610 Most Accounting Firms Are Using Claude Wrong [5 Steps To Automate Agentic Work]

My friends, we are in the thick of my workshop tour right now. I'm visiting 20 cities in the US, throwing half-day parties, continuing education, snacks, a dinner, an open bar. These are events that I have built with my team from scratch. We host them and we pull in 50 to 150 local accountants who all want to hang and hear a couple of talks from me and meet each other and just vibe, you know. These are unbelievably fun events. We are in the middle of running them right now. If you haven't checked out the tour schedule yet, please do head over to JasonOnFirms.com/tour. /events to see the list of all the cities. If you know somebody that's in one of those cities, give them the heads up. It's a super fun hang, and I want to see you there. Okay, on to the show. Enough about your workshops, let's get on to the actual episode. If you've done much with Claude inside your accounting firm yet, you've probably had some pretty novel back and forth conversations with Claude. Maybe you've built some But what does it look like to take that capability and have it run 24/7? Just have something happen when a project enters a status. Isn't that what we ultimately want to get to? Not having to have a back and forth conversation with the AI, but the AI to just do the work. That is what we are talking about today. The progression from the live conversation all the way to the, the AI that is in the walls that is just running on its own and how you can begin taking some of your workflows through this progression. Also today, because we pack way too much into every single episode, we are looking at a $700,000 a year bookkeeping firm. Got a really interesting niche, but they are struggling with kind of some operational weight, things getting less efficient as they've grown. And then we're gonna cover a bunch of your questions. Should we be buying like custom tax tools when AI agents can kind of just do this stuff now? You may have heard about the new Claude for small business. They literally released like accounting skills. They're getting normies to run cash flow statements. How should we feel about this? Some advice for accountants trying to partner with other accounting firms to bring in new work. And what's going to happen to the baby boomer firms out there? Folks who aren't learning into AI, let's talk about it because there's a lot of those firms on the market still. Is it worth scooping one of them up? Find the Client List. Come on in, man. Plenty of ground to cover today. How do you do, fellow accountants? My name is Jason Staats. Welcome to the Jason on Firms Podcast, where we talk all things accounting firm stuff. To your partner that has to overhear me talk about this stuff every single day, I apologize. But you know what? Your spouse, they are making the world a better place, okay? We're doing the accounting so you don't have to. I have always thought of us almost as like the financial garbage people of business. Almost like businesses just off-gas all this like finance and compliance stuff and we're like, we'll take care of that for you. You know what? You get back to what lights you up and let us do the rest. You're welcome. Okay. All right. First, what does it look like to take this novel, amazing conversation that you have with Claude doing this really powerful thing? What does it look like to progress that to something that just happens that can run automatically? I'm gonna walk you through 5 steps, sort of 5 levels of autonomy that you can give Claude right now. These are all things that are available today that are going to absolutely torch your cloud usage limits once you have it running in the background 24/7, but it's how you ultimately get AI doing even more of the production for you. I don't want an AI assistant. I want an AI grunt. I want AI to actually take the lead and be the primary staff person driving work forward. No longer am I using humans first and supporting humans with the tech. I'm using tech first and plugging the humans in around that. Like that is the paradigm shift that we are going through. Not a human with an AI assistant, an AI with a human assistant. Is that, is that a bad framing? I don't know. Kind of feels like what's happening. Okay, so first, every AI use case still should be validated in a live conversation back and forth with Claude. This is how you figure out, can it actually do this job or not? And to be really clear, a big blocker for many of us is we try to like think through all the things we do and we're like, yeah, AI could do this, AI can't do that. The reality is you don't know until you come to that AI and say, how can you help me with this job? And this is because AI, it can write code, it can execute code, it has access to all these open source libraries, skills that we don't know that it has until we come to it with a job and say, how can you help me with this? Sometimes when you do that, you'll be amazed. Other times you will be disappointed. But when it crushes, when it does a really, really good job, This is an opportunity to find a more autonomous way for the AI to do that thing for you and your team, kind of behind the scenes automatically going forward. Already stumbled into some really cool use cases from folks I've talked with on my tour. Stuff that's like, oh, that should absolutely be part of everybody's process because AI can handle 100% of that. By the way, in the middle of a 20-city tour all across the US, jasononfirms.com/events to grab tickets. We still go into 13 more cities. Having a ton of fun meeting a bunch of y'all in the wild. And so always start there with the live conversation. But step 2 of this, like the slightly more complex than live conversation version of this, is probably a scheduled task in Cowork. So if you're using the Claude desktop app under Cowork, there will be an option in the top left to create a scheduled task. And this is much harder than a live back and forth conversation because it forces you to figure out how to get this agent to just be able to solve its own problems and do a project from end to end without interaction from you. This is oftentimes harder than you think it will be because having a live conversation, like the agent can come back and ask for clarifying questions for this or that, or have you do one small part of the process, but you have to somehow get that process to the point where it can be done on its own without you having to jump in and get involved. And a scheduled task in CoWork is actually a pretty simple way to do this. Because it can run on its own, but it is still just kind of like one step away from you hopping in and giving it a hand. And when you do have to give it a hand, it ultimately means I've gotta find a better way for it to do this same thing going forward. Maybe that's access to another tool, maybe that's slightly different instructions. But the trickiest part of getting something to run in the background is ensuring that the agent has everything that it needs to do the work on its own and it doesn't end up getting stuck halfway through and need a hand from you. Because if it's constantly getting stuck, it's gonna be a nightmare to try to put this thing in the background and then stuff just like doesn't come through and you don't know why. Downside with scheduled tasks is they can only run chronologically, which is not how most of the workflows in our firm actually work, right? You may have like instances of a thing come in and need to be done sort of randomly, sporadically throughout the day. It's— most things are not like a once a day sort of thing. Some things are, but this is where the third step, the third progression. So first was live conversation. Second one's a scheduled task in CoWork. Third is a routine in Claude Code. And this might sound scary because I just said the word code. It's actually not that hard to set up. Inside the Claude desktop app, simply switch from Cowork to Code, the furthest right option. And in the top left, you will see an option called Routines. And these are very similar to scheduled tasks, but they can be triggered by some other service. And that is ultimately