in general, if you invest three months to train an employee on something and then say, okay, you're on your own. There's probably a high chance it's going to be successful.

Blake Oliver: [00:11:50] Potentially. But think about it this way, right? The what you just said, that first story, you said that report. Yeah. About the money that could be saved. It's all based on cutting headcount, right. These chatbots are not agents. They can't act on their own. Somebody has to copy and paste text into them. So there's no headcount being eliminated by these chatbots. It makes your existing workers more productive and more effective. But unless you have a bunch of people that are all just doing the same thing and you can make some of them more productive and then therefore fire others middle management.

David Leary: [00:12:22] Like I.

Blake Oliver: [00:12:23] Still believe. Yeah, maybe. But like, I mean, the stuff that's getting automated is not where the real value is. Like, and I'm going to show you that with some of the agents that I'm building that are like making my life easier and making me more effective as a manager and a leader of a company, but like, you couldn't replace me with that yet. So I think that's part of the problem is it's not integrated into the workflow. And that's what really has to happen to make AI really powerful in your business, to really up productivity, to eliminate the need to hire people. And that's what we've done at earmark. We have embedded AI tasks in our human workflow. So eight of the tasks might be human. And two of the tasks are AI. And we're gradually increasing more and more of the tasks to be AI tasks. So if your company doesn't have a defined workflow like a checklist process that you follow for the things you do, you can't really benefit from the AI because you can't slot it in to these tasks. And there's like so many companies that get away with having no defined workflow, including accounting firms. But I'm talking about like giant fortune 500 companies because they just throw people at the problem and they have a bunch of really smart people that work for them. And so they can't benefit from the AI because the AI agents are not capable enough to replace humans, yet they can do tasks. You got to have.

David Leary: [00:13:47] A workflow of automating things that, you know. If I go back eight years ago, ten years ago, you know, automating stuff, the bill went off for me is like, oh, you really can't automate anything until you document everything, every step. Then you automate each part of that documentation process. And it's kind of the same thing until you document it all. How do you hand it off to AI and you just prove that in your demo you had to you had to write out. Basically, I was thinking this when you were reading that, you basically gave it the standard operating procedures.

Blake Oliver: [00:14:16] Yeah. And actually I was really detailed, but I might need to be even more detailed in order to get it to be reliable every time. And most companies like, they don't they don't write this stuff down. They don't have like a manual of how to do everything. It's in people's heads, right? So until you get there, you can't benefit from it. Um, here's another fact. More than half of Gen. Gen I budgets are directed to sales and marketing, so we're investing in using Gen I to do sales better, to do marketing better. Well, the place where these tools are actually the best is the back office automation. So, you know.

David Leary: [00:14:54] In a lot of that has already been solved by other tools. Right. You've been able to scan OCR robotic process automation. Ai is not going to anybody who's going to replace staff or the back office staff has already done it with tools. It's not like there's a bunch of people still out there. And now AI comes. Now they're finally going to replace those people. Those people have already been replaced with either overseas labor or existing tools.

Blake Oliver: [00:15:19] And they're low value tasks, right? They don't add value to like the customer. It's just a cost for the business. Yeah. And generally if like you're applying AI, AI is best applied to those kind of tasks. And so you're just cutting costs. But that is way less effective than increasing revenue, right? The eye is not directly increasing revenue at these companies yet. So I think that's that's the main issue. That's my takeaway from this is it's it's like most of these tools are not making it into production. They're getting stuck at the at the pilot phase. And if you look at my dashboard for like Zapier agents or what we're building in, like any of the automation tools we use, you'll, you'll see, like this giant mess of automations that I've built that are not working, that are not in operation because I got it like 90% of the way, but not to 98 or 99%, which is what it needs to be to actually deploy it. And that last mile.

David Leary: [00:16:18] Problem, I always felt like that was a problem with Zapier and a lot of automation tools to begin with. You'd get them so close and then it can't handle the last the last little use case. You need to actually take advantage of it. I feel like so like, I feel like we're just suffering the exact same problems, but now it's just AI instead of automated tools. Robotic process automation. It's the same problem. You get 90% there, but it's the 10% that counts, right? That's what matters.

Blake Oliver: [00:16:44] And the other problem is that the last problem I'll highlight is that the tasks that AI is good at doing really reliably are tasks that are like short. And in order to automate those tasks, you need the people doing them to be on board with automating them. But there's really no reward. As an employee of a big company where you have a bunch of tasks to do to like help them automate the tasks you're doing because it feels like you're just getting rid of your own job. So there's no bottoms up grassroots effort in big companies to want to do any of this stuff. People are generally happy doing things the way they've always done them, and the bigger the company is, the more likely they are to just want to keep doing things the way they've been doing them there. There's resistance to change at the big company level. So it's startups like us that are benefiting from this because we have the mentality of we don't want to do this grunt work, we don't want to get paid just to do this task. Well, if.

David Leary: [00:17:40] You're a manager and you have a staff and now you're not going to roll AI to replace your staff because then you won't have a job, right? Because you're managing staff. And that's a that's a dilemma, right? When you have these people with castle walls of existing enterprises.

Blake Oliver: [00:17:54] Let's think our next sponsor. And then I want to show you what I've built in Zapier agents. Let's talk about team up. Are you turning away clients because you can't find qualified accountants? You're not alone. Accounting firms everywhere are struggling to find talent, even when they're willing to pay top dollar. The stress of being understaffed leads to overwork, sleepless nights, and burnout. Here's what most firms don't realize. The best accounting talent in the Philippines doesn't want to work for outsourcing companies. They want to work directly with firms like yours. Think about it. These are accountants. They can do the math. They know you're paying that BPO way more than they're getting. That's where Team Up comes in. They help you hire accountants in the Philippines directly. No middlemen, no corporate politics, no endless turnover. You get to build a lasting relationship with your people who become part of your team. Follow your processes and share your culture. It's just like hiring someone remotely in another state. Plus, you'll save money while your accountants earn more. Everybody wins. Want to learn more about why forward thinking accounting firms are making the switch to direct hiring? Head over to The Accounting Podcast. That's The Accounting Podcast forward slash up. Let's talk about PwC. The effect on the big four. Pwc was in the news talking about how they are changing, how they train junior staff because of AI. Jen Kosar SA is a is an AI assurance leader at PwC and said that new hires at the big four firm are going to perform manager level responsibilities within three years because they'll be supervising AI systems, handling routine audit tasks. So the entry level employees, as opposed to the current situation where they do a lot of the work, they're going to become reviewers and supervisors of AI work. So it's going to accelerate the development of the firm, where in three years you'll be doing the work of a fourth or a fifth year.

David Leary: [00:20:04] So what's interesting about that article, Blake, is there's a whole side article, but from PwC, because a slide from a slide deck got leaked, PwC is planning to recruit one third less grads in 2028. And it's because of this of automation. They truly believe that, um, between AI integrations and their offshoring hubs. They're just going to hire less in the US. So one third less grads is a lot. And then this is probably if PwC is doing it, that means all big four has a slide like this that they're so across the board. Big four is not going to recruit as many college grads in the next two years.

Blake Oliver: [00:20:41] You're cutting out an entire stage of the funnel so you just don't need as many people. And the question is, where are all those young accounting grads going to go if the Big Four isn't hiring them? It means that accounting, education, all the colleges and universities that have traditionally sent students into the Big Four and the big firms are going to have to figure something else out, because there's just not going to be as many seats.

David Leary: [00:21:09] So and I've popped into like, those job fair type places at the college campuses here in the high schools here in Tucson. And the big four have the biggest tables, the biggest booths. They hire the most people they claim. At least that's what they're claiming. You know, in many cases, um, you know, hey, if you're a business major, we're going to get you placed, but that's all going to go away. They specifically say they're going to restructure graduate hiring. So that means they're not going to be on campus going to these job fairs. It's a this is a major, major impact of AI that directly in our space.

