555 Using AI To Manage Accounting Firm Communications

Hey, I recently did a Q and A at one of what is now the biggest virtual conferences for accounting firms, WorkflowCon. And the theme of the Q and A was how to use AI for client communication. Like is that a thing that we should even be doing? And if so, like what are the boundaries? How do we wade into it? And it turned into actually a pretty fun conversation around the state of AI today and like where things are headed kind of just around the corner. I wasn't thinking we'd post this one on the podcast feed, but it turned into actually such a fun conversation that I wanted to share it with y'. All. It's part of a much bigger conference where you can actually go and get the recordings of all sorts of other talks from sort of a rock star lineup of thought leaders in the accounting profession. I'll put a link to how to get those talks in the show. Notes. Without further ado, check out this pretty cool Q and a around AI's place in our client communications. Gang. Thank you for being here. We got a couple questions stacked up so let's jump straight into this. What communication workflows will become standard and firms over the next 12 to 24 months because of AI? Boy, if you, if you think about like the, the biggest just pains in our hiney right now that get in the way of getting work out the door. Unfortunately I think it still is exchanging information with clients and just the old hokey pokey and how irritating that is and then having to follow up with them over and over again. This is an example of where I think like product AI can be tremendously helpful. So I think of AI falling into sort of two buckets. You have AI assistants, your chatgpts or whatever, these sort of like open ended assistants that aren't particularly connected to your tools but are still very useful for kind of specific to your workflows. And then you have product AI, like the stuff that is built into the tools that we have when it comes to chasing down clients for documents. This is where product AI will probably be helpful. Honestly, that's also one of those situations where like products are starting to solve for that even without AI. So like we don't really need to try to shoehorn AI in where it's not going to be totally necessary in terms of all of the communications that we struggle issues that we struggle with. I would say that's still the biggest, biggest one tracking down clients. So the solution is probably a combination of just good product with maybe a little bit of AI. I can tell you a Big thing that I'm excited for, for kind of the state of the profession is more help for juniors when they're still learning and are wildly inefficient. Like if you think back to your very first summers in the accounting profession and you're just, you are so stuck and you're, you're the spinning your wheels, as it were, right? And there's no good version of spinning your wheels because you either waited too long and you should have asked somebody or you asked somebody and you feel like an idiot now. And so those summer afternoons just like drag on and you're like, I can't ask my senior another question because I just asked them a question. I feel like an idiot right now. Imagine if we had like really good AI assistants when we were first starting out. Assistants that were like backed by our company's trust training documentation, assuming that exists, right. It's actually more valuable than ever before now to have really good training documentation because of AI, I would say. But imagine if that was you when you were a junior and you had really good access to advisory calls that everybody else in the firm has done, you know, and like training documentation, all that. I think our juniors become efficient wildly faster than they did when we were coming of age, which obviously that's a big unlock for all of us if we can upskill those folks faster. How do you justify the damage to the areas where data centers are located with the ease of using AI? AI is a funny thing where it's like we are going to just go headlong into a lot of problematic things and hope that AI fixes those problems before they come to a head. Right? And that's not even just an AI thing. That is a probably more of a capitalism thing. But that's for sure a sticky thing that's honestly probably way beyond my capabilities of speaking to as an accounting thought leader. But for sure there are problematic aspects of AI that the folks who are selling you the AI will say, don't worry, the AI will fix it. We just have to have enough compute and pour enough energy and investment and all that into it and then it'll fix those problems. And there's a version of that that's probably true. Like we're already seeing really cool medical stuff around AI. That's kind of the inherent nature of AI, right? The notion of like training on stuff that other people own, like the problems that it creates, but also the tremendous opportunity, like it is very much a double edged sword. There's a certain inevitability to it where like I mean, I don't know how you reel this back in. It would be the ultimate. Putting the toothpaste back in the tubening. So we kind of have to. I think we have to have these conversations for sure, but also can't totally turn our attention on, off of it when it comes to, like, how can this stuff support our business. Good question, though. How do we do training documentation in AI? All I use chat. All I use is chat GBT right now. Awesome question. Love this. So AI by itself, you just rock up to an AI model and it's frankly amazingly powerful. Like, it can do all these