what I need because in most cases, this thing that I'm trying to get to just run automatically like the AI in the walls. It is AI that is triggered when a project reaches a certain status. And this is really important as we are leaning into agents that will better help manage the work that we're doing is this probably looks like a, an evolution in the statuses of our projects and carving out specific parts of the project that an AI can just handle. Really good example of this that we've circled back to a couple of times now that someone shared on the tour. Is if you do tax prep in the US, taking the IRS transcripts and pointing Claude at your tax source docs and making sure that everything that's present in the transcripts is represented in your tax source docs. That is something that an AI agent is 100% capable of doing really, really well right now. And to me rises to the level of this should probably be a status in every tax project, but one that now can be fully automated. So I will show Claude how to do the work. I will give Claude an expectation of how it will document the work. That's probably just a one-page work paper output that I want it to produce to show the work that it's done. But right now, AI can do 100% of that. So what would it look like to sort of put that on rails? The next step in this progression, I would say, would be a routine in code. Now, these can be triggered chronologically, similar to scheduled tasks if you want, but the easiest way to trigger them would probably be a webhook. And many of the other platforms that we use, practice management systems, stuff like that, can trigger webhooks. If you're not familiar with webhooks yet, like go to YouTube and, and find a 5-minute explainer video on webhooks. Not very complex once you've got your head around them, but a very important kind of foundational thing to understand about building your own automations is the power of webhooks. Essentially what we want is a webhook to fire every time a project enters a certain status, and then you want the agent to run off and do that job. And the last step of the agent can be also updating the status depending on whether that task was completed successfully or not. And so that probably looks like making an API call back to the practice management system to update the status. There's other ways that you could achieve that. You could also have the expectation be that, you know, a human has to come back through and verify that before progressing into the next status. That's probably the best place to start. but routines in code can be triggered remotely. Now the downside is that routine, it still runs locally on the machine of, of wherever it was set up. That might be your machine initially. It might be a dedicated machine that you have that is just used for this sort of automated processing. I've got one of those now myself where it's a machine that I only use for Agentic stuff and I have other team members that also access those agents. but it's kind of the next progression from scheduled tasks because it can still be managed on your own machine. So it's easy to see when stuff breaks and sort of monitor it. And then maybe there's a sub-step here for progression number 3 where you actually move that to maybe a different machine where it's running. But then step number 4 in this progression is a hosted routine in Code. And when you create that routine, it will give you the option of creating that as a local routine or creating that as a hosted routine. And the only difference is just what it sounds like. If you select hosted, it will be run on Claude servers rather than on your local machine. And this can come with some other idiosyncrasies because the sandbox in which it's running in a hosted environment is, is slightly different than where it's running on your local machine. So for example, on your local machine, it's got access to certain resources that it may not have the same sort of access to in the hosted environment, like maybe getting it access to the file system is more complicated or just kind of comes with some other connectors that might be required to make that happen. The nice thing here is Anthropic has made it incredibly easy to take a routine that you've already validated on your local setup and then host it. But now you do— you just got to watch out for like what changes when it's in the hosted environment. It doesn't have access to anything on my local machine at that point, but that is kind of the next step, getting it to be something that just kind of runs and at that point isn't even reliant upon your own hardware. Now, to be really clear, we're, we're at step 4 now, but all of the steps up to here each have their own value because even the live conversation version of this is a real time saver, is very valuable. The version of this that runs on your local machine, that's still very valuable. That's more helpful than me having to do that work myself because it can just sort of run in the background. If you take this as far as having a dedicated machine in the office that is just all it does is run this agentic stuff. So it's not clogging up other people's machines. I say that agents, they really don't hold up your machine at all. I would say the, the bigger argument for moving it to a centralized processing location is so that both you and other team members can access that centralized device. So for example, I've got a Mac mini on my desk that I remote into. I mean, I'm, I'm literally in arm's reach from my normal machine and the Mac mini, but I remote into that machine from my laptop, from my normal desktop workstation, wherever I I can remote into that machine and other members of my team, we've got a VPN set up so that they can also remote into that same machine if they need to do something with the agents that are running on it. And so even if you don't get to like the more advanced versions of this, or it takes you, you know, 6 months of evolution on a process before you really totally automate it, that's totally fine. Like this is all progress. Just know there is a destination that is fully autonomous. As I read this ad, I am feasting my eyes upon talented accountants that you could literally, literally book an interview with right now. Like Ruru, a CPA, a tax pro, $3,500 a month. He's got experience with CCH Access, bless your heart, CCH Engagement, QuickBooks Online. Gang, he is one of many accounting pros that you can peruse at the Cloud Accountant Staffing portal. We'll link it down in the show notes. It is a veritable smorgasbord of accounting talent, folks that they have pre-vetted, where you can go in, you can see, see their experience. You can literally see an intro video that they've recorded, and you can book an interview with them without ever even having to hop on a call with Cloud Accountant Staffing. Amazing. We got tax pros, bookkeepers, auditors, executive assistants. Oh my gosh, I got a lot of people in here right now. 45 candidates. Because I know you need that next dream hire. You really do. What do you got to lose? Okay, go check out the portal, book an interview. Feels like a win-win, right? You're only going to move forward if you find somebody you love. Cloud Accountant Staffing, run by folks that used to run an accounting firm. They make it as easy as possible for you. To learn more, check out the link down in the show notes. Hey, if you're done licking your wounds from tax season, if you run more than a few hundred 1040s a year, you got to look at this episode sponsor, Sorban. One of my leaders in tax workflow automation can save you a tremendous amount of time tremendous amount of time right now specifically on the intake part of 1040 workflows, but they're also working on connecting to your tax software automatically, like pushing data in there. They're working on automating the delivery of the tax return. But this is like the fastest growing category in tax tech. If you're a U.S. tax firm, you absolutely need to have a good tax workflow solution nailed down. We did about 1,800 1040s a year in my accounting firm. This would have saved saves so much chasing clients for documents and just tedious organizing of all these PDFs. Oh