Blake Oliver: [00:21:43] All right. I want to show you what I've built in Zapier agents, because this is this is going to unlock, I think, a lot of automation capabilities, a lot of possibilities of AI for firms, because a lot of the AI agent building platforms have been highly technical up to now. And one thing I love about Zapier for automation is that it's just easy, relatively easy to use. It's a no code platform for automation. Now they've added this AI agent functionality, and it works in a very familiar way. And it's it can be locked down so I don't have to worry about my AI agents going rogue. And I think that's really important. And it works with all the apps that already integrate with Zapier. So you don't have to build a bunch of custom API stuff. So I mentioned that I was able to build the core functionality of Hubdoc, which is the OCR tool that scans, invoices, bills and enters them into your accounting system. And there's a ton of these David Wright auto entry.

David Leary: [00:22:49] Entry list goes on and on and on.

Blake Oliver: [00:22:51] Um, you know, QuickBooks now has its own Xero and it has its own. There's a bunch of tools that do this separately. There's tools that are built into apps. And one thing that's kind of frustrated me is that, like, I just felt like the tool built into Xero is not very good. Like where you forward the emails, like it's just it doesn't work very well.

David Leary: [00:23:12] Xero just got another one with Melio. It does it too. So now they have two tools that do it.

Blake Oliver: [00:23:16] But they're all kind of like janky in a lot of ways. They're like just not very. They don't companies don't really invest in these OCR tools all that well, I think because it's like not a core feature of their app and they just add it to add it and then they let it go and it's just not that great. And I want to customize it. Right. So what I've built here on the screen is a very simple agent in Zapier agents. And the instruction is very simple actually. I'll start with the trigger. So the trigger is it's a trigger via zap. So it's triggered by a zap, which is, you know, a defined workflow where if I forward, I forward an email with an attachment to a special secret email address, and that's what triggers the agent. Um, I've done that, and I should actually, I should show you the zap, because integrating your zaps with your agents is really powerful. And so I want to show you that as well. So I'm going to look that up right now. So what is it called? Email bill. I have to find this app. I'm missing it right now. I don't want to hold everything up for that. Let's just just assume that I've got a I've got a zap. Okay. And what it does is it takes the attachment and it uses a tool called PDF and converts the it extracts the text from the PDF.

David Leary: [00:24:38] Okay.

Blake Oliver: [00:24:39] So I get now I have the email and I have the PDF text. And I can pass that to the agent. So this is where the agent comes in. And that's what you're seeing on the screen. And the agent has some simple instructions. When I provide you with a document, extract the following details from the document. And then I've got a list of all of the typical details you would see in a bill. Step two enter the bill into Xero. Create a new bill using the extracted details. And I've linked the action in Zapier called create Bill, and I can define what organizations the agent may choose. And I've specified it has to choose Earmark, Inc., because I only want the bills to be created in Earmark, Inc. and this way I don't have to worry about the agent making a mistake. I also want to force the agent to choose draft status, because I do not want it to be able to approve an invoice. So because of this, it will only be able to create a draft invoice in the Earmark ink zero file. And I'm letting it select all the other values that I could have to program. So what is the contact name. What is the attachment. What is the line item type. All that stuff.

David Leary: [00:25:52] The description you're throwing some trust out there on it. You trust to feel these fields in.

Blake Oliver: [00:25:58] And what's beautiful about this is in the past, if I was creating just a zap workflow, I would have to define all of these fields and I would have to somehow map from the invoice what should be used in all these fields. And that's what all those OCR apps did for years in different ways, into different levels of success, and often has.

David Leary: [00:26:20] Human humans involved somewhere in the process that was doing that for you, you know, and that's what I think. If you think about what Hubdoc and dext and auto entry, that's there's lots of apps that did this, but they had humans kind of verifying the data and making sure it was.

Blake Oliver: [00:26:34] Yeah, not just kind of it was essential.

David Leary: [00:26:36] Because, yeah.

Blake Oliver: [00:26:37] I couldn't do this reliably, but now it can I can very reliably take that text dump of the PDF and the email and figure out how to map it into Xero through the API. So that's what it's saying to do here. I also, in the instruction set said use the attachment link for the file to attach it in zero, and then search for previous bills from this vendor to determine the proper expense account. If you can't find the previous. If you can find previous bills, code this bill. Similarly, if there are no previous bills, select the best fit expense account for each line item and I gave it access to an API request from zero, which is like a a different way of requesting information in case it needs to. Step three is to email Blake Oliver to let him know that a draft bill has been entered in zero and is ready for review and approval. Include the link to the draft bill's screen so he can easily go view and approve the draft bill. Include the details that were entered into zero and the body of the email and at the bottom of the email, include the URL to edit the agent instructions, because I just want to be able to click and fix it if something's not right, and then I say expected outcome. All receive bills are accurately entered in zero with the original documents attached, and are ready for review and approval.

Blake Oliver: [00:27:50] So that's it. Okay, that is the whole thing. And it's got these tools. It's got zero and email by Zapier. That's all it has. So that's all it can use right? It can't go do anything else outside of those tools and how I've configured them. And so what does that look like? Well Here's me in my email forwarding a bill that I received, and you can see there's an invoice PDF attached. And then I wait a little bit and what do I get back? Well, I get back this email from Zero Automation Bot and the subject is draft bill entered in zero. And it's got the name of the vendor and the bill number. And then in the email body it's got all the details about what it extracted and what it entered. And it says that the bill is in zero and it's ready for my review and approval. It shows the line items and there's a link to the draft bill PDF. Um, it's got and actually that, that that link is like labeled him correctly. If you click that link, it takes you to the draft bill's screen and zero where you can then approve it and it even attached the document to the email for me. I didn't ask it to do that, but it decided to do that.

David Leary: [00:29:02] Can you pull that up just because somebody is asking in the chat if they can see it in zero? The actual the bill got entered in the Attachment.

Blake Oliver: [00:29:09] So here is what it looks like in zero. And this is a screenshot. So I could blur out some information. But you can see it added in uh a the vendor name the date due date reference number, which is the invoice number of the vendor. Uh, bill it attached the file, it put in the description quantity unit price, and it selected the correct account. And you can see I've already matched a payment to it. But yeah it did it. And I haven't tried this with like a super complex bill with tons of items. But I'm pretty sure that it could do it. Maybe I would have to like work on the instruction set, but this is basically hubdoc. This is what most people were using that for or auto entry. And I've managed to build that in like an hour with a Zapier agent.

David Leary: [00:29:55] So because it's AI, right? Do you constantly, forever now have to keep telling it all these little scenarios forever? Or is it ever going to just learn this stuff on its own? And where I'm coming from on that, all that stuff you said to do, those rules you would configure, like I'm going to auto entry, you'd configure rules. That's how I would be able to tell it. If it's a GoDaddy receipt, look for office 365 and put that to SaaS expense. But if it's a GoDaddy and it's a domain renewal, yeah. Or a website that needs to go to marketing and you would set up rules to do that, but you're basically just typing in the rules instead of like going through a UI to do the rules. You're kind of getting to the same end. That's where where's the actual AI coming in? Just the clicking the building.

