442 How To Know When Work is “Good Enough”

How do you know when, when the work product is good enough, when you're doing too much, something that isn't worthwhile, the juice isn't worth the squeeze, the client's not going to pay any more for it, versus the other end of the spectrum. That's problematic. Where it's not enough, right? Hot take, maybe. Hot take. I think most accounting firms, I think because we're doing things to our own standard, we oftentimes take the technical too far. But the client experience, not far enough. That's what we're going to talk about today, drawing on what was a actually fun conversation on social media recently. But because it's a Q and A Friday, we're also talking about, hey, isn't Intuit just going to like or pick a company, big company, just going to build like their own version of ChatGPT or some AI tool that ultimately just kind of replaces what I do also, is local AI, local large language models? Is that the future, the secure path for accounts to work with AI? We're going to, we're going to dig all into it. Come on in, let's do it. So how to know when work is good enough? Megan Blair kicked off an interesting conversation on LinkedIn. I'll give you the bullet points here. She says quality and accuracy matters in accounting and bookkeeping. But what happens when everyone's standard for quality and what is accurate enough is different. Some industries have seals or ratings. We grade lumber, we give health departments ratings to restaurants. The tuna industry gets certified as Dolphin Safe. Schools get annual reviews and scores. So what is the industry standard for bookkeeping, accuracy and quality? As AI accounting tools and firms continue to pop up, what will be the seal? Like, what will that quality seal be and what rubric will they be scored on? My immediate reaction to that, I don't know if you've, you've thought this like, there's a wave of and not that great sort of productized solutions and this isn't a new thing, but it's becoming more and more of a thing with AI, where there's the level of how good it can be with just like a purely automated productized solution. You give it a pile of stuff and here's the level that it can do that work to. And what we do as professionals who run firms, and we are literally the tippy top, the alphas, the ones who do it to the highest technical level, what we do is we look at those solutions and we go, that thing sucks. Sucks because it's not as good as me or not as Good as what a collection of really talented people in my firm would put out, but it's still light years ahead of maybe what my client was capable of doing themselves. I think we see this with AI almost everywhere right now, because just rocking up to Chat GPT, it is out of the box, better than you at everything except for what you do. And that's, that's a great way to think about how to use Chat GPT is. It's still not as good as me at my domain of expertise, but at everything else it's probably better. And it's, it's not as good as hiring an expert and pulling them in to help you with it, but it's a whole lot better than if you were just left to your own devices to do, do something that you're not at all an expert in. So I, I actually think in the coming years this is going to become more and more of a conversation of what is good enough. Because even a version of say, bookkeeping that could be fully automated, let's say we force the banks to get onto like a standard reporting sort of thing where we have good API connections for transaction feeds and maybe we can get a little bit more metadata into what those transactions are so they're easier to classify. They have this in the uk, there's a standard that all banks have to adhere to. Let's say we did something like this in the US and we can get all that banking data reliably. And then regulators say, okay, we can see that there's a fully automated version of bookkeeping that we could potentially do here, but it's going to be missing a lot of the more detailed information we have with how current regulations are set up. Is there a middle, like a middle ground there where we can maybe in a more aggregated way, like without the same granular level of detail. Is there a version of, of this that's fully automated, that is good enough? Where we'll say, okay, this is the new standard for what is good enough and it's a lower standard than what professionals have done in the past, but it unlocks X. Maybe that's it being fully automated or it unlocks greater compliance because the problematic small businesses out there today, they're not the ones that are working with firms by and large. There are like crappy and fraudulent firms out there for sure. But, but, but by and large, non compliance is like the folks who are not working with firms at all and who are actively trying to, you know, avoid paying tax and stuff like that. So there's with us being like the top of the heap and the, the most accurate, there's almost certainly a trade off there that would make sense that unlocks simplification or greater compliance or something like that. That is a lower standard than what you and I are used to today. The big blocker is just, it requires everybody kind of getting on the same page and saying, here's how we're going to do this. Right? And so elsewhere in the UK for example, they've had making tax digital, which is like a simplification and in some ways like the E filing mandate that we had in the states. But it's like a much more streamlined approach to business tax reporting. In Australia they do a version of this that's mandatory quarterly. So that totally changes how an Australian firm works with their clients. They have to be doing that touchpoint with the client every single quarter to ensure that the books are correct. Be a little more hands on for that quarterly tax filing. Right now in the UK that's annual, but they're talking about going to quarterly and that, that just totally changes your year as an accounting firm and how you serve your clients. In the U.S. it's, it's an annual thing. Maybe someday they get their act together and do a more frequent reporting. But, but you