212 I Just Cloned My Accounting Expertise
4 Apr 2024 · 30 min
Podcast episode (Jason Staats) describing his experiment training an AI avatar on ~200+ episodes of his podcast (automatically ingested transcripts) to create a searchable, advisor-style clone. He demonstrates a sample query ("I'm overwhelmed — how do I get better clients?"), explains the underlying retrieval-augmented generation approach (the assistant cited multiple "chunks" from past episodes), and walks through the product he used (Delphi.ai): its data inputs, customization sliders, tone prompts, pricing tiers (free → Prodigy ~$400/month), scripting, monetization, integrations (CRM, Zapier, SMS, embeds), and limitations (voice cloning quality, model backends). He situates the tech for accounting firms — potential advisory augmentation, new lead magnets, automated triggers/alerts, and revenue-sharing/paid-access models — and issues practical advice: start documenting (podcasts, meeting transcripts via Fathom/Fireflies/Microsoft/Zoom), capture advisory conversations, and experiment with embedding chatbots as client-facing tools. Sponsors mentioned: Interval, Team Up, Live Flow, Tech Guru. Episode teases a follow-up Q&A where the assistant will be tested live.
Chapters
I just trained an AI on 200 episodes of my podcast. Jason, why don't you go ahead and introduce yourself? I'm going to have you on as a guest for the podcast today. Oh, hey there. I'm Jason Stats, your go to guy for all things small accounting firm related. I'm here to share some nuggets of wisdom on running the show in the accounting world. How about we dive into the nitty gritty of creating a killer client experience? That's something I'm really passionate about. Okay, so he doesn't sound a whole lot like me yet, but boy, does he know a lot. I've literally pumped everything I've said on this podcast into an AI avatar. It is a fascinating glimpse into the future of can you really create an avatar of someone's expertise simply by documenting all the things they ever say? And the answer is. The answer is. Find out after the intro jingle. So I'm going to run you through how this tech works, what it can do, what it can't do today. It's really fascinating and show you some very, like, actual real examples of me posing questions that people frequently pose to me. See what the answer is and how it got to that answer. So just real practical, off the top. I'll give you an example. I threw the question at it. I'm overwhelmed with how to get better clients. That that was it. Here's what he has to say. Oh, boy. Getting better clients is like the holy grail, isn't it? First off, breathe. It's a common struggle, but not insurmountable. You've got to lean into new media and specificity. Traditional word of mouth is fine, but it's not going to change the game for you. You need to be where your ideal clients are, whether that's on social media, in online communities, or through content like newsletters and podcasts. I mean, define who you want to work with in terms of work style, expectations and the problems you solve for them. And remember, it's about attracting clients who value what you do and are willing to pay for it. What strategies have you tried so far to attract better clients? If you have listened to this podcast very much before, you are convinced that was pretty darn accurate, right? Don't get caught up on the voice reproduction quality. This tool isn't that great at it. In fact, I could go out and make a voice clone on Elevenlabs that would sound way better than very good. Honestly, indistinguishable version of a monotone version of me talking. Ryan Lazanis. I don't know if he still does this. For a while, he was literally putting out his podcast with an AI generated voice of him. I like to tease him it's because he already sounded like a robot. I get pretty animated when I do this, and AI can't do that. So it could do a monotone version of my voice. So don't get caught up on that here because it's not great at it. But the text of that answer, honestly, that's every bit as good as I think what I would have put out there. Now it's obviously a huge question. I'm overwhelmed. How do I get better clients? Well, where do you even start with that? But it also ended this with a follow up question. What strategies have you tried so far? And honestly, that's probably the best place to start. But I'm gonna break down the answer that it gave for you to show you just how powerful this technology is and how it got there. So in this one paragraph response, there's three citations. By the way, it responded in a single paragraph. Because I have set the app to on average provide X words with each response. So I've dictated how long I want the responses to be. But within this paragraph, there's three citations. The first citation comes midway through it, and it takes me directly to the sources for the first two or three sentences. The sources are eight chunks from my podcast, each chunk. Chunk's actually like a technical term they use for this type of retrieval technology. Each chunk is about two or three paragraphs actually longer than the entire response it gave. So it pulled eight chunks that were about two or three paragraphs in length just to form the first two or three sentences. This technology is called retrieval Augmented generation based. Basically, it searches through that entire body of information. We're talking about 211 episodes now of 30 minute podcasts, five to 7,000 words each. It generated this answer in less than 10 seconds. And just this first citation pulls out eight chunks across all those episodes. There's two subsequent citations as well. They each pulled eight chunks also. Now zooming