my gosh. Can't say enough about Sortman. We actually did a full demo of their platform on my YouTube channel not so long ago if you want to see it. But to learn more about Sortman, check out the link down in the show notes. And so step number 5 is Claude's managed agent infrastructure. And this is where you set up an agent that fully runs on Claude servers. It's kind of the next step evolution off of these hosted routines. That comes with even more functionality and managed agents. Like, this is something that is ultimately built for developers, but like all things technical and all things code now with the help of AI, people like you and I without software development skills, we can now do a version of this ourselves with the help of AI. So you are 100% capable of setting up a managed agent through Claude just by doing a back and forth with Claude and getting an understanding of how all this stuff works and having Claude build this for you. And so this is kind of the ceiling. This is like when you have an agent that is running fully in the cloud and is even more capable than these hosted routines. And the big thing that managed agents unlock is the ability to run agent teams, that is multiple agents in parallel, multiple agents in sequence. If you get to where you are running really big, really complex tasks, oftentimes it will make sense to hand off different parts of that task to different agents. If you're doing really big complex processes that are costly and eating up your usage limits, this is an opportunity to actually delegate parts of the process that are lower lift to lower cost models. So you have an agent that uses a model that's a tenth of the cost because it doesn't need all the brains. Now we can carve up different parts of a bigger project between agents. To optimize for cost rather than just throwing humongous amounts of work at the biggest, most expensive models. All of this can be done within Claude's managed agent framework, and this is the most autonomous version of this. It's the sort of thing where it's going to be watching your projects and hoovering something up as soon as it enters a status and hopefully then bumping it to the next status. Maybe hard in the abstract, hard when you haven't done it yet, but that's where all this stuff is going. Right? It is not the AI that we are having this back and forth with. It is the AI that is like the infrastructure of our accounting firm and is just making things happen. I think that's really exciting. As with all things, though, step 2 sounds hard until you're comfortable with step 1. Step 4 sounds hard until you're comfortable with step 3. So just think about this progression and see what you can do to get to the next step. The first step is live conversation. The second step is schedule tasks and co-work. Third step is a local routine and code. Fourth step is a hosted routine and code. Fifth step is a managed agent. And the fastest way to learn this stuff is with other people who are learning it as well, trying to put the same stuff into action in their accounting firm. I mean, find those folks online. I mean, I host events, meet those people in person. Chad Davis runs a great community for this stuff called Automation Town, and I've got a community of my own called Realize with 800 firms in it, and a lot of folks that are going really, really hard on AI. Try to make some of those friends who are trying to do this stuff just like you are, and you're going to learn so, so much faster. What an exciting time to both be in an accounting firm but also kind of like be a bit of a nerd. Like, so much of me wishes I was 16 years old again and I just had more time to like go deep on this stuff and, and not quite so many adult responsibilities. If you've got any experience with this progression and putting stuff on Rails, tell me about it in the comments. Tell me what you did, what worked, what didn't work. Is this the path that you recommend to people. This is the path that I've used in some of my own stuff that I've automated, but I'd love to hear from you if you have done the same. Now, one brave soul submitted to me a whole bunch of intimate details about their accounting firm. We're going to pull back the curtain today. It's your chance to kind of put the coach hat on, give some consulting to this accounting firm to tell them how they can take their $700,000 a year bookkeeping firm to the next level. Let's roast a firm. Roast a firm. Roast a firm! Let's roast a firm on Jason on Firms! If you want the chance to have your firm anonymously roasted like this firm we are talking about today, just complete the form in the show notes. This is a fun thing we've done now with over 100 accounting firms where they give us some details about their firm. We talk about what's going great for them. We talk about what can improve, and it's a chance for you and I to like see somebody else's thing and like give them advice, give them some coaching, because that can be a hard thing when we're looking at our own mess to like really be able to think clearly about. Let me tell you about this firm. This is a 100% bookkeeping and CFO firm. No tax. They say about 10% of their revenue is CFO, so largely bookkeeping. They'll do $700 grand this year in revenue. They are in a rural location. They have a hybrid team. 3 of them work in the office in a small resort town and the other 3 team members are remote. Most of their clients are from this small town, but they do have some clients in a closer big city. Their best client is specialized contractors. So not your typical trades. They say, think GCs, custom cabinetry, electricians, engineers. They work with a lot of contractors, but the ones they like best are the high-level specialized ones. They're high margin, smart and driven owners, and tend to want to go beyond just bookkeeping. Second best type of client, they said they have a lot of restaurant groups here. I like restaurants and the owners that have these multiple businesses because the groups tend to be very successful restaurants, lower margins and lower level bookkeeping services, but they tend to understand the work that goes into it. And my admin team can do a lot of it. So it's profitable but time-consuming. So it is more legwork for the team, but less legwork for the business owner. On the subject of their team, they've got 2 full-time account managers, 2 full-time bookkeepers, a full-time admin assistant, And then the owner works full time as well. Their team went through this big transition recently, elevating two of the bookkeepers to be account managers. And so there are now those, those two AMs, two bookkeepers under them and an admin under them as well, while the business owner kind of hops around and does what's needed. Pricing: $500 a month minimum, $1,500 a month on average. Sales and marketing: they've never had to advertise because they are mainly working with people in this small town. They said they didn't think it would make sense. But they have a massive referral partnership with a local CPA firm that's been game-changing. They're very well known in town, but they also do bookkeeping, so they usually just get the ones they can't take on internally. That is always worth keeping in mind. If you have a referral relationship with somebody who does that work, also know that you are getting the scraps. And there are situations where the scraps are amazing, but know that, like, they're always kind of cherry-picking the best ones. I had a friend that ran a firm who at one point there was a big firm that just drew a line. And said, we will no longer work with clients who don't pay us at least $10,000 a year. And this was a very small firm, like 5-person team, and they are picking up like $6,000, $7,000, $8,000 a year tax clients just like they're nothing. And so for a small firm, like, that's amazing, right? But know that it's still the scraps and there's always something better. Tech stack, they are using QuickBooks Online, Financial Sense, Notion, Front, that's like an email management platform, RoboForm, Loom, and Gusto. Let me just give you their pains, the stuff they're struggling with, and then we can get into the tactical stuff. What