cool things. It just kind of knows stuff. But the even better AI is the one that goes out and, like, fetches from authoritative sources, right? And it'll, like, show you the sources in a format that you can see, and it kind of proves what it did for you. Not unlike tasking an intern with doing something. I mean, I never trusted anything that an intern came back to me with. And in the same way, I don't trust anything an AI comes back to me with. And so what we would tell the intern to do is like, here's how to go off and do this research and give it to me in a provable format. Because what I really want is the fastest path to trusting the answer that you gave me. And this is where, like, work papers and all of these things sort of originated. We need to ask the same thing of AI. It's not enough to give me an answer. You need to give me an answer that is provable. And so when it comes to what those sources of context can be, like the authoritative sources, when it comes to tax stuff, that's all pretty clear, right? You know what the authoritative sources there are. But when it comes to internal firm stuff, our expectations of team members, our processes, how we work with clients, how we use our tools, that is context that oftentimes will be proprietary to us. So AI obviously can't use your training materials if your training materials don't exist. So how far we can get with AI and how useful it will ultimately be is very much dependent upon the context that it has access to. And it's really important right now to be thinking about what context can we give our AI to make it better. Really good examples. I'm sure we're going to talk about mating transcripts a bunch here. Really good example is your meeting transcripts. Say you have a monthly advising call with a client after you deliver the financials every single month, and somebody else is going to take over that account and wouldn't it be nice if they had recordings of those and they could be like, hey, do we ever talk about this? Do we ever talk about that? Explain how the flow of the meeting usually goes. Those meeting transcripts are an example of context. So context can be like, you know, how do I use this tool specifically? Or what's the expectations and how we communicate with clients can be pretty much anything that you could imagine building into training materials. The logistics of how the AI accesses it, I mean it depends on the tool that you use. But just using ChatGPT as an example, usually that's going to be what's called a custom GPT. And then you give it that knowledge, your internal documentation, and then when you ask it questions, it only references that documentation when it gives you answers. And then that custom GPT can be shared with your team members. We're getting a little nerdy here, aren't we? But worth thinking right now. What are the sources of context that'll be an advantage for you in the future? What are the things that only you know? What are the frameworks that are proprietary to use so that if somebody's running out and asking ChatGPT the same question, they're going to get a lower quality answer. If I just work with taxidermists and I'm an expert on exactly what their problems are and what the solutions that are specific to them are, that stuff is proprietary to me. Where we can start thinking about what resources can we leverage with AI that other folks won't have access to. How do you feel about linking ChatGPT with the Google Drive or third party apps? Company knowledge looks intriguing, but security concerns me. Great question. I'm not a security researcher, I just pretend to be one on the Internet. I think it's super important with AI things to come back to very first principles, considerations around what is secure and what is not. Right now a lot of people, they just hear the word AI and they're like, that's I'm not going to test that. That's not secure. Just like kind of a 101 AI. It is important to understand when you have AI models historically, in the very early days and in the beginning with ChatGPT, anytime you interacted with it, your prompts, anything you gave it, those were trained into future models, they retained all that information and it's those prompts and those conversations that continue to make the models better. And in the beginning of ChatGPT was the only option. There was no way to work with AI where your stuff wouldn't be trained into the model. Now you could have a, we could do a whole 50 minute session about whether data being trained into a model is actually problematic or not. Because there is not a way to retrieve specific details like a database out of a large language model that's trained on billions of parameters. We need to be conservative in what we do, right? Like at the end of the day, like I'm totally fine. Like I do not. Yeah, I don't want my client stuff going into an AI model. These days we have very secure ways to use AI where our prompts are not trained into the model. So chatgpt for business, for example, very clear from the beginning, nothing you ever get is trained into the model unless you opt into it. Like there actually is a way to do that deep in the settings, but nobody's going to go in and dig that out. So then how do we feel about it being able to go out and pull stuff from a third party system? Well, there's many versions of that. There's a read only version of that where I can just go out and see information and better inform the response. Something like that. Like I don't think I have a problem with that for the same reason that I don't have a problem with the information that I'm putting in ChatGPT. Like if I