out. That's really cool tech, right? Like, wow. Like I just did that. And it instantly pulled up all of that information. And in my testing, it's pretty darn accurate. Like it hit on all of a whole bunch of words that I'd say. I mean, said, oh boy, it really likes to say buddy. And it's actually kind of funny. One of the things you find as you stumble into a novel thing like this is maybe new limitations you weren't expecting. Like when I see my spoken word in text. It's really weird. I write very different than I talk. I don't say oh buddy when I start sentences, like writing it down. I say that all the time when I'm like palling around on the podcast, just talking, but stuff that it hit in the first three sentences. New media specificity, traditional word of mouth is fine, but it's not going to change the game for you. Social media, online communities, newsletters, podcasts, those seven things were in that first block of citations. And most of those things are not what other thought leaders will tell you. Or if you just turn up to ChatGPT or something, it will not say those things. So this isn't general advice. It's very specific to the things that I say. New media. Have you ever heard anyone else in the accounting profession say new media? Like, obviously it's pulling all this vernacular to sound like me in a very convincing way. So what is this beyond a razzle dazzle thing for influencers? Because the way that this company specifically is productizing is they think that people will ultimately pay for access to these things. And the only way that's not true, this company makes money from you paying them a monthly fee. So I'm paying them $29 a month. But they also give you the tools to set up paid access in different usage tiers. For this, they take 20% off the top at this tier. At the higher tiers, they take less and I take the other 80%. Like, if there's actually a market for this thing, for people to come and talk with it, that's how this company makes money. And then ultimately I make money off of that. And I don't know if that's. I don't know if that's actually where we're headed, if that's like the future of AI chat tech. So let's talk about what does this mean for accounting firms? Because we've done entire episodes in the past about how like the age of the AI accounting and tax advisor and cfo, like, man, it's here and people are gonna make these things. And honestly, they're gonna be good. Like, they really are. Like, the level of help you'll be able to get by uploading a copy of your tax returner or connecting it to a QuickBooks file, something like that. It's gonna be pretty darn good. Especially if you compare it to the current solution of like Googling a question. Like, the point of entry for professional expertise right now is so high and so expensive. There's a Massive gap between Googling things, Investopedia, like wherever you're gonna get your answers, and true professional help that's putting it in the context of your company's numbers and, and that's like a huge void that a productized solution is going to come in and fill, or at least do some version that's better than Google. I'm sure there's some out there right now. And the reality is like, they'll get good and those things will be compassionate. Like they will be many things that actually accountants and firms will not be. So there will totally be a place for them. And I'm not even like, necessarily against them. They're just a new thing. I don't think that they're a replacement necessarily for some of the things that we do, but they're a different thing. And that's where maybe it gets hard to know how should we feel about them, right? But as you know, let's say Intuit really nails this. Let's say they build it into QuickBooks or into TurboTax or something like that, and they have versions of this in the product right now that frankly suck. Like, they're not good, but they'll get much better. But let's say they nail it. And the upside for clients is they have this thing that they can get instant answers from where it's not as good as talking to an actual human advisor, but it's much more available and much cheaper. This episode is sponsored in part by Interval. Interval helps accounting firms unlock opportunities and optimize results for your SMB clients. This one's pretty cool. It's basically like a connector that will then surface automated insights based on the ledger connection. So with more business owners to serve and less time to serve them, Interval is the conduit to information and insights that put your advisory practice on ahead of the curve. Here's the curve. How do you get ahead of a curve just from a. Just from a geometric standpoint? Is that like another curve with a bigger radius? Through a quick connection to your client's financial data, you get access to automated data discovery and advanced analytics. INTERVAL delivers actionable insights on valuation, industry benchmarking, excess, working capital, tax, advisory opportunities, and more. Filtering out the noise and highlighting only what you need to spark advisory opportunities. Interval provides foresight into financial forecasts and overall business health while protecting and growing SMBs. Creating a shared understanding for. Because when businesses grow, so do you. Man, I wish I had stuff like this when I was running my firm. If you've got access to the ledger like that is a treasure trove of opportunity and good advice that you can give people. I was always like, if only we could like wire up all these ledgers to specific service packages that we have where anytime this triggers, we propose this package and we can set it up in a way where it's like a no brainer. We're