they're struggling with right now is operational heaviness. Over the last few years, we've been in the hamster wheel and it feels like it's finally caught up with us. We struggle with updating SOPs, proposals, following through with leads, all the stuff on the back end. This team transition was also a challenge. I shouldn't have done the transition all at once. It's made it difficult to grow and has taken longer than we originally predicted. The team change. They're referencing as moving 2 bookkeepers up to being account managers. Now their backstory, they never thought that they would grow a team like this or absolutely love running this type of business. They tend to work too much because they love what they do. They were working in an accounting department for an engineering and construction company, basically outgrew all the available positions. They had a CFO that they got to the same level as, but they didn't need us both, so they were stuck. They went out on their own in 2019. And the first few years they doubled every year. That's amazing. So they were just in a role at a construction company and one day decided, let me go do this for other people. And here they are now with a team of 5 or 6 doing $700 grand a year. Amazing. Okay, real quick, let's, let's do tech stack first. So this is a bookkeeping firm. They're using QuickBooks Online, Financial Sense, Notion, Front for email, RoboForm, Loom, and Gusto. Notable things missing: WriteTool. If you're a QuickBooks user, you should almost certainly be using WriteTool. It's a Chrome extension that improves upon QuickBooks Online in pretty much every way. Curious what they're using for any sort of document processing. They're working with tradies, and oftentimes you end up with a lot of like handwritten invoices and receipts and stuff like that. Like just a weird collection of like sometimes carbon paper stuff, which like Claude is amazingly good at now. I actually think Claude and ChatGPT are now better at reading human handwriting than you and I are. Like so, so good. And our mind is kind of still stuck in the OCR days where OCR couldn't read handwriting. And so at least if you're like me, you almost assume that the computer can't see these handwritten things. But for folks in the trades where you end up with a bunch of handwritten stuff, AI is actually now very good. So I wonder if they're using something for volume document processing like a Dext or maybe just an AI assistant. Otherwise, tech stack looks great. This is the sort of client profile that is very well suited for Maker's Hub. Maker's Hub is a sponsor of the channel, but I think has the, the highest ceiling right now for AI automation in complex accounts payable work. So project inventory tracking, that sort of thing. You got a contractor that's doing a whole bunch of jobs with very ugly itemized details. Like, that is right in MakersHub's wheelhouse. They're crushing construction stuff like that. They're doing really well there. Okay, let's talk about the pains, the operational heaviness. I can tell you a hard lesson that I learned that I've seen as a recurring theme across accounting firms is frustration with the level of management that is required to keep everybody doing what it feels like is just their jobs. And like you set out and you're like, well, if I just make the systems and I set the expectations clear and all of that, everything will just follow the system and people will show up and their projects will get done on time. I mean that like, I wish that would happen. In my experience, it doesn't. In the experience of a lot of folks who I've heard express the same frustration, it just doesn't. Management is still required. Somebody following up with the team on how are these projects going? What do you need? What are you blocked by? Why did this conversation not happen? You're blocked on X, Y, or Z. And this can feel inefficient, but like to need this management layer, but the price of it not happening, like there's also a price to that as well. And things slipping and people not being checked in with, I guess. That also creates waste of a different type. And then it falls back on you to be the one following up with folks and just doing the check-ins. So that is frustrating, but that is a realistic part of growing a business. There's a day where that wasn't necessary and you were plugged into everything and everybody was so close to you that just kind of by proximity you knew what everybody was doing. But as soon as it grows out beyond that, there's a different level of oversight that's required to keep all the projects on track. Hey, real quick, the most important question accounting firms are asking right now in 2026 is how do I use AI securely in my production workflows. I've said this many times, don't take security advice from a YouTuber. What you need is actual, like, proper IT help, like this episode sponsor, Farodil. If I had to shoot from the hip, uh, with what I've seen from folks, Farodil might be the fastest growing IT solution for small accounting firms. I say small, I mean that like 100 employees and under. You need an IT group that just works with accounting professionals. Otherwise, you are paying this IT group to learn accounting firm things. Like, I had a local IT guy that I once brought over and spent 2 days watching him learn CCH engagement, like how the, like, database worked on my server. Oh my gosh, please don't do that. Work with an IT group that only supports accounting firms, particularly right now as it is so important that we are beginning to implement AI effectively inside of our firms. You want to do that responsibly too, right? You need better help with this stuff? Look no go no further than Verito. They do hosting, they'll be your entire IT group, they'll do your hwisp. Can't recommend them enough. Check out the link down in the show notes to learn more. Hey, I know you love Cloud Cowork as much as I do, but have you heard the news about Canopy Coworker? Subtle. It's a new AI assistant they launched inside of Canopy, one of this episode's sponsors. Canopy, a practice management system that is a feature nerd's dream. They're trying to cover as much of everything you need in an accounting firm in a single place. But as we talk about on this podcast often, it is much easier to roll out an AI agent inside of a closed system than it is to roll out an AI agent that works across like all of my systems, right? That's the trade-off. I want the AI that works across all my systems, but we're struggling now with like, what's the most secure way to do that? The AI agent that works inside my all-in-one system, buddy, that is easy. Flip the Switch, count me in. Strap in, buckle up, I want it. Canopy is so convinced that you're going to love what they got to show you, they're going to cut you a check for 75 big ones. They'll give you a $75 gift card. Gift card to Bed Bath and Beyond. What's that? Okay, I'm not sure where the gift card is for, but if you book a demo with Canopy via getcanopy.com/jason, I'll put that link down in the show notes, they give you a $75 gift card just for coming and demoing. Do it. What do you got to lose? One thing they mentioned they're struggling with is updating SOPs. A really common trap with SOPs, process documentation, is that the documentation is built around your client engagements rather than the firm's processes. Building an SOP around every single client engagement is a monumental undertaking, is an insane amount of work, and we have to like fight back against the natural forces of these engagements, which take them in a different direction over time. Like, absent any sort of system or us trying to like enforce some structure to this, your engagements will get different over time because clients will ask for different things and staff will say yes because it's much easier to say yes to a client than to say no. But when you have a system and you have processes that are your preferred way of getting things done, the opposite starts happening over time. Those engagements can actually start to look more similar as a client raises their hand and they're like, benefits, benefits for my employees. And you're like, oh