trust say the security of financial sense and I'm, you know, storing tax returns in there, why, like, why do I trust that system? Like, hopefully we have some third party help that helps us validate these things and all that. I would say right now, ChatGPT for business, security wise, if the bar is as high or higher than the standard in the accounting space for security. So for me right now I don't have a problem doing sensitive things in a ChatGPT for business plan if I'm connecting it to a third party system, I think I'm okay with read access, write access is a little more tricky. Last I, as far as I'm aware right now, all of this stuff with ChatGPT is with the company knowledge. To my knowledge right now it all, it all is just read only. So I can see that stuff, I can pull that context in the conversations, but I don't believe right now it's capable of modifying that stuff. You want to talk about like how we are headed into a super weird decade right now, still very, very early on AI agents, right? And ChatGPT's agent mode is probably the best example of a very accessible one where we literally go create a ChatGPT for Business account, you toggle on agent mode and you can send it to any website to do a task for you. Kind of wild. And it's cool and it works sometimes and it doesn't in other times, and it's going to keep getting better. But how do you feel about something going out on its own unsupervised and, like, acting on behalf of you, like, doing your bidding? Like, that's pretty spooky, right? It's kind of weird. That's like a new thing that we're all having to get our heads around. What does that look like? And that will be the entire next decade, honestly, like, the agents are so good and so close to being able to go out and handle all these things for us. Then we kind of become like these puppet masters where we're like, yeah, we've created all these sort of agentic workflows and they go off and do these things for us and you just kind of trust that they will work. It's a weird thing to get your head around. I would say. It's not entirely different to do your bidding. You hire folks to do stuff for you, and boy, do they make mistakes. I would say with AI, there's more of a fear of, like, new novel ways of maybe hijacking an agent. There's ways to hide, you know, things in an image that will give a specific instruction to an AI that, that you can't see, and maybe that makes it go and do something that you don't want it to do. It's really only an issue on websites that are not like QuickBooks, you know, like, it's mainly an issue on, like, malicious websites. Are team members subject to the same thing? I mean, what percentage of security breaches happen from a team member getting tricked by phishing, you know, getting an email and ending up somewhere they shouldn't be, or putting credentials into something that looks like something that it isn't. We kind of got the same risks with AI right now, and it's still so new. People are kind of unclear on like, okay to use it right now or not. So where I'm at right now with agents is I think we need to be testing them. I'm not testing them in an unsupervised way. I'm supervising them if I am testing them at all. There are some cool use cases where it's like you just give it a spreadsheet and it actually, like, give it an existing spreadsheet of your own and have it complete a specific workflow and it can actually do that stuff end to end right now. And that's not going to the web. Like, that's not. There's nothing weird happening there. Those use cases I get excited about, but things are about to get really sci fi, man. Like the notion that we're going to be cutting these things loose to do our bidding. And you know, even back to this question, the day that ChatGPT is able to modify things in my Google Drive or the Microsoft equivalent, like, are we going to be okay with that? Are we going to be able to trust it? Will it prove to us that can be trusted? Or will it go in there and just like be like the intern that you never wanted? Right. That's kind of what I think about AI as right now is kind of an intern. On the one hand, if you had unlimited really low cost interns, that'd probably be handy, right? On the other hand, if all you did was work with a hundred interns, that would just be hell. I wouldn't wish that on my worst enemy. Have you seen examples where AI actually improves client relationships, not just internal efficiency? Great question. I can think of a ton of ways where it hurts the client relationship, right? And to be clear, I'm a big time AI nerd. Like, I love, like, let's find the best ways to use it for sure. But as someone who creates content that people watch and who gets a bunch of emails and DMs and all that stuff every day, as soon as people get a whiff that the response they're getting is written by an AI, they're not going to get in touch anymore, right? Like somebody emails me a question, they get something back and they're like, well, for one, it came back in seven seconds and I don't believe the JSON's that fast. For two, it looks like AI, like I realize I'm now talking to a bot like this kind of ruins it, right? And so professional services, you can devalue what you do real fast. So the upside here, like the most useful aspects of it are indirect. They're not usually going to be AI writing a response or auto sending a response, although there will be a version of that that happens someday that will actually be pretty cool. In fact, if you follow this through all the way, the notion of like, well, what if, what if I had an AI that could just like respond to my client's emails