Blake Oliver: [00:30:38] That whole workflow traditionally would have just taken a long, long time. It would have had to map all the fields. I would have had to say, extract all this information and then map it to every single field in Xero and doing line items. I don't even know how I would do that. The logic for coding because some bills have one line items, some bills have ten, right? Yeah, the AI can just do that. It can figure out how to map the API. So that's a big time savings. What you're talking about is the policies that the bookkeeper would follow and the way I would do this. And we can experiment with this, David, and I'm pretty sure it would work, is we'd create a Google doc and call it Earmark, Inc. Accounting Manual, and we would begin to document how we want everything coded in that document. So we could have a section for the coding of different vendors. And we could say for this vendor, we always want it to this unless it's is this case, in which case we want it coded this way, just like a human, you'd give a human right. And we could add that document to the knowledge of the agent so that it could go search it and find the answer for what to do. And we could make that part of the instruction set.

David Leary: [00:31:54] This is why companies are not observing savings, because if you already had auto entry configured and all those rules were set up, now I hear it's not going to like it already works. The process works. It's not. It's not that revolutionary.

Blake Oliver: [00:32:07] It's not a ten x.

David Leary: [00:32:08] It's a yes. Exactly. It's it's it's like if somebody came with a new microwave and it cooked food in 15 seconds instead of 25, you're probably going to buy a new microwave. This is why.

Blake Oliver: [00:32:19] It took like a whole generation for households or longer for households to start using dishwashers when the dishwasher was invented, because you had an entire generation of people, women who had been trained how to wash the dishes really, really well, and the dishwasher was not necessarily better, and you had to convince them to want to use it. And I is the same way, because we have a lot of people who are really good at doing this stuff. And the AI is like, okay, you know what I mean? Does that make sense? Yeah. Like and that's.

David Leary: [00:32:56] And I think the only difference is the, the okay is all right, because it can do it 24 hours a day, seven days a week, nonstop, constantly. And that's that's.

Blake Oliver: [00:33:05] Now that I've set this up, I mean, but like, think about how easy it was for me to set this up. Now, granted, I've been kind of playing with automation for years and whatnot, but these this, this platform and if more developed like it, it's still in beta, but it's just going to get easier and easier.

David Leary: [00:33:21] You build an app that got acquired for $30 million basically.

Blake Oliver: [00:33:26] Yeah. Right. I was able to do that evening.

David Leary: [00:33:28] How long how long did it take you to build this app?

Blake Oliver: [00:33:29] I did this in a morning in like an hour. Yeah. I built the basic instruction set and connected it to zero and tested it. And, you know, the rest is just debugging. But it works like we can now just forward any PDF attachment. Now could it handle, uh, how would it handle just like the if the bill is just the email itself and there's no attachment. Well, I could add some logic for that, but that would be pretty easy because I just add in like an I step in the workflow to figure out what to do if that happens. Yeah.

David Leary: [00:33:59] And I'm sorry You built a $70 million acquired app. Zero bought Hubdoc for $70 million. So.

Blake Oliver: [00:34:06] And you know, now imagine that we have like, that build entry agent. And let's say we add a receipt entry agent. And let's say we add a, I don't know, reconciliation agent. And we have all of those connected to an accounting manager agent that forms the workout to the different agents with their instruction set. It's basically turning your companies or your departments accounting manual into the real, into action in the real world. But you have to have the manual. There has to be something the AI can read and understand and follow. So this is where the gap is. If you don't have the accounting manual and it's all in people's heads, you can't automate it.

David Leary: [00:34:57] So this is where there's again, I want dumb AI or an AI only trained. What if you had AI only trained on QuickBooks zero and the top 200 apps in both ecosystems and some standard operating procedures like that's a valuable piece of AI that would scale across thousands and tens of hundreds of thousands of accounting firms. Right.

Blake Oliver: [00:35:19] Versus, you know, what you could do? I mean, it's like, I wonder if I could just set like ChatGPT Agent Loose on Xero and say, you know, go click around and figure out how everything works and read the help documentation and create like a custom instruction set for an agent whose job is to reconcile the books. Like, could I actually get the agent to create the most of the instruction set for all the standard stuff?

David Leary: [00:35:44] Yeah, because you have to eliminate the guessing. We just had this bug. We couldn't figure out why a bunch of users of earmark were putting earmark as the company they were, or their firm name, we can figure out. And when.

Blake Oliver: [00:35:54] They were.

David Leary: [00:35:54] Saying that, that Uh, was not not. I wouldn't say it went rogue. It's just if there was no company, it would be like the company must be earmarked. It would just. Yeah. That's like again. Well, so.

Blake Oliver: [00:36:06] Here's what was missing.

David Leary: [00:36:07] Yeah.

Blake Oliver: [00:36:08] Here's what was missing from that. And how we could solve it is we could have a review agent that reviews the output of the worker agent before it goes into the CRM.

David Leary: [00:36:20] Yeah, but a human wouldn't like. Like if there was a blank field on data entry, a human wouldn't. Just like I'm going to put in this value, just pull out of their butt and just put it in.

Blake Oliver: [00:36:28] Well, David, I've, I've worked with some people who would. Don't underestimate the stupidity of humans.

David Leary: [00:36:37] That's true, that's true.

Blake Oliver: [00:36:39] Right. Um. Anyway, so that's that's sap your agents. What else have I built with it? And I want to hear if you're building AI agents, I want to see them. And I don't want to just talk about them. I get vendors, You know, software developers all day long reaching out to us and saying, like, we've built this amazing technology we want to show you. And then I get on a call and there's nothing to see, okay. Like what I just demonstrated on this program. I want to see that. I want to see how it works from end to end. And send me a video. And if it's really good, if it's interesting and does something cool, I will showcase it on this program for our listeners who are really interested in learning this stuff.

David Leary: [00:37:21] Just like even if it fails, like that's okay. Like it's okay because it's.

Blake Oliver: [00:37:26] Okay if it's unreliable, it has to work. But if it's like not every time we want to know and why and how, even.

David Leary: [00:37:31] If you're doing it fails and you share it, this is how you got the knowledge to go retrain yours, right? It was on you, shared on social media, and you got input and you were able to improve it. So it's okay to share it. If your I think you tried to build doesn't work.

Blake Oliver: [00:37:46] Here's one that I built that I'm really proud of. And once I get this working, how I like it. I'm going to share this with everybody because I think it's super cool. So this is a meeting agenda organizer agent. And the trigger is a new meeting. Transcript in fireflies. I use fireflies to record my meetings. When the meeting is over, it saves the recording, the transcript, and then action items and notes. In fireflies. So Zapier watches fireflies, and when there's a new meeting, it grabs the transcript and the action items, and the agent then extracts all the follow up items, including action items and any unresolved issues from the fireflies. Transcript. It organizes the extracted items and issues into a structured agenda for the next meeting in plain text. And then this is the part where you couldn't easily do this with workflow manual workflows. It searches my Google Calendar for the next meeting in the series based on the current meetings details. So we have a sales stand up every day. It if the meeting is transcript is related to the sales standup. It looks for the next one. So it finds that next event. And then it adds the follow up agenda into the description of that calendar invite. So now automatically at the end of every meeting, the next meeting now has an agenda that is focused on everything we need to follow up on. So nothing gets dropped. Now, is this 100% effective? We've been doing this for one week now, David. Have we have we missed anything? Have we noticed that we're missing anything? I feel like because we've been using this, we have talked about stuff that we would have dropped. What do you think?

David Leary: [00:39:40] I think the big issue is the it's just getting into your workflow and making sure you go and read it and then you have to you have to operate yourself a little bit differently, because if three of us are in a room and we're all talking about the next item, you have to be very deliberate to say now we are now talking about item peanut butter. Now we're talking about item banana. And you go through it. You have to be very conscious when you're doing this. So the AI knows what you're talking about and picks up from there. But it works. But like you're we're four days in.