yeah, no, here's what I recommend. Do X, Y, Z. And now you're taking the lead on that conversation rather than the client just being like, what about this or that? When like they have no idea the best way to do it and you're like, sure, let's do it. Now it's just a one-off thing that you only do for that engagement. As much as possible, we need to get 20, 30, 40, 50, 90% of an engagement to be the standardized approach that we recommend. This is also a benefit of serving a more specific type of client. The more specific a client base, the easier it is to put this stuff on rails. And when you have that language of like, what are the common ways that we do things for people, then you're able to build SOPs for your systems rather than the client. And 100 clients can go through this system that is documented once rather than serving 100 clients and having to then document 100 clients. Now, if 60% of the engagement is standardized and 40% of it is client-specific, great. We can still document the last 40% that is specific to that client and describe What bespoke things do we do for them? But that is a lot better than documenting 100% of the client engagement. In fact, it is 60% better, right? We've only documented the 40% that is specific to that client. Early days in a firm, more often than not, you're taking whatever clients will come in the door and doing whatever they ask you to do. And you can run that business. You need to charge people a lot to be able to run that business, but it gets harder over time. It becomes more and more of a tangled mess to have 100 different projects that are completely unique. It's hard to do profitably as well. I think I've shared this story before on the podcast, but I ended up in this weird niche in my firm where we were like the go-to provider for this chain of retail flooring stores. And these retail flooring stores had this incredibly old, coddly, not only accounting ledger, but like whole operating system that these stores were supposed to run on. And it had like an accounting module. And this thing was so awful. I mean, so, so rudimentary from like the debits and credits don't even have to balance in something that's posted. And so total nightmare legacy system. But all of the franchisees in this large nationally known flooring retail store would use this platform. And so we had one client come in and it was at an early stage in me building this Kass practice, our bookkeeping practice. This client came in and they're like, they're a pretty good-sized client. I'm like, I don't want to say no to this, even though it's not QuickBooks. It's a whole new system. We've got all these things that we got to figure out. It was early enough in that journey where I'm like, this is like a $5,000 or $6,000 a month engagement. Let's just say yes and we'll figure it out. Maybe you've been in this situation before where the number you're saying yes to feels so big that you're like, how could we not make money? And then on month 1 to month 2, you're like, we found a way to not make money. That was a lot of time. That was, that was definitely the experience here. We ate it for like the first 6 months. But on the other side of that, once we had figured this system out, We were so unique, so scarce. Imagine any other franchisee that is using that platform going to an accountant and be like, being like, hey, can you help us with this? And they're like, holy shamoly. Like, the answer is probably no. Like this, to only support one client with this. So we became the experts in this because we knew that software platform as well as anybody else did. And so we kind of just became the default for that type of client. And it's when you get that specific that there becomes a really exciting opportunity to operationalize, to standardize a whole bunch of the work. And so if we're serving 100 of these retail flooring stores all on the same system, I mean, the franchisor wants all these people operating the same. Now, us being kind of the accounting department in the box serving the backend, we can do all this stuff in a super turnkey way, the same way across these different stores. And then it is highly automatable and much, much easier to scale. Think then about our documentation. When we try to document this around the system rather than every single client in its own sort of unique way. This is the power as a firm matures of being able to lean more into specificity. And not everybody has that luxury in the beginning. If you start by taking in anything and everything, it's not to say that you can't start niche day one. You absolutely can. But over time, this is the beauty of getting more specific as it gets much easier to systematize. And that's a, that's usually a more profitable business to run. One other thought on this firm, as they are largely working with folks in this local town, is the thing that got you here won't always be the thing that gets you to the next thing, you know? And I do think just as a framework for life, always think about how you got here as your means to doing what's next. So often we get caught up on what we've built. We fall in love with what we have, and it stands in the way of becoming like the next version of you or building the next I think we've got to just be much more forward-looking about this stuff where what we've done to date, that may have all worked, but that doesn't mean that that's what we're going to keep doing. This firm has built their business around local clients, and a lot of it has come from this big referral partnership from a CPA firm. And that's an amazing thing, and that has gotten them here. But that doesn't necessarily mean that that is what is going to take you forward. Forever. And right now they're struggling with this operational heaviness that I'm guessing comes from doing a lot of different things for a lot of different people. Flipping from local to national, especially if you're a bookkeeping firm, is an opportunity to solve for that problem. Within these specialized contractors, what is the subset even more specific than that that you love working with the most? That is the most profitable that you could see yourself standardizing around. What would it look like to take that small little sub-niche national? This is much harder for tax shops because then you have the added complexity of SALT compliance all around the country. If you just do bookkeeping, there's a really juicy opportunity because every little corner of professional business has now influencers, has committees, has events, has all these places that collect that specific type of person. And if you can show up as the only person that can speak their language, that's doing bookkeeping or doing tax or whatever it is that you do, people will eat that up. They will say, finally, there's a person here that is just for us. So while local has gotten them here, an option here might be national, getting them to what's next and solving for the operational headache along the way of the work just kind of being a lot to manage because it's a little bit all over the the board. To you, I say thank you. We salute you for your courage and sharing some background about your firm. You want to do the same? Check out the link down in the show notes. A chance to get some free coaching, right? What do you got to lose? Hey, the mailbag overfloweth. Let's get into some listener questions. Mailbag, mailbag. The segment is called Mailbag. Questions come from email, from social media DMs, from my cellular telephone. Text me at 1516-980-4968. Leave me a voicemail. Let's get into the first one. It feels like custom workflow tools might be replaced soon by general purpose AI agents just getting better. Does it still make sense to buy custom software or just build on top of some sort of core intelligence layer? Have you built many apps with Claude yet? I can tell you I have. I built like 3 apps last week. Good golly, Miss Molly, is it getting good? So good. Just these little like quick micro apps that takes an hour or 2 to put together to solve a simple problem. Literally on the toolbar of this Mac is an app that I've built that is running locally 24/7 now. I do think for the peripheral tech that that is