for me and maybe it could even be pretty good because it could see into all my stuff in financial sense, then it can see the answers, so it can give good Answers. And if it's confident, I can just tell it, hey, just auto respond. We keep following that through. What about when the client's AI agent is. Is like, okay, I'm pretty sure I can handle this. Then that one responds right away. And then our AI agents just spend all day talking to each other. What does that mean? That's pretty weird, right? What are they talking about? Are they talking about me? That's like just a really strange timeline. I can tell you right now. If I have to ask Steve for his W2 and it takes him nine days to get back to me, that doesn't feel like the future. That doesn't feel optimized. Like if there's a world where I could ask somebody for a thing and the privacy and all, there's a million considerations with this. But if there's a world where I could ask somebody for something and their AI agent just like bounces that back like that, and we can all get behind that, right? Like, that's pretty cool. That feels further away. So like the direct AI interaction, all that stuff right now feels a little further away and feels risky. Like if the client gets that vibe, that's just not that nobody wants that right now. The more today ready version of this is other useful ways that we can use AI to support our own human communication. And so we talked about meeting transcripts. Imagine if you had the last 10 years of meeting recorded. You had transcripts for all those things. You had perfect recall. You could say, man, two years ago, did they say this or did they say that? Or another member of your team has a meeting with a client that you weren't even at, and you can search across all those transcripts. That helps. Just not making me look like a dummy, right? Like I'll look like I have superhuman recall because I never have to ask clients those questions a second time. We talked about upskilling juniors. That's a big factor here. I mean, I think everybody, if they understand what the expectation is in the firm and you know exactly what the scope of an engagement is, and when something falls outside of that boundary, like that leads to more intelligent conversations and making sure that we are maybe not overextending ourselves. I was talking with Dan Luthi not that long ago, and they do quite a bit of advisory work. And one really cool thing or a question that I asked him was like, what conversations in your firm fall into the bucket of a CFO conversation versus a controller conversation versus a bookkeeper conversation? Because nobody likes doing that gatekeeping thing where you're like, oh, we're only engaged to do this stuff. And what you're asking is up here, and it's almost like you're giving them this, like, verbal paywall. And it's like, that just sucks, right? Nobody likes doing that. But they sell advisory work, so they kind of have to be careful about that, of not crossing the boundary into what is a bigger, more complex engagement. And you know what's really, really good at this is AI. So if we can define what a CFO conversation looks like, what a controller conversation looks like, what a bookkeeping conversation looks like, and then we have the transcripts, we can pull those transcripts in and be like, hey, did we, like, get outside of the lines anywhere here? Like, did this grow outside of something? What I'm really excited for is more automated workflows around those transcripts so that your manager gets pinged to say, like, hey, this conversation, like, is maybe a little bit on the edge. Might actually be a sales opportunity to talk about doing CFO stuff, right? So I think when we think about AI communications, we think about it in the direct context, the going right back and forth with clients, that feels a ways away. It's more the indirect ways that it can support us that I think are best right now. Thanks for the question, by the way. What are the areas you see that we need to have adopted AI yesterday in order to stay relevant and competitive in today's landscape? One thing to think about is what context do we need to be capturing that we could never go back and get again? And that's like meeting transcripts, right? Like, every day that we're not capturing these client conversations is the day that we'll never go back and be able to get. If we had last 10 years, that'd be amazing, right? Man, think about if you've taken, like, five years of monthly advisory calls with clients and you're like, man, I gotta scale this stuff. I gotta get out of these calls. But how do I teach my team how I advise when it's something that you kind of just shoot from the hip on, right? AI's very good at helping you, like, build your own frameworks for the first time and a more clear language and a more systematic way of how you advise. And if you had the last five years transcripts, that'd be a tremendous resource to help you build that. So definitely be thinking right now, what is the context that's slipping through the cracks that I could never go back and get if we don't start capturing it now? My favorite tools right now for this are Fireflies. Also got Vinyl, which is a meeting recorder specifically for accounting firms benefit. There is. They're building integrations with our practice management systems. So I don't have all those transcripts living in some other system. We can pull them directly into the tools that we already use. Another cool one is Ping Ping is working more on like kind of the practice intelligence side where it's trying to do kind of some more advanced like