Blake Oliver: [00:40:10] Like right. But already David, I used to have to go and extract all that stuff and then add it into like a Google doc. And we had a Google doc attached to every meeting, and I would have to like the work that I had to do as a manager, just to stay on top of all of these open items and make sure they don't get dropped from meeting to meeting. That is now handled by the AI, and it's doing it well enough where I'm I'm going to stop doing it manually. It's good enough for our sales process, and we have some other ways to make sure that stuff doesn't get dropped. So I'm really confident that this is going to save a ton of time. And and like it's that kind of stuff that these agents are really great at. It's so convenient.

David Leary: [00:40:56] I'm going to jump into our next ad. Blake, will you prepare the next story?

Blake Oliver: [00:41:00] Yes. So let's thank human at scale. Are you the bottleneck?

David Leary: [00:41:06] Sorry.

Blake Oliver: [00:41:07] Are we the bottleneck in reading these ads simultaneously? You go for it, David.

David Leary: [00:41:13] Are you the bottleneck in your own accounting firm? You've built a successful practice, and you have a team. But now you're drowning in client work instead of growing your business. That's where human at scale comes in with their operations audit. They're not just consultants who drop recommendations and disappear. Their team actually works alongside you to transform how your firm operates. And here's what makes them different. They guarantee measurable outcomes, real improvements in your margins and team productivity. They build systems that allow your junior staff to step up, freeing you from constant firefighting and quality control. Imagine consistent client experiences, smoother onboarding, and processes that don't require your personal touch on every project. Their clients seem stronger, margins, increased firm valuation, and maybe most importantly, they regain. They regain the time on to work on their firm instead of in it. Human at scale helps six and seven figure accounting firms create scalable systems that run without the partner being involved in every decision. Recently, in a mere 60 days, they helped a firm reduce client onboarding time by ten days and improve team response time by 75%. To see how they can transform your accounting practice from chaos to consistency, head over to The Accounting Podcast The Accounting Podcast.

Blake Oliver: [00:42:30] Thank you. Human at scale. All right, let's go to our live stream chat. Lots of lots of chat going on today. Thanks everyone who joined us live and has commented Gator NYC boring accountant Big Four transparency. The coffee emojis are flying this morning. Now, this afternoon, as we proceed into our Friday night light, Zephiel Zephiel wants to know how do we know that you're not AI agents? Uh, because thankfully, an AI agent is not reliable enough to host a podcast yet. Although at the current pace doubling every seven months in capability, it could be, uh, maybe in a few years. So, David, that's why we got to make earmark a success, right? We got to exit before, uh, AI takes over the world. Boring. Accountant says AI agents don't drink coffee like David just sipped.

David Leary: [00:43:29] I think it could be an empty cup.

Blake Oliver: [00:43:30] It's just the deepfake videos are getting so good. Honestly, you wouldn't be able to tell. Nightlight says I will replace all accountants, I disagree. Well, it depends on the time frame, right? Like, will I replace all human labor that we do today on a long enough time horizon? Possibly the same way that robots have replaced most manual labor. Will that be a bad situation? I'd like to. I'm an optimist. I like to think that, you know, we'll all get to, like, do more interesting stuff.

David Leary: [00:44:05] I mean, it's going to do it's going to eliminate some. So I just got a new air conditioning unit. Mhm. People had to come and install it, but people didn't have to lift it off the ground and put it up onto my roof. A crane came and did that. So there's going to be advantages to using technology but it's not going to replace. Yes, it replaced 50 people that had to work together to build a ramp and push it up a ramp to get it onto a roof, but now with a crane, you don't need to do that. So yes, it's going to replace some jobs, but not all jobs. Right?

Blake Oliver: [00:44:38] Right. And we will we will create new ones. And also the cost of goods is going to go towards zero, because we're going to build automatic automated manufacturing facilities where you just input raw materials and electricity and out comes finished goods on the other side. We're going to have like factories like that in this country. So just imagine like basically print on demand products that will happen eventually. We'll get there. Um, and that'll make things really cheap. And then we just have to deal with our housing and health care crisis. All right. Continuing down the chat, the deck says the partners at our office definitely got burned by an implementation attempt, but are now looking toward outsourcing to Asia for all work below senior level. You know, there's an argument that outsourcing and offshoring is like, if you're deciding between that and doing AI stuff, I think you're better off offshoring if you are choosing that because it's been done for a long time, so it's reliable now. Where's all the AI stuff is new. So if you're just looking for, like, an easy win. Um. And you think that'll help? But the other option is you skip that entirely and you just say, you know what? We're going to use us people, and we're going to give them all these awesome tools, and we're going to reorient our firm to be based around AI workflows. And then you were going to get way better results than if you offshore and just keep your horrible, uh, business model. But even if you.

David Leary: [00:46:08] Offshore.

Blake Oliver: [00:46:08] For time.

David Leary: [00:46:09] Offshoring could actually force you to document your processes and create rules in these things you want.

Blake Oliver: [00:46:15] That's true.

David Leary: [00:46:15] And then now you can take that next step. Yeah.

Blake Oliver: [00:46:17] Often I think the problem with the offshoring is that these firms don't create good documentation and processes. And so the work comes back and then you just end up having us people fixing it, and it just wastes their time even more. Right. That's the complaint about offshoring.

David Leary: [00:46:31] Which is why maybe offshoring is probably the biggest risk of AI.

Blake Oliver: [00:46:34] Yeah. Um, Boring accountant says AI is automating the job of AI. We've added AI to the payroll, but we don't know what exactly their job description and role is. And I constantly needs direct supervision. Yes, you must have a human in the loop. And the key is to build in those approval processes and workflows. And some great examples are you can you can have an approval step in an agent that goes to slack. And you can add a button to approve or disapprove so that the agent will not proceed to like sending an email until somebody is approved. That's a great way to like add the human in the loop and and then you have a link back to the agent so you can update the instructions and rerun it if something is wrong. And you just need to keep doing that for a long time until it gets reliable. I think my favorite part about building an app with an AI agent is that I can completely customize how it works, versus buying off the shelf solutions where I just have to take what I get. So you're basically building completely custom software?

David Leary: [00:47:41] Yeah.

Blake Oliver: [00:47:42] Without code. That's what it is. Matt. Cases I for now is mostly hype. Yes, I agree that most of it is hype right now, but we're finally getting hype, man.

David Leary: [00:47:53] This is this is Blake's current role. He's the biggest.

Blake Oliver: [00:47:55] Finally getting to the point where it's getting really good. Um, but you're right. Like all these, like, CEOs that are investing in AI, I feel like a lot of them don't know anything about what they're talking about. And it's just like the.com boom where everyone's like, you got to have a website and do all this stuff, and they don't really know what to do. They're just throwing money at it because it looks good to the investors. They feel like they have to. It's just the hot fad, right? And so they're doing it in ways that just waste money. And it's going to take a while for people to figure out how to do this the right way. And it's going to take a long time to document this stuff and to get it into production.

David Leary: [00:48:32] So I have a question. Um, Intuit released their Q4 earnings call, their earnings report for Q4 and the earnings call. So and it's typically like everything's great in Intuit. Things keep growing. Qbo advanced Enterprise Suite customers grew 23%. The Qbo base grew 8%. Um, TurboTax live is one of the biggest success stories for them. For fiscal year 2025, revenue is up 47%, customer growth is up 24%. Um, and it's they're saying it's a key driver to 10% of the revenue growth in the consumer group. So super love of TurboTax live going on. But nowhere in their earnings call was QuickBooks Live mentioned. So TurboTax live, they're shooting off rockets, shooting off fireworks, celebrating how great success, how how successful TurboTax live is. But QuickBooks Live wasn't mentioned at all. Instead, the focus is on QuickBooks Online Advanced and Intuit Enterprise Suite isn't to it. Giving up on QuickBooks Live, it makes you wonder, like.