the stuff that is most at risk from AI right now. The core stuff like your practice management system, where all your data lives, that feels the safest right now. I mean, software people are very scared of AI today, but that sort of core technology feels like the stuff that is the least likely to be replaced. That being said, what is becoming very important, very fast, is the ability of our AI agents to talk with the data inside of those platforms. I know it has never been harder than it is today to make the argument for an all-in-one that won't integrate with anything because we're all using these AI agents that can work across all these tools. They can even log into these apps and use these apps. If an agent is good enough to log into your app and use your app, just give me an MCP. Give me the way easier, way faster way for, and more secure way for the AI agent to talk directly with the information in that app. Don't make it so that we have to like log into it with an AI agent. So yes, I, I do think a lot of the peripheral tools, there's a version of this that we can build ourselves. Now the reality is when you get into it, you find like, oh geez, this is like more work than I thought it would be for certain problems. And then you've got the whole headache of like the overhead of maintaining it over time. That being said, think of the rate of improvement of AI agents the last 12 months and then just consider where we're going to be 12 months from now. My goodness. Right now, AI is incredibly good at writing code. And I mean, the majority of code now globally, I think, is probably written by AI. But the ongoing oversight, the durability of it and all that stuff, I got to think AI is going to just keep getting better and better at making that stuff trivial for us as well. But there is definitely a certain craft that goes into like making an app as simple as it can be, making it work with your other apps. A craft that is not easily replicated by one person banging away on a keyboard in the middle of the night versus, you know, a team of 20 really talented people. But yeah, the core systems, I think they are the most safe. The peripheral stuff, I don't know about you, I'm already building solutions for the peripheral stuff that's really exciting. Next up, what was your tech education background? I've heard you say that you had a background in software. Kinda sorta. So I basically got through 3 years of undergrad doing computer science, and that was always what I had planned to do. I mean, from the from like my middle school years, I was building PCs and that was kind of the first thing I was super, super into was, was anything to do with computers. Junior year of college, I had a couple of bad professors that convinced me that I wasn't any good at computer science. I got like my first D in my life, decided I sucked at this, and accounting was literally the degree that was the fewest number of credits away from what I had up until that point. That is how I ended up in accounting and I just still haven't found a way to leave yet. If only I could talk to my 19, 20-year-old self. Just shows you how big of an impact a great teacher or a bad teacher can have on kind of your whole journey. Since then, I've dabbled in loads of software stuff. I mean, I even after graduating, I wrote all sorts of custom software for our accounting firm. I built basically a practice management system. In fact, I remember when Carbon first came out. Carbon was like kind of the first global practice management system that really picked up traction with accounting firms as like a cloud-based way to manage all your projects and tasks. I remember showing that to one of the cadre de old partners in my firm and saying, look at this, look at this thing called Carbon. And what he said was, that looks like the thing that you've already built for us, but worse. And at the time he might have been right. This was early 2010s when I had so much idle time in the summers, it was a tack firm. I was the most junior person there and I was like, hey, I could probably build some software that would like help us automate more stuff. And they're like, knock yourself out because we got nothing else for you to do. And so I, I do— I don't really don't have a formal software background, just a whole lot of banging around and a close call with a larger software career. Hey, real quick, I want to invite you to my super excellent weekly newsletter. My newsletter was like the start of my thought leadership journey. It's where I try to collect what I think are the most important things for you to see every single week, sent to you every Wednesday morning. Make sure you don't miss any of the good stuff, right? Increasingly, we're also building into that some, like, of the best evergreen thought leadership I've ever done, kind of make that, like, the most action-packed, highest quality email you get to your inbox each week. Because listen, we're all busy, all right? You're not gonna see every single little thing that I put out. Goodness sakes, there's just too too much. If you're not on my newsletter yet, check that out, jasononfirms.com, or I'll put a link to it down in the show notes. Hey, quick word from one of today's sponsors, me. Me, yes. I run a community, Realize, that is for accounting firm owners. And Realize is kind of the secret behind how I've done all of my content and been able to like financially turn thought leadership into a business. About 100 firm owners in there. Realize is a systematic way to build connections with other peer firm owners. It's not a framework for how to run a firm. It's a way to introduce you to other super cool firm owners. Right now we're putting together small groups so you can have a group of like 5, 6, 7 other firm owners just like you that you meet with every single month, run very similar firms. Having this for me totally transformed firm running. My North Star, like my goal for you coming into this group is to make a new best friend, somebody that runs a firm because that was what I needed in my journey. The whole community is built around that. It's called Realize. Link down in the show notes if you want to learn more. Next up, what can we expect from the recently released Cloud for Small Business? Have you heard about this one? So Cloud released, quote, Cloud for Small Business, promising all sorts of things. On that landing page, they've got a video, and the video is a person coming to Cloud saying, do I have enough cash for the next pay run? I don't know yet. And then Claude goes through this kind of step-by-step process. I'm gonna build a week-over-week cash flow. It asks for the pay dates and the pay schedules. It connects to QuickBooks. It must not have been using QuickBooks for payroll, but it connects to QuickBooks and pulls some data from there. It connects to PayPal, which must be how they are receiving payments. Puts together this whole plan and then it lets you export it to a spreadsheet and it says then, so that you can dig into it yourself or send it to your accountant. Nice of them to at least throw us a little more of a bone than Intuit would've, right? Now what Cloud for Business is, is a plugin, a Cloud plugin. And plugins are just like collections of skills. That's kind of it, a bundled up little batch of skills. And you know who can make a plugin? Any of us. You, you my friend can go out and make a plugin. You can make a plugin and give it to your clients so that your clients can use it. Now packaged in this plugin is a bunch of stuff around sort of like financial analysis and just kind of day-to-day small business tasks. And skills inside of Claude kind of are this amazing marketing surface area where you can make a skill for anything and then you can be like, ooh, Claude, you know, really good at this thing because I made this one-page document that's Claude skill. And that's what they're doing here is they've just like they've given it some skills that are relevant to small businesses. In reality, you ask the average client, can I make the next pay run? There are so many different systems that a different business could use and so many sources of information for that. That is it realistic