advisory sort of things around what we're talking about in those conversations. So for sure be capturing that stuff as. As soon as you can. I would say the biggest risk to firms though, a lot of people right now where it's like, how do we make time to learn AI and train our team on a. And boy, $30 a month for an AI assistant. That feels like a lot of money. People still say this and even two years ago I'd go and I'd do an AI keynote somewhere and whole room of accountants back in the early days of life and usually hadn't even touched ChatGPT and you get them all excited and they're all worked up into a lather. They can see the future. They understand why everybody's talking about AI and they're like, okay, yeah, I can actually see how this will probably change things. And we get to the end and they're pumped and I'm like, and it's $30 a month per user. And they're like, I don't know. You're like civilization changing technology like the biggest thing since fire. A lot of people say feels like a little much, right? But very big civilization shifting thing. And they're like, yep, yep, yep. Get it? Yeah, totally. I agree. Actually, then you're like 30 bucks. And they're like, maybe if it was $20 a month. The real risk here is the cost. Having a team that can't use AI three years from now. The number of roles right now that say like AI expertise is required are an important part of the job to find a job. If you employ people right now, those people probably aren't going to go out on their own, learn AI unless you encourage them to and you give them the tools to and all of that. So consider if AI continues on the path that it's on right now, where three years from now is tremendously powerful and has a huge impact on our overall efficiency. You don't want to end up in a spot where your entire team can't use AI. That's expensive. And I would say almost a moral standpoint. I don't Want to be the one that puts those people on that spot, right? Like, this is, like this is such an important thing to people's long term relevance in their career. Like, if feels like, you know, being able to type 40 words per minute, Microsoft Office proficiency. Like, it just feels like this sort of foundational aspect of being an efficient knowledge worker now. And the last thing I want to do is to be the one to stand in the way of my team developing that skill. It's a standard AI tool that doesn't solve all problems in your bookkeeping flow, like AP versus building your own custom APIs and Python scripts. Okay, okay, so, you know, we'll go nerdy. That's fine. That's a great question. I would say one really cool thing about AI that is relevant to all of us. We made a. It was a short form video that we put on YouTube a couple months ago, and it was something about somehow using ChatGPT with Excel to like write a formula or something like that. And the video wasn't packaged in a way that was specific for accounting firms. And so YouTube served it up for, like, anyone and their mom. We went way outside the accounting firm bubble. And in those comments did I get roasted by all the Excel gurus who are super proud how good they are with Excel. And they're like, well, you could have done this with a SQL query. You could have written some code, you could do all these different things. All these, like very advanced Excel things, they're like, this isn't the right way to do this in Excel when in reality, like, we're making YouTube videos. I'm like trying to bring it down to a level that's like universally helpful for folks. Right? The power of AI is regardless of where you're at right now, it can get you here. Like, it's the best, most patient, cheapest tutor of all time. And so if, if you can write Python scripts, sweet, man. We can write better Python scripts now. Like, we're doing some marketing stuff and I had to export every single email that I'd sent out of my Gmail and then filter it down like some specific subset and remove all the unique things and all that stuff. I wrote a Google Apps script because Chad Davis talks about those all the time. And I'm like, fine, I'll try it by telling ChatGPT give me a script to do XYZ. And it worked on the first freaking try, like 20,000 emails. And it gives me exactly what I wanted. On the other side, I was just following instructions and copying and pasting where it told me to copy and paste super cool. I think everybody needs to be aware of regardless of where they're at right now, like it can get them to the next level. Like it's such an incredible tutor. The question of like, do I build my own cool stuff or do I wait for other people to build my stuff matters a lot on your role in the profession. So this is coming from a guy who first started making YouTube videos for accountants about like no code automation and zapier and airtable and all that stuff. So all that like super geeky automation stuff is very near and dear to my heart. There was a time in my accounting firm when I was building all sorts of that stuff. And then there was a time in my accounting firm when I owned it. It was a 40 person firm. I was out of the client work and my job was to manage the strategy of the firm. And the last thing that I should have been doing was playing software nightjam, like building all these automations and then being the only one that can fix them when they break and all that. Right. So it very much depends on where you're at in your journey. It depends on how painful and costly of a problem it is. I've talked with folks who will have like a very niche client base and a specific really high volume problem that they have with every single one of their