Blake Oliver: [00:49:34] Yeah. Yeah, absolutely. If I had to bet on it. Yes. Because it's really hard. It looks simple. Bookkeeping always looks simple from the outside to people who have never done it. And then you get in there, just like we saw with our AI agent. When you have a transaction to reconcile, it's not always obvious what you have to do. And there's a big long list of things you have to check to do it right. And if you do it wrong, it just compounds. And the problems get worse and worse and worse month over month. And then you have a giant mess to undo. Yeah, it's not easy.

David Leary: [00:50:08] When we first discovered QuickBooks Live, we were going back three years ago. I think we broke the story on the podcast 3 or 4 years ago. Quickbooks Live. And one of the thoughts I always had was, this is really into it, learning the last mile of automation they need to do. And with AI. So now, now, now Intuit has AI, they have automation. They've been having quote unquote employees. They've been employed employing QuickBooks bookkeepers, essentially. And now you're studying how they've been doing all the work in QuickBooks. Maybe things will just be automated and they don't even need it'll be called QuickBooks I and just instead of QuickBooks Live. Right. And that's what you that's the service you'll subscribe to. And they're going to get rid of the humans entirely. But it's just interesting that for something that was they were so big on just was not in an earnings call. Transcript. And not even questions from the analyst either. Like nobody talked about it.

Blake Oliver: [00:51:01] Regarding the job market for accounting grads. Jeff said, my son said that in his big four firm, they offered only one job to the interns this year, as opposed to the usual 5 to 6. Larry says regarding the AI agents in Zapier, the big unlock here is more endpoints exposed in Zapier or similar tools. Yes. So the fact that the endpoints, the apps you can connect to are already defined in Zapier is what makes the agent platform powerful and easy to use, and something I would use even ahead of a more sophisticated tool. Because it's already set up. I don't have to create a bunch of API connections that I don't know how to code and program. It's just all there.

David Leary: [00:51:52] I saw a piece of AI. I forgot what the name of it was, but that's what that company does. They just build integrations back and forth. So. So if you don't know how to use the AI of this app or the API of this app, the API of this app, but you want to move the data around, it'll go and configure all that and build that for you to move that data around.

Blake Oliver: [00:52:10] And boring. Accountant says the new concern about AI is losing your AI bot in rules and functions. When a new AI model gets rolled out, like with the ChatGPT five fallout that happened recently. So if you've built projects in ChatGPT or custom GPT, My experience is that when they roll out a new model, it just gets upgraded. Now, it might not behave the same as you expected because they changed how the model works. And that was what people were complaining about with GPT five. Is that because GPT five has a router in it that routes your requests to different models, the prompts that people had customized for GPT four zero, which doesn't have that routing agent, didn't work as well in certain cases. So they had to learn a new method of prompting. So all this prompt engineering that people learned for 45 tries to eliminate that. So you don't have to engineer your prompts. But if you did engineer your prompts and you tried to use the old prompts with the new GPT five, it doesn't work as well. Does that make sense? Yeah. So it's like you got to learn the system again. Now, if you're doing something in like a Zapier agents. The beauty is that and this is also disadvantage is they abstract all of it from the LLM.

Blake Oliver: [00:53:25] Lm so you don't know what LM Zapier is using when the agent is working. I can't figure out how to customize that. So Zapier is doing all that. The good news is that theoretically, when the models are upgraded, Zapier will just plug in a new model and everything will work as expected. But the disadvantage is you might not get the new model as soon, because you can't just plug in your own API key the way you can with other tools, so that's the disadvantage. But based on how Zapier works with companies, I'm confident that they're not going to break everyone's automations. Like when they upgrade their API connections to apps, they leave the old one as a legacy connection that you can then upgrade later. So it doesn't break everything. So that's that's what's going on there. Um, but if you do want to do a specific prompt with a specific model. Zapier lets you do that, and you could actually give your agent a tool to run a zap that does that prompt with that specific model. Does that make sense? So you can get around these limitations, and you can do like a really sophisticated prompt with a specific model if you want to, and then feed that back into the agent. You know, I didn't explain what an agent is.

Blake Oliver: [00:54:50] And maybe that would be helpful because I think there's some confusion in the discussion in the chat about what is a workflow and what is an agent. So Zapier, the Zapier everyone knows and loves, is a workflow app. I can define steps. I have an input, like a trigger that starts the workflow, and then I have steps in my workflow that transform information, add information, subtract it, massage it, change it, Send it to different apps. Pull it from different apps. That's all rules based. You can add in an AI step that lets you prompt a particular AI model using inputs a prompt, and then define specific outputs. So that's how you integrate AI into your workflow. But that's not an agent, that's just an AI workflow. You're just using AI to transform your inputs into different outputs. Yeah. The difference between that and an agent is three things. The agent has a goal. Actually, you could say that the workflow has a goal to because the prompt itself should have a goal. So they both have a goal. Okay. So the agent has a goal. It has access to tools multiple tools often. And this is the important part. It can make decisions as to what to do next. So in a workflow.

David Leary: [00:56:14] You're not telling it exactly what to do. You're giving it guidelines of what to do and then it goes off on its own.

Blake Oliver: [00:56:21] Yeah, I'm telling it. And I can be as detailed as I want in the instructions. I'm giving it instructions in English or whatever language you want. Right? I'm just giving it an instruction set, and then it reads those instructions and decides what to do next. So I'm going to do an API request to research the transaction history of this contact in zero, and then gets the information and then decides what to do next okay. Based on what I found or didn't find. Now I know how to code this transaction. Or maybe I don't have enough confidence. So I'm going to go send a slack message to somebody asking them what to do. And then that comes back, and then it decides what to do next. So it can it can basically cycle through this process to achieve a goal. And the reason people have trouble with agents is because they'll often give it a goal that is just too big. You're way better off at this point with simple goals, tasks that take a few minutes. That's where you can very, very, very reliably get it to do the task. Like with almost 100% accuracy, is if it's a task that takes a few minutes, no more than an hour or two. And typically, those kind of tasks that take an hour or two are research based tasks, which is also really good at, like going out and searching hundreds and hundreds of web pages and pulling all that information back in. But when it comes to tasks to take a lot of time, because there's a lot of different things you have to do, that's where these agents time out get confused. They have very short memories. They're like goldfish compared to us, right? The context window is how much like the agent can actually remember, which is really not that much. So you want to give it tasks that that don't take a lot of memory that that don't require it to like remember what it did five steps ago that only required to remember what it did in the last step or two?

David Leary: [00:58:14] I've been kind of thinking about this in the real world. Basically, you're hypnotizing your agent or the I remember how on the prompts you used to have to say, like, you are a marketing expert that's going to do this project. It's like you're hypnotizing somebody, right? And the only thing they're going to do is the function you tell them to do. Mhm.

Blake Oliver: [00:58:31] Let's take our final sponsor digits. Let's be honest accounting software hasn't changed much in decades except for rising costs and declining service. Now there's finally a reason to switch and never look back. You can now do your bookkeeping on digits. Digits is an AI native accounting software that works for you, not the other way around. While other platforms just slap ChatGPT on old workflows and call it AI, bookkeeping demands more than a chatbot. It demands precision, auditability, and trust. Digits rebuilt ledger software from the ground up with probabilistic categorization and human review prompts the result. Over 95% of transactions auto book with unmatched accuracy. 54% better than ChatGPT style models so you can close faster. Stay in control and finally stop wrestling with your accounting software. Low value clients are now high value four years worth of cleanups. Now take just two hours, all while your clients get visually stunning. Reports streamline collaboration and the insights they need to make better decisions to see why hundreds of firms are making the switch to digits, head over to The Accounting Podcast. That's The Accounting Podcast.

David Leary: [00:59:42] I have a two surveys, but I don't know about if you have any major news to chat through still.