like, hey, we see this thing, you should totally have us do this. And they'd be like, yeah, but where did that fall on its face? Like monitoring the ledger data to actually make that connection. Like, and I got cool tools like Interval to do that for you. We're still in an ad read. Learn more about Interval in the show notes below. This episode is brought to you in part by Team up, helping you recruit top Filipino accountants without any ongoing monthly fees. The difference between Team up and all the other offshoring options. So Team up helps you hire staff directly. No middleman. You work directly with your new hire in the Philippines. Hire the person, not the company. Guys, gals, gang. Here's just a few reasons to hire directly. You have access to higher level talent. Makes sense. You have complete control over team culture and training. They keep 100% of what you pay them and it's a lot more affordable for you. So you can retain your team for the long term. Team up can source accountants with experience working at US or Australian firms familiar with tools like Zero, QBO and DeX. Also recruit specialist roles, team leaders, tax specialists, administrative assistants. Thought experiment. What if you had an executive assistant for the first time this tax season? Just. Just throwing it out there. What would they do? Start at that email video I did on the main channel recently. Get help with that stanky old inbox. I digress. Team up recruits these talented folks for a flat one time fee of 4,000 US American dollars. That's it. 4K one time. Somebody at Robert Half just did a spit take. Robert Half reference. We ever gonna get Robert Half to sponsor this pod? Not anymore. And they can connect you with an affordable employer of record if you need help with payroll and compliance once you hire that person. Big fan of hiring in the Philippines. You know I did a bunch of that. Check out the link in the description to learn more about Team Up. And so where we've explored this in the past on the podcast is is there a future where there is an avatar of your own own expertise that clients can get access to? Probably for a fee, maybe as a lead magnet, but probably for a fee. And so I'll put that question out there. I don't have an ultimate answer for it. I just think it's an interesting question, and I'll walk you through how this tool works, because I think when you understand how the tool works, you see different ways that it could actually be plugged in to an accounting firm. And it's really interesting. So the tool that I'm using here, and there are actually a number of, like, create an AI clone, types of tool, type of tools out there, and we're seeing explosion of just a ton of them. This one is called Delphi D E L, P H I A I. The setup process goes much deeper than other tools I've shared in the past where, like, you upload a document and now you can chat with the document. This tool goes well beyond that in some really interesting ways. So you train clones as, like, their vernacular. And each clone has, you know, a wide, like, huge amount of different settings you can use to tailor it. But the big thing is what information are you training it on? And so right now, the only thing that this, this clone knows is everything that I've ever said on a podcast, which is amazing, but you also got to think about what's in there. So, like, ad reads are in there right now, right? If you do a podcast, it's conversational, people talking back and forth. Maybe that gets a little bit more complex, but really cool. Right now, it can pull from a massive volume of information of the things that I've. All the things I've ever said in this podcast, and pick out very specific little things now going at like, a level of depth past that within the things that you train it on. For every one of those items, I can actually adjust a slider that impacts the importance of that document. So, for example, some of the AI stuff we've done early on in the podcast, I slid that document importance way down because out of date now I've got a global setting enabled that biases it to fetch more from recent stuff because my expertise, the way I talk about things, the way I think about things, has actually changed from the first hundred episodes to the second hundred episodes. Right now, this is just connected to a podcast feed and it went out and it fetched all that stuff and got all the transcripts automatically. A new podcast published this morning as I'm recording this, and it automatically pulled that in. So every time a new one publishes, it pulls that in. I could connect it to my YouTube channel. I've actually decided not to because I think the format of that is so built around the visuals that the text only might be a little confusing or at Least isn't as good as the podcast. And there isn't really anything that we do on the YouTube channel that we don't do in even more depth on the podcast. So it's pulling all that stuff in automatically. Each clone has a purpose that you define. So the purpose that I set for the clone is to serve as an advisor for individuals running or aspiring to run small accounting firms. So that drives the way that it responds to questions. It's trained on my voice from the podcast, like we said, like we talked about. Not great right now, but not a big deal. It has a prompt that defines the tone of voice. It uses like, not like speech voice, but when it responds in text, how does it sound? And it's actually automatically generated a tone prompt based on all the, all the things that I say in the podcast. So it mimics the voice that I use in the podcast. You can add custom instructions. Avoid talking about this or that frequently, bring