clients. Where it's like you got a case like that where it can save you an hour a day on like tens of clients or more for sure. Like that's a situation where it may make sense to you either yourself build that or pull in an agency that can help you build that. There are also some of us that just have to do that. Almost like we enjoy it so much. You like you just need it to get that to get it out of you. And that's cool. And that whole like self exploration of like the stuff that I'm into and learning, learning the next thing and feeling fulfilled by that, that's kind of a whole different calculation. When you look at an accounting firm and consider what is the place for that person that is doing custom API stuff and building Python scripts and all of that. For most small firms, like most of the folks who are here, unless you have someone on the team who is really lit up by that, that is usually best done by pulling in an agency. Even if you're into that stuff to a degree. Oftentimes agencies, the thing that you're thinking, they've built that thing for 30 other people already and so they can Give you the shortcut and come in and yes, it'll cost more. Right answer there depends on like what is the path that everybody in your firm is on? Does it make sense to send a person on my firm down this path? Like is this a long term thing where I want them building this stuff for us or not? If I'm a hundred person firm, it could be a different answer about whether that should be an internal person or an external person. I'd say don't be afraid to like follow the stuff that you enjoy and that you want to learn more about. But there's also a lot of people out there who are doing this for a bunch of firms that can help. What's the best AI for us to use as accountants and what is the best version security to use while not paying too much? I want to work this into my firm but need it to be cost effective. It's hard to like bucket what AI is because there's so many products that have AI in them. Right. By the way, thank you for the question. AI is built into a bunch of different things now. So it's like a meeting recorder. We've actually had meeting recorders back to well before generative AI. So it's kind of hard to say what is an AI app and what is. I think most people, what they mean here is like an AI assistant, like a ChatGPT. I've been saying this for years. ChatGPT is the only AI assistant that has always been either the best or very nearly the best. Clear back to ChatGPT first launching in 2022. Only version you should be touching in the context of work though is ChatGPT for business, which is $30 a month. I still always come back to that. So let's talk about, you know, Blue J, for example. They do tax research. It's really good. It is. My take right now on AI tax research apps is I actually can get that a ChatGPT power user can get just as far with ChatGPT, but the reality is everybody on your team is not a power user. And so like point solutions, AI apps that do one specific job, they lower the bar of expertise that's required to be able to effectively use AI to do that job. The rub is they're more expensive. And so when you go out and like you're, you kind of got to do the math. It doesn't make sense to put my team on this research platform. I would tell you in most cases you're paying for research platform licenses that were made in like the 18th century. And the research platform that most of your team is using is actually called Google, or even worse, free version of ChatGPT. So if you get sticker shock at an AI research tool like cut the AI research tool that you get, cut the research, the legacy research tool that you got and pull in the AI one, if that's just a budget swap, that's an easier argument. But desert island, I can only pick one AI app. It's gotta be ChatGPT for business. It's so good. And I would say there's a lot of people right now that are like, well, it's risky to use ChatGPT for business. You know, we just, we just don't know yet. We just don't know how any of this stuff works. You want to know what's super risky is not giving your team anything to use and then incentivizing them on getting work out the door. That puts somebody in a tough spot. Like, yes, I want my team to be honest and all that and, but you want to talk about risk, especially this whole generation of kids that are coming out of school that are freaking wizards with ChatGPT, man, just imagine if ChatGPT had come out when you were in school, right? I mean, we would have been a lot better with that thing than we are right now. So all the cheaters that are entering the workforce now, which by the way, are exactly the people that we want because they can most effectively deploy tech to get stuff done, many of those people struggle with like basic writing tasks, like knowing how to operate without chat. And that's a kind of a whole different set of problems. But you turn your back and you cut that person loose to get a job done. I mean, I would almost be shocked if they're not using the free version version of ChatGPT, which isn't a huge deal if they're not putting client information into it. But I almost see it as more risky right now not to put a secure AI in your team's hand just because of, for some people right now, there's such tremendous time savings that that's a really hard thing to not use. And so if somebody believes that they can cut their eight hour workday down to six hours if they use an AI tool and you won't give them access to it and they got other stuff going on in their lives and a whole lot of reasons why they want to be done working in six hours instead