Blake Oliver: [00:59:47] Well, in the description of this episode, we said we'd talk about the UK using AI to spot tax fraud on social media. So I think we got to talk about that.

David Leary: [00:59:56] Story.

Blake Oliver: [00:59:56] Because it's kind of wild actually. Like so His Majesty's royal, uh, what is it? Hmrc, Majesty's royal collectors. It's the it's the IRS of the UK.

David Leary: [01:00:13] I know. I can't remember. I'm blanking on it too. I've trained myself to say it and now I cannot remember exactly how it goes.

Blake Oliver: [01:00:21] It's so this.

David Leary: [01:00:22] Is the UK.

Blake Oliver: [01:00:24] The UK is IRS. They have confirmed that they are now using AI to scan taxpayers social media posts for tax fraud. Hmrc is now using AI to scan taxpayer social media posts.

David Leary: [01:00:35] Not scan tax forms. Tax reports. What you're doing on social media.

Blake Oliver: [01:00:41] They're comparing it to what you reported. And if you're living a lifestyle that's in excess of the income you claimed, they're targeting you for audits. That expensive vacation or new car you posted about. It could be a red flag if you declared income that doesn't match your lifestyle. It's not just for criminal cases anymore. Hmrc is going to expand AI into everyday tax processes to close their 7 pound, 7 billion pound tax gap. The system will cross-reference your financial data, tax returns and public social posts looking for inconsistencies. Hmrc has quietly changed their privacy policy language from guaranteeing human judgment to just human involvement.

David Leary: [01:01:23] This is horrible, really, really horrible. If we go back to the UK Post Office. Post office problem, right? The UK has a law basically that says you have to prove that the computer was wrong. You're, you're, you're you're basically guilty until you prove the computer was wrong. Well.

Blake Oliver: [01:01:42] And guilty until proven innocent. That's a very European thing. Like people don't realize how unique that is to the American system of justice that we have this. And so they give the computer proven guilty.

David Leary: [01:01:55] The the computer got the benefit of the doubt over the human. And then humans were put in jail. So now AI is going to look and make these calls based on your social media post. And then you could be put in prison or get fined or whoever. Whatever the punishment is. And now you have to prove that the I is nuts and didn't work correctly. This is bad.

Blake Oliver: [01:02:16] You can already be put. You can already be put in prison in the UK for your social media posts.

David Leary: [01:02:21] That's true, that's true.

Blake Oliver: [01:02:22] They don't have free speech there.

David Leary: [01:02:23] Yeah.

Blake Oliver: [01:02:25] Not the way we do. So I mean, here's the thing though. It's like this is actually a great way to catch fraud is monitor what people do. Like in all of the fraud cases that I've listened to on the podcast, oh my fraud, one of the biggest red flags is people living a lifestyle greater than their income would suggest they could. So if if we just had an algorithm watching for that, we could probably catch a lot of bad actors.

David Leary: [01:02:54] Well, this is why in Miami all the Pip fraud was out in the most of it was in Miami, some huge percentage, because there's the place to have the lifestyle to spend that kind of fraudulent money on. You can buy a Lambo. I can't buy a Lambo in Tucson if I pull off a big old fraud. But you could do that.

Blake Oliver: [01:03:10] Well, and you can just take the cash on, like a one of those little, like, boats. What are those called? The ones you like. Stand on.

David Leary: [01:03:17] The jetties. Yeah.

Blake Oliver: [01:03:18] You just take a jet ski to, like, an island somewhere and deposit your cash. That's.

David Leary: [01:03:22] Yeah.

Blake Oliver: [01:03:23] Um. All right, last story we have to hit on before we go. The results of Accounting Today salary survey. So this is not doom and gloom. In a world of AI, we are seeing salaries continue to go up. Entry level staff accountant salaries jumped $10,000 in one year. The median is now at $75,000 a year. Now that's the median across all firms in the survey. So that may be more or less than your local market, especially if you're in a high cost of living area. But the change from 65,000 to 75,000 is pretty significant. So here we are zero.

David Leary: [01:04:01] Because they don't plan to hire anybody. It's actually zero.

Blake Oliver: [01:04:04] It's fewer hires. But now the money is going to the people that do get hired because the firms are saving some money and they're investing more in the people who are more capable. So this is the thing about AI is that people who can use it, who can become those managers in three years. And I would put myself into that category, are going to have a lot of opportunity. And the people who are just looking to follow an instruction manual, they're going to get replaced by AI because the AI can follow that instruction manual if the firm can actually get it written down.

David Leary: [01:04:39] I have two other pieces of survey data from two other surveys.

Speaker3: [01:04:42] Wait wait wait wait wait. Don't you want to know what other people are making?

David Leary: [01:04:43] I thought you were done. Sorry. You paused too long there.

Blake Oliver: [01:04:45] Sorry, sorry. Well, okay. So staff accountants are making 75,000. Median senior accountants are now at 93,000 median managers. If you're a small firm manager, you're making 100 K. Large firm 130 K. Remember these are median numbers. So there's a range Uh, partners. This is interesting. Small firm partners. 155,000. Median salary. Large firms. 210,000. Median. That seems so low to me. It doesn't seem right, but I swear to God, there just must be, like, a huge range in the partner comp area. Like, if you're an equity partner who's been around for 20 years, you must just be making like millions of dollars. And it indicates to me that, like, the new partners are working way more than they should for the amount of money they're making because, like, I would not, at this point in my career take $200,000 to work 60 hours a week. Zero chance.

David Leary: [01:05:47] Even 50 lying on the survey because, you know, a little bit.

Blake Oliver: [01:05:51] But that would make it higher. Generally, people inflate their salary.

David Leary: [01:05:55] Well, unless you kind of once you get to what.

Blake Oliver: [01:05:58] This.

David Leary: [01:05:58] Means, you don't want people to know, right.

Blake Oliver: [01:06:00] What this without any further investigation. This is my hot take on this, is that the people that are coming in as new partners are not getting, uh, are basically getting the lowest salary that the other partners can give them. And that's why think about it. You know, if you're a large, firm partner, you're making 210. You could go off on your own and start your own firm and be making 155 and work a lot fewer hours. I mean, you probably make more on an hourly basis. You could I mean, I know people who are making 200, who are working 20 hours a week as a small, firm, independent CPA or close to it. All right, David, I'll let you take us out.

David Leary: [01:06:50] Yeah. So two more pieces of data from two other surveys. So ignition has their new 2025 US accounting and tax pricing benchmark report out. And the big the big takeaway in this one is two thirds of firms raised their prices last year saw no or little client loss. So two thirds, two thirds saw no client loss raise their prices.

Blake Oliver: [01:07:12] That means they didn't raise their prices enough.

David Leary: [01:07:14] Okay. That's what it could be as well.

Blake Oliver: [01:07:16] You should raise them if you don't see if you don't start losing clients. That's that's how you know, when you've hit the limit is when you start losing clients.

David Leary: [01:07:23] Yeah. So the the study is based on Twitter 19 US firms. And their typical fees for an individual returns is 400 to $599. Business returns 1000 to $1499. Advisory 2000 plus annually. Bookkeeping 250 to 500 a month and CFO services at $2,500 a month. So those are those.

Blake Oliver: [01:07:44] That you went through that real fast, David, but that like tax return pricing is is so low.

David Leary: [01:07:49] It's for an individual return depressing you.

Blake Oliver: [01:07:53] We as an industry could charge so much more. And I know what did you say the business returns were going for.

David Leary: [01:08:00] So the business returns are 1000 to $1500.

Blake Oliver: [01:08:05] Yeah, that could be 2 to $3000.