for most of your clients to be able to belly up to Claude and get a complete picture of that? It's not a simple thing. It's kind of a big lift. If you've done cash flow planning before, you know, like just how every client looks different and then every month to month looks different. So many different inputs factoring into this, right? Like what about the hours for the next pay run? Anybody think to ask for that? So Claude, like this might be an upgrade to Claude for the average small business owner. To me, what's more interesting about it is how they are leveraging it with its whole own landing page and promo videos and all that stuff when it's literally just a plugin and there's nothing stopping you from doing the same. Let's go back to this niche firm we were talking about here. Try to double down around a more specific, like, person in the trades. Let's say we're gonna double down on US-based Finnish carpenters. I say US-based 'cause I don't wanna create any confusion. You just work with these Finnish carpenters. Imagine making— a Claude plugin just for Finnish carpenters, giving them all the information that they need. You know what the best tools to use are, common tax questions, how to meet other Finnish carpenters in your area. You can make a plugin for that. You can stand it up in a landing page and be like, Claude for Finnish carpenters in the United States. That's the thing that you could do. So that's, that's the more interesting angle to this for me is how they're using it for marketing rather than the capability. At the end of the day, it's just a skill that anybody can swipe and modify. Next up, "New accounting firm here. Any advice on where to find clients? Is there a website where accountants can refer clients not in their niche?" I wish that was, there was. I mean, that may turn into like a total gross kind of meat market sort of thing if anybody can hop up and like start buying and selling clients. "Is there an app for this?" Yes and no. I mean, you could say the app is called the internet or more specifically Internet Explorer. Like the app is building a network and meeting people and earning their trust. That's the hard thing is we just talked about a $700,000 a year bookkeeping firm that's built most of their business on a partnership with a local CPA firm. How do you get somebody to trust you enough to send you work? How do you stand out from the noise and all of the people soliciting accounting firm owners online these days? Goodness sakes. The app, it is human connection. It is breaking bread. It is not being a weirdo. And it is following through on the opportunities that you're given so that they trust you and they bring more work back to you the next time. Time. It's honestly, it's the sort of thing that I don't think would work all that well on like a web platform, buying and selling clients. Like it just, that just kind of devolves into something kind of messy. This is also yet another argument for not winning business on breadth of services. This is a real trap in an accounting firm to try to win clients on the basis of doing more. Well, we will do your tax and your bookkeeping and your payroll and your CFO and wealth management. We'll figure it out. Oh, need your payroll checks delivered in a hollowed-out wooden mallard? We got it. We bring those ducks to you. The more things your business does, the harder that business is to run, but also the harder it becomes to partner with other people who could send you work. The people that have the most success partnering with other firms, they do one thing really well. They do tax. They only do books, something like that. As soon as you also do the same thing the other firm does, the notion of referring work between each other, that just gets a little more messy, right? So if you want to build your business around partnerships, know that it will be easier to do so if you can say, we are amazing at X. In fact, I saw this in the last few weeks at my live events. A lady is like, all I do is sales tax stuff. And everybody's like, grrr, like craning their heads. They're like, I hate sales tax. Let me get your number. Saw this last year at another event I hosted in Oklahoma City. Lady's like, we just do weed stuff. Everybody's like, ooh, you do cannabis? I gotta get your number. What is the thing that's so specific that when you say, we just do this, people will be like, praise be, praise be, I finally found somebody I can trust to do this thing. And literally in these situations, the bar for trust was they were at one of my events and that like that just immediately earned them evidently enough trust to send work to them. But if you're on the other side of that, is the thing you're doing so specific that other people will rejoice when they hear that that is the thing that you will do. If it's not, then it's going to be a little harder to get those referrals. 3 more to go. How do I decide on what agent to build in the firm and then how to scale it? Here an agent, there an agent, everywhere an agent agent. You can literally plug AI into just about anything in your accounting firm right now. How do you know where to start? A lot of people in the past have said automate the stuff you don't enjoy doing. Automate the stuff that's painful. And we can automate that stuff, but I don't know if that's a logical explanation for where to start. Historically, best practice here in a business would be what is the constraint of the business? Where is the bottleneck? Like if we are manufacturing plastic tubes and we got to hit 20,000 tubes this month, but we're trending towards 16 because this one machine keeps breaking and Johnson didn't order enough tube filament. Those two things, those are the two things that are holding us back. So what is, what is the only thing that matters in the business if we're trying to hit 20K tubes? It's Johnson, his ordering process, and it's this machine that stops breaking. Everybody, everybody in the business, any progress we can make on those two things, that is all that matters. And accounting firms work the same way. Like when we have this huge volatility throughout the year and you have these spikes in your workflow, literally any time you can spend in the non-spiky season to make your spike less spiky is time well spent in the business. And so the place I would probably start is what is the constraint of the business right now? Where are we bottlenecked? This is where I would pull AI in first to see where it can help because solving for the constraint helps the business as a whole. And that won't always map to the thing that you just don't like or the thing that's uncomfortable to do. But truly, it's not hyperbole to say AI can help just about anywhere in the business right now because we're in this really messy transition from humans doing the work and tech supporting to tech doing the work and humans supporting. Really, the whole dystopian framing of like, AI is quote unquote taking over the world and us doing their bidding. Are we not already there? Because we talk about that almost as like a thing that we aspire to do of like, let the tech do the stuff and then you just manage it and you do it when it breaks. Is that not tech ruling the world? Maybe ruling is too strong of a word. Running the world. Whew. And then you see the robotic stuff going on right now. Goodness sakes. It is a new type of nightmare fuel. Every single day. You know, they got robot horses now. What, like four-legged robot horses? Are you kidding me? They're actually pretty darn cool looking. And what, like, what a way to get around, like rolling up to an accounting event on your cyber horse. Okay, next up. Do you have a preferred AI tool/platform that can help manage a high-volume email inbox? There's a growing number of people that are using Claude's built-in connector for this. The problem is the built-in AI connectors don't let you delete or archive email messages. And for me, I'm like, brother, that's the most important thing you can do for me, AI friend, is get all the stuff outta here that I don't want. So little inside baseball here. I can tell you what I've done is I had Claude build a custom Apps Script in, in Google. Isn't