up this or that custom suggested questions. So when a user comes in, there will be a couple little bubbles where it's like you suggest a starter question to kind of, to kind of get things going. Really cool. But I'm going to hop over to the pricing page because I'm doing actually very basic, like entry level things with this. And when you look at the pricing tiers, you can see sort of what their ambitions are for this product and all of the various different ways you can use it. I'm definitely just doing the simple stuff right now. By the way, I covered this in depth with some visuals and all of that on my newsletter today. I can't get like too detailed with visual stuff on the podcast because it is a podcast. But if you want to go deeper, check out my newsletter. Newsletter. Jason, cpa. You can read today's edition without even having to subscribe. So on the pricing pages, there's basically four tiers. The cheapest one is a free tier. The biggest tier is called Prodigy. It's 400 bucks a month. They each come with like volume limitations. So number of messages, number of call minutes. Call minutes is like the speech if you're actually talking with them. Now, on the upper two plans, you can define scripts, which is like guided paths. So for example, let's say I work with beekeepers and anytime they start talking about, you know, converting to an S corp, I want to walk them through an explicit set of things, almost like a form, but conversational. It will trigger a script and take them down that path. And I can then go and get that information, like basically get like the same output you would from a form builder. So you can pull scripts into this. You can create different instances of a clone on these higher tiers. So ones that are tailored maybe a little bit one way or another. So for example, if you work with three different niches, you could have three instances of a clone each which are kind of fine tuned to do a little more for that specific niche. So you have like three slightly modified versions of the same things on the upper two tiers. It will collect common questions and extract those. Man, if you want to do a weekly podcast and you've got people using this, maybe this is like a lead magnet that you put on the website, people can chat with it. Again, specificity really helps here. But if you can extract the most common questions from that chatbot and the weekly podcast is just a Q and A on a question or three questions every week, something like that, man, that would be gold. All the paid plans support monetization, so you can charge for access at varying limits, like number of messages, that sort of thing. Different pricing, whatever price you want to set, be it $5, be it $5,000. Interestingly, you can also do affiliate products in here. So you can link to like sign up for your own service or buy this thing off Amazon with an affiliate link and it will actually plug those things in the course of a conversation. I don't know how I feel about that, honestly. I get, I get like, affiliate is big business. I get the value of it. You can extract popular topics, you can set up alerts anytime someone wades into a specific thing. Maybe they started talking about cost segregation studies and that's something you're pushing within your firm and you want to be alerted anytime somebody brings that up so that you can reach out to them directly. You can connect this with like mass messaging services like sms, your Instagram dms, sync it with third party CRMs, embed it on your website. The top tier lets you white label the embed. Zapier, telegram, Slack, discord. They've got an API this episode. It is sponsored in part by the fine folks at Live Flow. Tell me if you've ever had this situation before. Somebody needs some information from the QuickBooks file. Maybe it's an advisor, maybe it's like the financial advisor, some other stakeholder, a board member, maybe even the client. But you don't want that person to have access to the actual QuickBooks file itself. Or maybe you're at like number of user limits for your QuickBooks license. Man, I had this all the time where I Wanted to give them like a subset of information but oh lord knows I do not want them to have full access to the book books Gang is actually a really good use case for Live Flow because you can selectively sync any little detail out of the QuickBooks file to Google Sheets situation where we actually had this a lot. You've got somebody within the company that like manages ad spend but I need them to tell me like what events certain ads are associated with and they need that data for their own analysis. But you can't give them the full P and liff Flow can be a really great solution. Solution here to just get the information out of there that they need to see without getting them everything else. Now Live Flow does a whole bunch of stuff. Syncs data out of your QuickBooks file to Google Sheets. Can be a one time sync or it can keep updating that stuff on an automated basis. They've got a killer consolidations product now. We can roll up theoretically an unlimited number of companies with different sets of chart of accounts all in the same like really simple, easy to use liflow interface. People are using this for like Live Flow for everything from like managing month end closes, even internally. Like they set up all their work papers in Google Sheets. Kind of wild. That is Live Flow. Learn more about that one in the link in the show notes. This episode is sponsored in part by tech guru. Pop quiz. Your computer just explodes right now. What do you do? A, we're gonna go down to Costco and get the best deal on a new tower that I can. B, call my IT group and let