David Leary: [01:08:07] Yeah. And then there's a second survey from Tax Dome. Tax dome released their 2025 niche business accounting report. And there's some really interesting numbers that came out of that survey. Um, 50. So how many was this? They interviewed uh, 350, uh, small US small midsize businesses. So basically they're asking business owners how they look at accountants and accountants. So 57% of small businesses found their accountant via referral, and only 3% found their accountant through advertising. So you might think about the money you're spending in advertising with this um, and business, small businesses are willing to pay up to 25% more for premium services tailored to their industry. So if you're a niche accountant and you're doing so whatever those numbers we just talked about with the CFO Services and Advisory Services that. So $2,500 a month, you could actually charge 25% more than those those rates that we just talked about from the other survey. Um, and this is interesting. They when they're choosing their accounting firms, they are looking at tech. They want they want to be working with a tech enabled accounting firm. So people that are paying 10,000 plus annually to an accountant, 83% say that tech is the deciding factor when they choose the accountant. The tech, the accountants.

Blake Oliver: [01:09:30] Using what percent?

David Leary: [01:09:32] Uh, 83%. So? So for small businesses deciding factor, it is a deciding factor. Yes. So a.

Blake Oliver: [01:09:38] Deciding factor. So it's like the most you're saying that the that like I'm trying to get an idea like because I've always felt that tech savvy is important. But ultimately a prospect chooses their accountant based on their expertise.

David Leary: [01:09:55] Yeah. And then it's going to be it's going to be small businesses that are spending ten K a year on accounting services. Yeah. So, so they they're they probably have the right to expect have a.

Blake Oliver: [01:10:08] Good experience.

David Leary: [01:10:09] Because you're not paying $300 for accounting services. Um, firms or small businesses with a million plus in revenue are twice as likely to hire a niche specialist accountant. So as the business gets bigger and more serious, they're more likely to choose a niche accountant. And then some things that are interesting when they switch. So if they leave a specialist accountant, they're not going to a generalist. They're switching to another specialist. 98% of small businesses will if they have an e-commerce accountant, when they leave, they're going to do another e-commerce account. They're going to stay in that space. Now, some of the top reasons for switching. One is the lack of industry expertise. This is why you would switch accountants. Then the other one is poor responsiveness. So you might not have industry expertise, but poor responsiveness. Every firm can develop that and reply back better to people.

Blake Oliver: [01:11:00] And that's where this is actually something I want to work on. Note to self we don't have my fireflies meeting note taker in here, so I actually have to write this down. David. It's so painful to get a hand cramp. Ah. Note to self build an AI agent that. Oh, what'd you just say I forgot. Responsiveness. Oh, yeah. I want to build an AI agent that automatically labels emails based on, like, priority that they can look at the.

David Leary: [01:11:34] Get a plug in, they exist fix or I don't.

Blake Oliver: [01:11:37] Need a plug in. David. I can build it for free very easily. And, um, basically it'll read the email and decide like, well, a good example is like, is somebody upset? Yeah. So decide like is the is the person writing this email Angry. Upset. Sad. Like. Categorize the email based on their feeling and then escalate emotional emails to management. That's one. That's the way you deal with like problems in your firm before they become fires. And then the other is like urgency. So if there's something in the email that indicates that this needs to be done by a particular date, and that date is soon, like label that email, you know, escalate it, put it in my inbox, that sort of thing. Like agents are really good at this. So you just have to give it like examples. And just like a human assistant, you have to say, okay, well here's the definition of urgent. You know, the client saying, um, where is my tax return? I just gave you the stuff three days ago. That's not urgent. You can deprioritize that person and raise their fee next year, right? So. I mean, you could build an agent that, like, evaluates. Actually, this would be a fun experiment is let's build an agent that just watches the email inbox and makes a list of of Pita clients, and then suggests raising their fee periodically. Right. Like just automate. Automate. Watch the email communications going back and forth and figure out what clients are causing the most pain for your team.

David Leary: [01:13:10] You can have it. Just generate the price, increase email for you and send it to them automatically.

Blake Oliver: [01:13:15] Just automatically like automatic Pita client price increase. And just do it every week until they either leave the firm or stop complaining. I'm sure there's you know, I'm joking. Of course, for those who cannot detect sarcasm.

David Leary: [01:13:32] So you made a comment earlier about the UK about how you don't have free speech there. So there's one last accounting related free speech related thing. So, Blake, what is the number one thing you're not supposed to talk to your clients about?

Blake Oliver: [01:13:46] Oh, the number one thing that you are not supposed to talk to your clients about.

David Leary: [01:13:52] Well, maybe the two things, right? There's the two things.

Blake Oliver: [01:13:56] Um.

David Leary: [01:13:58] Religion is one of them.

Blake Oliver: [01:14:00] One thing. Oh, religion.

David Leary: [01:14:02] Unless.

Blake Oliver: [01:14:02] Unless that's how you get your clients. Yeah. Um, but I, I always thought that, like, it's better, like, don't ever talk about your other clients with your clients.

David Leary: [01:14:12] I always thought it was like religion and politics.

Blake Oliver: [01:14:15] Oh, okay. Yeah.

David Leary: [01:14:16] You keep the relationship at a certain level, right?

Blake Oliver: [01:14:18] Well, you know, we said that for a long time about that on the podcast, and then we started talking about politics and it's going great. So maybe we should add in religion. I don't know.

David Leary: [01:14:25] So there's a UK musician, Paul Weller. Um, I don't know him, but he's like one of the UK's most enduring and influential artists, blah blah blah. Um, he is suing his former accounting firm because, uh, Harrison Trotter. Because they fired him based on his outspoken criticism of Israel's war in Gaza.

Blake Oliver: [01:14:45] Hmm.

David Leary: [01:14:46] So the firm told Weller it will no longer work with him or his companies because he was offended by his description of Israel's actions of genocide.

Blake Oliver: [01:14:54] Where was this that he sued them in the UK?

David Leary: [01:14:56] This UK? Okay, um. Weller's legal team argues that this was unlawful discrimination against his philosophical beliefs. So they ended the relationship in 2025. He's now suing, and he says he's suing. They're going to donate the money, but he's suing not just for himself, but to protect other people's free speech. Right. But it's interesting. Accounting firms are now firing clients because something the client said.

Blake Oliver: [01:15:21] Well, I mean, if they it's like a Jewish accounting firm and the client is like criticizing Israel. And you're a big supporter of Israel, I could see you not wanting to work with that client for sure. And, uh, I would I would say like, I mean, in this country, that's I think you have a right to do that, don't you? But definitely don't tell the client that's why you're firing them. Yeah. That would that's a bad idea. That's just inviting a lawsuit.

David Leary: [01:15:44] But you could get work from other accountants, so I don't even know if there's a real case here. Right? Yeah. It's not like he's being hindered by this. Um.

Blake Oliver: [01:15:52] Well, he's just mad.

David Leary: [01:15:54] He's mad. That's what this is. He's mad. Yeah.

Blake Oliver: [01:15:56] Don't piss off your clients. Yeah. That's why, like, don't talk bad about your clients on social media. Even anonymously, I think is just bad form.

David Leary: [01:16:03] Which a lot of you listeners do. I see it on Twitter all the time.