that— I don't know anything about Apps Scripts. I had Claude do all this stuff. Basically a totally custom connector that has more capabilities that will allow Claude to delete emails. Or move them to a different folder or something like that. Whatever I'm having it doing, it's not permanent. It's basically like a bot archive, like the bot archived it. And so the built-in Claude connector, it is limited. You can like draft messages, it can read stuff, but it's limited in the triage that it can do. So there are a growing number of people that are using Claude for this, but it doesn't ultimately solve the longer-term problem of email delegation. And email delegation is notoriously hard in Gmail or in Outlook because these are not like team-based email platforms. But that's what most accounting firms need is a team-based way to manage email. I got punched in the face with this one day, had like a come-to-Jesus moment when a member of my team walks into my office and they're like, hey, headed out. Remember, I'm going to be on that trip the next week. But don't worry, I don't have— I don't have internet access there, but we're going to drive into a town once a day. I've got Outlook on my phone and I'll take a look, make sure nothing Jason Staats. And they tell me this and they are like trying to save me any stress about their being gone and all that. And I'm like, this person works for my company and they felt the need to tell me to this, tell this to me when there are 40 other people at this company. And the person that's leaving is like junior. I'm like, you really, like there's really nobody else in this building that can do what you do? I'm pretty sure almost everybody in this building can do what you do. The problem was the expectations, was the systems that we had for helping each other with our email inboxes. And I mean, I tell this person, Summer Child, you don't need to go into town. It's fine. We don't entrust anything to you that is of that importance right now. But it was like a sort of wake-up call of like, no, how can people help each other with their inboxes? So many people are a slave to their own inbox because it is quote unquote, "their." inbox. Let's be really clear. Nobody in your firm should own their inbox. If somebody leaves the company, are you going to be like, I'm not going to look in there. That's, that's their inbox. Hell no, man. That like that. This is a company inbox. We treat email as if almost as if your accounting firm, if you had like physical offices, like everyone there had their own street address and clients would email things into your own little mailboxes. But only you had permission to go downstairs and open your own mailbox. Nobody else look in my mailbox. And this usually stems from like 1 or 2 HR emails a month that go out via email. Just find a new home for the HR stuff because in an accounting firm, your email inbox, it is the receiving bay of the accounting firm. Like it is how information comes in. It is how new work comes in. That is not a hill that anybody should be dying on alone. We need to share how we approach that work and that you do that with shared inbox tools. The leaders on that right now in accounting firms are probably Missive and Front. And both of those just sit on top of your Outlook. They just sit on top of your, your Gmail. And so Claude can help with the triaging of email, but the long-term strategy is an actual email system where stuff is systematically moved between teams and managed by more than just one person and by the team instead. Last up, what is going to happen to the baby boomer generation of firms. They're gonna go away one way or another. And we shouldn't generalize too much because I know a load of baby boomers that are like way ahead of the curve on AI stuff. So let's not generalize. But what they're saying is like the old school firms right now. I'm honestly amazed at how much value they're still holding. Like walking into a firm that is a paper firm, I am shocked that people are still buying paper firms. And I'm convinced that the people who are don't know just how hard it is to change a firm from being paperful to being paperless. Especially in a world where there are paperless firms for sale. Why you buying the paperful firm? I probably don't have as much compassion here as I should because I started my career at an awful baby boomer firm. And I suffered, but I was like 21 years old. I had nothing better to do, nowhere else to be. The people I worked with though had families and had kids and we were working 60, 80, 100 hour weeks. And what I saw there was awful. And like, no white-collar worker should ever have to endure that. So when I think the old-school firm, that is what I think of. And they're not all that bad, but that is why I probably don't have a ton of compassion here. For me, I would never buy a firm like that. I would go and find another firm very similar to my firm and buy a firm like that because that is— even if I got to pay 50% more for that, that is a vastly better acquisition to make. The staff are going to be aligned. With the way I work. That is the riskiest part of buying a legacy firm is now you have to teach the entire staff a completely different way of working. And these are accountants in that sort of firm. They've probably been doing it that way for decades. Good freaking luck, brother. But if they don't sell, all those clients end up back in the wild. So I don't know, circle of life. Some of those firms will get bought. Many of them won't. I think over time most of them won't. But all those clients will go back out and have to be served by somebody. There's for sure like a thing that you could do. You could do a roll-up of like old codgerly firms and you get them at a big discount and find your way of serving them. The problem is like most of these clients, they can't really be served in scalable ways. Like they're used to walking into an office and just showing up and doing this or that. And it's inherently a non-scalable model. But I don't know, maybe someone will figure that out. I mean, I know accounting firms where they like go into retirement home centers and do that stuff stuff in person with those people. That sort of firm, I freaking love that, man. Talk about meeting an unbelievable need, like to meet those people where they are and sit down and do that stuff with them. That's freaking amazing. Like, bless those people. But what I don't need is to go by some old fart's firm to be able to do that. So I don't— yeah, I don't totally get the value of like the aging out firms. Some of those firms are really cool. And again, I don't want to generalize. There are firms that are doing cool stuff that have valuable clients and valuable relationships. But I'm going to assume from the question being baby boomer generation of firms that these are just kind of like the aging 1040 shops. That stuff, not all that interesting to me. I got better things to focus on, especially if you are making friends with other accounting firms, man. Like, as the world gets smaller, I think it opens up so many cool M&A opportunities. A number of deals now, so many deals now that I've seen go through where like I found out, find out a friend is doing something and I'm like, And now I just ask, who'd you do the deal with? And a big percentage of the time I know the person on the other side of it because we kind of travel in the same circles. Man, those are the best type of acquisitions to make because you have so much common ground and it's not this total revolution of like having to flip the whole machine on its head and get staff to work a totally different way. As the world gets smaller, M&A gets so much simpler. I always push people, build your network, make more friends doing this stuff. I appreciate that you're on a recumbent bicycle right now and it's just me and you, but let's get a third person in on this. Not like in a weird way, like how do you make your next friend in running an accounting firm? All right. That was— I'm sorry, that was an unfortunate way to end this episode, but I appreciate you making it all the way to that moment. Until the next one, I'll see you there. Take care. Happy accounting.