them figure it out because they're the professionals. That the correct answer is B. He thought there was gonna be four options. I'll be honest, when I started that I thought there would be four options too, but I couldn't think of anymore. Gang, we talk about delegation and the stuff that you ought to be worried about and stuff that you really need to get off your own plate. IT tech. Please just stop, okay? There's professionals for this sort of thing. I don't want Tina spinning up her new Etsy shop, going out and filing her own tax return. But computers, you got computers figured out, do ya? You just don't. Okay? You just don't. Tech guru. They just work with accounting firms. They know the stupid apps we have to use. They know the great apps we get to use. They know the, oh the cyclical natures of our business. So they're gonna get up off your back during busy season. But yeah, no, I'm sure that you know More about how to manage it than a company that literally just has a big old team of people that just talks with accounting firms about, about all their tech stuff all day, all night, deals with it nonstop. But yeah, no, I'm sure that you know better than they do. Why am I so upset? Check out the link in the show notes to learn more. We're getting to deadline season. If you don't have a good plan for it, just book a call with them now so you got something on the calendar to force yourself to circle back to it and put together a better plan going forward. So loads of different ways that you could use this. Obviously, we don't want people taking advice from the AI. So, like, we're gonna have disclosures all over this, just like you see on all of the chat services out there. But for the concern of, like, oh, this is gonna replace me. Like, no, it's a new version of access to you. I think, like, and this hasn't been normalized yet, but I think this is where we're headed is it is a new version of getting access to you that isn't as good as the real deal, but it's better than nothing because it's instant. Because so many of the questions clients ask are the same questions over and over again, and they probably could have gone out and got an easy answer somewhere. Some questions, like, need to be answered in the context of their situation, but a lot don't. And honestly, the ones that don't are all the same information that we're putting out in blog posts and podcasts and all that. And that's the beauty of this here, is we don't want AI to give a black box answer, but it's not doing that. It's giving answers according to all of the stuff you've already said, said. That's the beauty here. If you have a blog, if you have a podcast, whatever it is, you connect all those things to it, and all it's going to do is regurgitate the things that you've already said and point them to the sources. And this is an example, as we talked about many times, of how AI enables a totally new version of search, like semantic search on a massive, massive scale. Now this tech's going to continue to get way, way better. They're using, it looks like GPT 3.5 and GPT 4. Behind the scenes, those are the language models they're using, which are no longer the best models for this type of thing. There are better models that are more expensive that do this even Better. But the main point here is, you know, we talked about Claude on Tuesday this week that how that does this stuff much, much better and is actually a really big step from GPT4. The main thing here is we are, I mean, really, really close to this, continuing to get much, much better. Like each quarter, like as they pull in different models and models improve, improve. And OpenAI is going to get us a new model this year. This is good today, but it's going to get way more compelling quickly. Like, it will just keep getting better. So what do I do about this today? This advice is going to sound familiar. Start documenting. This is all about documentation. One of the reasons I started this podcast last year was for exactly this tool, because this was coming. We were, you know, a few months after ChatGPT and everybody's still trying to use ChatGPT as a knowledge base and just trying to get it to like make up answers for them. And it's like, no, the future of this is when you provide the answers and this is just the chat, like the homogenization of all of those answers, of all of those things that you've already said. And so start thinking about what all the sources of those data that data could be. Maybe you have a blog, maybe you have a podcast, maybe you don't. Is there like something you're already doing that could be repurposed into expertise based stuff like client meetings? I mean, how many meetings do you take with clients where you're giving them advice every single day and the only person that ever sees that advice is that client? Like, that is kind of the sawdust of our business, where maybe there's a way to sell the saddest of your business. Maybe there's a way to like roll up those advisory conversations and turn them into something valuable. Like maybe you're using this tool internally and it's helping you build podcast outlines or write social media posts or something like that based on all the data that you're training it on? Myself and my team will absolutely be using this to pull up. Have we ever discussed this? Lord knows I don't remember, man. We've done so many of these things. Or what are some adjacent concepts to this thing where maybe we're doing a podcast episode and I'm like, have I thought through this all the way? And I'll kind of drop the outline in here and be like, are there adjacent things to this that I haven't covered? And it'll be like, well, in this episode you cover