Blake Oliver: [01:16:07] Well, I don't know if those Twitter. I don't know if those people are listening, but yeah, don't bash your clients on social media. Don't talk about your clients to other clients. And, um, I don't know, it's like you're taking their money. Treat them with respect and professionalism. And if you disagree with them, that's okay. And if it's such a big deal that you need to, like, let them go do it. But also like, don't you have air in your firm like talk to like an HR person or a lawyer, at least to cover your butt like you're an accountant? You know, if you're a good accountant, you should know to protect yourself from this sort of thing. Um. Yeah, it's but it's interesting because there's also the like the people want to work more and more with people who are personal. And so the professional like wall you put up can actually be a detriment too. So as social media blends our work in our personal lives and we share more with our, we share more between those two worlds, right? There's no longer the madman era where Mad Men era where you had your home life and your work life, right? It's all blended together now. Um, it's really hard to know what to do. Like, that's a real challenge, I think, for being a quote unquote, a professional these days. I mean, there's plenty of people who have called me out as, like, a totally unprofessional CPA because of the way that I talk and the way that I, you know, the way that I express opinions and the way I talk about the profession. And these are people who, like, show up at every event wearing a suit and a tie, and they're on LinkedIn in their suit and their tie, and they're like, acting like a CPA from like, the 1960s. And I'm sorry, but like, you think that was a great era for the profession. That was also the era where we were, like, completely misogynistic and racist, too.

David Leary: [01:18:03] Yeah.

Blake Oliver: [01:18:03] Yeah. You know, like, the profession was not a great place to be. Maybe for you it would have been great, you know, but like, um, I'd rather have it mixed up a little bit more, you know, and people be able to be themselves. Well, David, we've gone on too long and we're going to get in trouble because every now and then when we do long episode like this, we only give one CPE for it. And we actually had somebody complain once. I had to listen to an hour and 20 minutes and I only got one CPE. And I said to say, I'm so sorry we make it. You know, we had you had to listen to an extra 20 minutes of audio, but, um, you're just warned, you know, you're going to get one hour because it just doesn't make sense for us to like, add more quiz questions so you can get 1.2 or whatever.

David Leary: [01:18:46] To give you a 0.2 of a credit hour. It's so much.

Blake Oliver: [01:18:48] It's a lot of extra work. It's a lot of extra work. So don't forget you can earn one CPE for listening to this episode of the Accounting Podcast. Download the earmark app on the App Store or create your free account at earmark. Find the The Accounting Podcast channel and register for the course. Take a quick five question quiz and you'll get your CPE certificate emailed to you almost instantly. You can earn one free per week, and we should mention, David, that we now have live CPE in beta. This is for in-person events, so if you are hosting an in-person event and you want to offer CPE for it, it can be an event of literally any size. It could be like a lunch and learn. You and one other person, Even we want to test it. We will help you set it up in the app and you can issue CPE. All people have to do is scan a QR code with their phone and check in on the earmark app, and we handle everything for you. Um, I mean, there's a form you got to fill out to get the courses up there too. But like, the goal is to make it as easy as possible to offer CPE to your live attendees for your events. So if you're interested in that, reach out to us sales at earmark comm, sales at Earmark Comm. And we'll, uh, you know, we'll hook you up with a sweet, uh, price for the beta, you know?

David Leary: [01:20:18] And then the other thing, speaking of price, you probably want to try to lock in your current price for your earmark app because we are going to raise the prices on October 1st.

Blake Oliver: [01:20:27] Yes. The price today is 149 .99 per year. Lock in that price. Subscribe to the earmark app and you get to keep that price for as long as David and I feel like it. And so far, everybody who has subscribed, we have honored their original subscription price. So there are people that subscribed back when it was dirt cheap at like 80 bucks.

David Leary: [01:20:52] $79 or $99.

Blake Oliver: [01:20:53] $79. Yeah. And you get to keep it. It's like a gym membership. So lock in that price because it is going to go up because we've now got over 2000 courses. We've got amazing subscriber only content for tax and audit. Uh, we're adding all this other amazing stuff to the app. So lock in that price, support the work that we do and get your CPU done. Now it's August. You could get it all done before the end of the year if you've got that calendar year renewal. If you're like me, you. David, I have to tell you this story before we go. All right? Okay. Because this just kills me. So I'm an Arizona CPA. I moved my license here when I move from California to Arizona? One thing I did not know about Arizona is that the law requires that you earn 16 CP credits every two years that are live or webinar. They cannot be on demand. For some reason, the state of Arizona has determined that.

David Leary: [01:21:55] That has to really just get you. It has to drive you crazy.

Blake Oliver: [01:21:59] On demand. Cp is not good enough. No, you have to go sit in a seminar or you have to sit on a webinar because I guess otherwise you're not learning. So I've had to register for our competitors webinars because we don't do webinar CPE right. It's all on demand on the app. So I've had to be sitting through webinars at times that are not convenient to me, or if they are convenient, I have to choose a topic that is completely irrelevant to what I want to learn, just because it fits into my calendar.

David Leary: [01:22:27] Just so you can smash the stupid button that says, I'm here, I'm here, I'm here.

Blake Oliver: [01:22:31] And having to like, press the like I'm here button. And it just reminded me why we created earmark, because it's such a better way to learn than the traditional webinars and in-person, not necessarily in-person events, but definitely webinars. I feel like it's so much better because I can just choose what I want to learn, and it's not dictated by what's on the calendar and what's available. So I'm getting with earmark. I'm getting content that is, you know, education that is tailored to me. Whereas with these, these webinars, I have to do it's just like random stuff that is completely irrelevant.

David Leary: [01:23:09] And if you use earmark, you don't have to be distracted by a stupid comment section of a bunch of people saying, where are the polling questions? Did I miss the polling question? It's the worst. Then nobody's paying attention to the actual content because the chat's out of control.

Blake Oliver: [01:23:22] So I hope Arizona sees the light and I know changes this rule because it's ridiculous. And, um, it just reminded me that, like, what we were doing is making education better. Because when you take classes, when you're able to pay attention and you can focus, you know, I'm driving, I'm focused while I'm driving, I'm listening. I'm actually paying attention. I'm not doing emails on my computer. I'm learning. And so we're not just making education easier for accountants. We're making it better. And all this AI stuff, all this new tech stuff, the traditional education just can't keep up. So I think continuing education is going to become even more important in a world where technology is accelerating faster and faster and faster.

David Leary: [01:24:13] Yeah, your degree.

Blake Oliver: [01:24:14] Is still.

David Leary: [01:24:14] Changing. By the time you graduate. It's completely stale. Yeah.

Blake Oliver: [01:24:17] I mean, and I think this is the way there's going to be a crisis of traditional education. Is that like the four year degree cannot change fast enough to keep up with all this stuff. You have to learn everything. You have to learn so much. Not everything. But I would say like more than half of what you need to know to be an accountant. You don't learn in school. And so, um, that's just going to become more and more of a problem. And so I think we're actually going to see more people just getting bachelor's degrees in a subject that's completely irrelevant to what they end up doing, because the professional education in the traditional institutions can't keep up. And that's why I hope that our leaders in the accounting profession will make it easier for career changers to become accountants. Don't set up a system where people have to go major in accounting in order to sit for the CPA exam. Make it easy for anyone who has a bachelor's degree to sit for the CPA exam and make the exam good enough where they have to prove their knowledge, and if they acquired it on the job, or they acquired it from a CPA provider like earmark, where they acquired it from an I tutor, It shouldn't matter that they didn't sit in a classroom. So that's my hope. And with that, thank you again to everyone who joined us live here on this Friday. We will see you all on another live stream of the accounting podcast. Check us out on YouTube if you have not, you can see what we look like. You can chat with us. Subscribe. Hit that notification icon and you'll get notified when we go live, which is always at random times. You will never know exactly. Normally it's Friday mornings, but we we try to keep your calendar clear. We're working on that one to.

David Leary: [01:26:07] Keep your calendar clear the whole week. You never know. We're going to go live.

Blake Oliver: [01:26:10] We're going to figure out how how you can subscribe to our live streams. But for that to work, we actually have to start on time. And that's all I got. I'm going to stop talking now. All right. Bye, David.

David Leary: [01:26:20] Bye.