of eight, that puts somebody in a tough spot. Like at the end of the day, that's still a problem. They shouldn't be doing that. But that's a great rationale to me to make sure that everybody has at a minimum, like the business version of ChatGPT in their hands. Megan Mason, can you give us a summary of how ChatGPT agent mode can be used for typical accounting or bookkeeping firm tasks? I'm not the creative type and want to know how people are using this smartly. Megan, I love that question so much. So we got a bunch of videos on this that will go through and like actually show you the best ways to use ChatGPT and all of that. And obviously there's only so much of that that I can do talking into a camera. But some cool examples of agent mode. Specifically, the agent mode stuff that I'm most interested right now in is offline. It's not the stuff that's doing stuff online because I still got issues with trusting agent mode to go browse the web and complete a form or do whatever that looks like on the web without me supervising it. Because at the end of the day, I'm still accountable for what it end up doing. And if I'm not babysitting it, like I'm just not there yet. There will be a day when saying that makes you a Luddite. I don't know if that's a year from now or 20 years from now, but at least right now, like I'm just not there yet. What it is amazing for is spreadsheet stuff. It's funny how all roads always lead back to spreadsheets. Why is that? If you simply enable agent mode, if you're asking ChatGPT to do anything with a spreadsheet, if you enable agent mode, it is much, much better with spreadsheet, it like takes a big step up. And the best way to use it with spreadsheets is not asking it to create a spreadsheet from scratch. It is to give it a spreadsheet, like a work paper template or something like that and have it complete the work in that spreadsheet. So I'll give you a good example. Super cool thing that this Dell1040.com, it's totally free. It's this ridiculously huge Excel workbook. It has 60 tabs and it rebuilds like a whole bunch of 1040 sketches schedule. And so there's all these input tabs that are just like the input screens. And a tax on bookkeepers are like, get me, get me out of here. Don't worry, I'll bring it back around. Bookkeepers. This Giant workbook has all these input sheets and it's just like your tax software inputs. And then on the other tabs they're like, you know, 1040 page one. It like creates what the 1040 will look like. So you have this big old workbook with 60 tabs. You can give that Excel template to ChatGPT, enable agent mode and give it like tax docs. And if you're uncomfortable with that with your clients, like just grab some example ones off the Internet. You can give it a pile of W2s, brokerage statements, 1099 interest, like government forms. We're not asking too much of it. Well, in some ways we are. You can give it a pile of PDFs and it will take all of the information from those PDF and put it into all of the correct tabs and all of the inputs fields in the workbook by simply telling ChatGPT, take all these source documents and enter the data into the Excel workbook. And it does it really good in like less than five minutes. A more bookkeeping related example that we did in a video, we took a payroll journal export from a payroll platform for a 30 person pay run. And you know, a payroll journal like CSV Exports look like it's just, it's a lot. And like if I hand that off to an intern to try to make sense of, like, good luck, you know how dense those reports are. And there are umpteen different columns for, you know, the 12 benefits and the employee paid portion, the employer paid portion and the local tax. Like it's just, there's column after column after column. Big ugly report. A pay run for 30 employees. I gave it a spreadsheet that was like a payroll allocation template, right? Like we all have some version of this and we're not proud of it, but. But we do because the owner wants to allocate all 1 of all 30 of these employees, like oh, 20% of this department of 5% to this department. And so it just makes like the payroll adjustment an absolute nightmare, right? Whether it's payroll, whether it's something else, whether it's Craig wanting to split the AMEX bill between six different businesses. Like you have a version of this spreadsheet and I don't believe you if you say you don't. I gave ChatGPT this spreadsheet, but then I gave it the payroll journal export for that pay run. And I said, enter the data from the payroll journal into the parallel allocation spreadsheet. In four minutes it entered over a thousand values into the parallel allocation spreadsheet. Like performed that task perfectly, taking all the data from one and putting it in the other. Now to be clear, that's an elegant solution for a problem that shouldn't exist in the first place. But isn't that what accounting is right now? Like, people think we do repetitive, monotonous things. It's like we don't. We already automated those things. Accountant's job right now of doing all the crap that falls between the cracks, like all this stuff that there isn't a perfect solution for when that system doesn't talk to this other system. Like, that's just, that's kind of what being an accountant is, is being capable and having the expertise of facilitating stuff and automating it as much as we can and applying our expertise. When you're taking that thing from a to balance, like, that is accounting these days. And that's why engineers don't understand what we do. So to be able to, like, if you zoom out on this and like, what are all the ways that