this. This episode you covered that, man. That's super helpful. The whole notion that, you know, you already have everything that you need in your head to write a book, it's just not in the right format yet. That applies to so many things. Maybe somebody asks you to go give a talk at some little business meeting. What are like the knowledge sources that would get you to that being a good talk? Have you captured those sources of knowledge of you talking about those things, your expertise? Can you combine them and then prompt out an outline for how you could give that talk? The bullet points slide by slide. This stuff is so going to change the delivery of expertise. And I feel like we're at a tipping point where it's good enough but not yet normalized. So folks don't quite know like how and where to plug it in yet. You know, we get, we do Q and A Friday every week. We're going to do it tomorrow and we're going to throw the questions at this as I answer them and I'll kind of assess, like, is that what roundabout what I would have said or not? That'll be really interesting to see. The notion of like a co host, that is me, or actually not me, a co host. That is everything I've said to date. Right. And does that end up being accurate or not? I mean, that's a fascinating thing to look at, but I mean, I get so many questions every single week. What we don't know today, and this applies for me just like it applies for you and your clients. What we don't know today is if there was a place where people could go to ask that question and get an immediate answer, would they go there or do they want the answer from the human? Like there is. Is there like that interpersonal thing there? Would that apply to some questions, but maybe not some other questions? Is it useful just as like a search function where you're gonna drop it in there so that you can jump to the episode where it's discussed? I don't know yet. But I'll tell you what, that's pretty darn mind blowing, like how much it sounds like me, it talks like me, it gives the same answers I would, and in seconds fetches paragraphs from a hundred plus hours of content of things that I said. And honestly all I did was I connected the podcast feed. And so the reason it works really well for me right now is because I'm documented. So how documented are you? That's what I want to keep coming back to is how can we begin documenting what we're doing better AI Meeting Transcript Tools Just start, Just turn it on. Fathom is a good one. Fathom Video Fireflies AI. If you use Microsoft, Microsoft Teams Premium, they've got built in transcript stuff. I think Zoom has some transcript stuff. Start capturing meetings. Start there. You can literally repurpose those into like an outline for your next weekly newsletter or a podcast recording or anything like that. Start collecting that stuff. Because the tools that we have to mine those transcripts are getting really dang good, but are only ever gonna be as good as you have documented things. When AI models are super, super powerful, they only end up being as good as the data that they have. And how do I fight back against that? Is having data, having my expertise documented. These days when it comes to accounting for marketing, it is not a high bar to like dazzle someone and put something out that's like, wow, that's pretty impressive. Like, you know we talked about like including a chatbot with like a monthly financial statement delivery, like things like that, where it's like, yeah, this is kind of useful. Maybe the most useful thing there is that you're signaling that you are a forward looking advisor and not a run of the mill accountant who is still like stuck in the past. Right? So consider, is there a way that you could use this as a lead magnet, as something that you just give your clients access to? Are there useful things you can extract from those conversations? Like the most common questions, setting up triggers to like notify you if somebody wades into a specific thing where you may want to like follow up with them. Really interesting. This tool, specifically Delphi, kind of opened my mind to just how far you can take something like this and start building things into the conversation in an interesting way. Pretty wild stuff. You got any questions? Drop them in the comments. Tomorrow we're doing a Q and A and we will bring this assistant along for the Q and A to see how it stacks up. Because not unlike the way that I answered questions for listeners, like a huge part of being an accountant is sitting there and responding to emails and answering questions all day, every day. And if this will answer questions the same way that I do. Oh boy. Isn't that technology that you would like to have, like built into your practice management system, like in all these different contexts in how we run an accounting firm? Right. Is that a tease? Come back tomorrow. I'll see you there.
More from Jason


622 Accounting Firms: Watch This Before Selling Advisory [What The Industry Isn’t Saying About The Future]


621 The 9 Most Asked Questions About Accounting Firm AI [AI for accounting firms]
620 The Real Reason Clients and Staff Resist Change [Here’s what to do about it]
619 The NEW AI Agent Approach for Accounting Firms [An entirely new AI agent platform for firms]
618 Accounting Firms: Stop Taking New Clients. Watch This First. [How to know when you’re digging yourself a deeper hole]
615 The First $1B Accounting Firm Rollup [What this means for small accounting firms]
614 The Three Types of Accounting Firm AI [And where to deploy each in your accounting firm]
613 Stop Writing SOPs for Your Accounting Firm. Do This Instead [The NEW way to build your SOPs]