we can use agent mode right now? I would say it's most powerful at spreadsheet based workflows and can genuinely do over a thousand entries in our testing in less than five minutes. And it's weird. It kind of makes the spreadsheet cool again. But if you have like work paper templates that are in Excel, start there. Like, it can help you prepare those things. It's really, really interesting. And so I'd say right now, if you want to play around with agent mode more. I love using spreadsheets because it's not doing stuff online, it's just modifying your spreadsheet. And don't ask it to create a spreadsheet from scratch. It doesn't do that very well. And also your expectation of what that should look like will always be different than what it actually does. So instead, give it a spreadsheet to work in, maybe give it some other information and then give it a job of how to complete that workbook. You can even protect the cells of the workbook that you don't want to touch. And you will be shocked at how good and how fast it is. What is next? And Jason Stats the journey. You're absolutely crushing social media in many ways and I love what you're doing with taking yourself on the road. Where do you foresee your journey going? That is a great question, Yuri, and if you could tell me, I'd appreciate it. I don't know, man. I believe it was Bendy's the one Fast and the Furious who said, I live My life a quarter mile at a time. Honestly don't know. I can tell you right now I'm in Phoenix. We are on the, the third stop of a 14 city tour hosting our own live events. And it's amazing. We had like 65 folks out last night just talking about running accounting firms and the sort of things that you can only vibe on with other accountants in person. So we've been going hard on live events lately. We will definitely be doing more of that stuff. You can only stare down the barrel of a camera for so many hours before you start going, what am I doing with my life? And I need to talk to another human being. I have three very little human beings right now. A four six and an eight year old. So they take up a lot of my time. I do think with the online audience that we have now, I think the coolest thing that we can do is create amazing in person experiences for accountants. Like, I love the online stuff. Obviously my whole business is the online content and virtual conferences like this. And that's all really, really cool. But if we're like uniquely suited to do one thing, I think it is like pull people out to a live thing, have a really cool experience. Experience, just make those connections with other firms. Like that was the most powerful thing that I ever experienced in my firm running journey. We burned so much freaking money on coaches and training and all these, you know, advisory things and all that. And then when I sat down with a bunch of other people that intimately understood what I did, because they do it too. And you can throw those questions at them and all that, there was nothing better than that. My wife, incredibly supportive. I get home at the end of the day and she like, she'd be like, man, you're the best of this stuff. You're so, you're, you're doing great. You're, you're so good at this stuff. And I'm like, I appreciate it, but I don't. You don't know, you don't know anything about running calibers, but I appreciate that. And so like that support, it is all, it is all very important. But like getting in a room with a bunch of other people to do the stuff you do, there's something that just like recharges your soul about that because they can relate to what you do. And so if I can create that experience for folks like that's, that is the thing that I think most lights me up right now. This is probably the last one. AI Agents, is there a lot out there right now in terms of Answering service, website agents, et cetera, that can help with admin based duties. That's a great question. So there are kind of some AI native workflow builders. So like Zapier actually does this pretty well right now. So you have like workflow builders that can kind of make custom things like Zapier. One struggle with what you're describing and you know, admin duties and website agents and customer service and all of that is that Bridge is like silver, many different apps and everybody does that so many different ways that it's very hard to build a platform that will do it in a universally relevant way. So this is where I think this is actually the golden age of agencies of being able to pull in an AI agency to support your firm. If you have high volume, annoying workflows that you would love to have be more automated. There's now a growing number of groups and really talented people that you can pull in to either train your team how to do that or to build it for you. And man, there's a part of me that's like, I just had all the time in the world to just geek out on this stuff and make all these cool things and. But you grow up and you end up with grown up responsibilities. There is a really cool opportunity for talented people right now to come out and support accounting firms that all need help with this stuff. Right? Because we're all busy and we're struggling to even make the time to learn this stuff. In most cases, the solution for what you're describing is probably going out and hiring somebody talented on a a contract basis to come in and help your firm build what you need. Thank you so much for being here. Great job. To the workflow con stuff. I've heard so many good things. Y' all are crushing it and I hope this is valuable for you and I will see you in the wild soon. Take care.