can imagine if that was quarterly, you're just a much more embedded part of that, that small business than you would be otherwise if you just did work annually. And, and that's why we have in Australia like a fundamentally different relationship with the accounting professional because they are just by the product of compliance, they are much more integrated into the business because you have much more frequent touch points. Now, couple other thoughts on this notion of like getting to a, a standard is I would argue that, that right now if, if there were a standard and if there even was an absolute way to do this, of, of what's correct and what's good enough. And right now there isn't. Because the, the tax rules and the perspective of whoever's consuming the financial statements, all those things come through different lenses. So you can have eight different classifications for the same transaction. The marketing team wants to see it this way, the manager wants to see it this way, the business owner, the board, whoever that is, the, the federal government, the, you know, state and local compliance issues there, maybe they need a different version of that classification. Right now there is no absolutely correct answer because of all of those different perspectives and contexts that could impact the classification. But that's also why to date it hasn't been able to be fully automated so like many things in accounting, it is still subjective. While externally non accountants, they look at what we do and they think it's all deterministic. And that's why accounting, accountants and the profession at large has always been the target of automation is the perception is that it is deterministic when the reality is when you're doing a month end close, you're not spending 80% of your time classifying transactions, you're spending 80% of your time tracking down people for information you need to answer questions, integrating different systems that don't talk to each other. We're already doing all the leftover stuff now like, like not every firm is, is as, as efficient as leading firms. For sure, there's ways to automate stuff. Some firms have not leaned into that. But the perception that accountants are just like going down the list and classifying things, that, that's just not really what accountants do. Because if that was all that we do, then that would have been automated by now. Now kind of the, the secret sauce and the, the exciting thing about AI is that it is kind of like a version of intelligence in a box. So it is much better at subjective things than any other tech that we've had before. And that's the exciting thing is AI can actually help us automate some of that subjective stuff. Given your wisdom and knowledge and perspective, the same way that you would instruct another human to make subjective decisions. So if it depends on xyz, you define to the large language model. Here's what XYZ are, I'll give you the information and then you decide and that's, that's something. Honestly that AI right now already does pretty good. And you can kind of test a version of this and say chat, GPT and the, the, the simple kind of universal example I always come back to is creating an automated AI workflow to handle your emails. So you get emails of all sorts of different types. Right. And let's say we wanted to create, use and create an AI agent that would identify any emails that were sales outreach from a, from a software company or something else and put it in a specific folder and you'll revisit it later to do that with say an inbox rule that's very, very challenging. Right. Because it's going to be coming in from new vendors. The verbiage that they use looks a little bit different. There isn't like a hard coded way for you to indicate here is what sales outreach looks like every single time. Throw it into this bucket. So there's no built in way to do this in say your Outlooker or a Gmail filter. But if you took every single email that you ever got the body of it, you take one email as an example and you go to Chat GPT and you say, hey, tell me if this is a, like a sales outreach email. Just respond yes or no and then you give it the body of that email. You could give it even more instructions if you wanted to say like, you know, common ones here are like software companies, staffing agencies, like it's very common that, that folks will get in touch, act like they know me already, but are ultimately just kind of trying to sell my firm something now only respond, is this a sales outreach email? Yes or no? And it responds yes or no. AI is excellent at that right now. In fact, I would argue that it is at human levels of accuracy on that. Because when you think about things like phishing and how people can easily get tricked, if you run a team, you have seen your team get tricked by things that are like, yikes, really you, you went for that, you thought that you should go out and get a, a Walmart gift card. You thought that would be a normal request, like whatever that is. We've all seen humans get duped by this. And I would argue that AI right now is much better than many humans and so at the very least as good at a human at identifying these different types of things. But you, like, in practice, you're not going to go to Chat GPT and copy and paste every single email that you ever get in into Chat GPT and say, is it this or that? What you want is that to run automatically. There's ways to do this with tools like, you know, Zapier is probably the most obvious one. Basically a zapier if you haven't played with it. Z A P I E R We can put a link to that down in the show notes. These are just like custom workflow builders that will run automatically for you and they can ping the OpenAI servers to have AI like handle a request the exact same way that you would chuck something into Chat GPT. So for example, you create a zap that was like on every single email. Every time you receive an email. If all we're going to do is check to see if it's a sales email, then it, it triggers on a new email. The next step in the workflow is going out to OpenAI and giving it the body of the email with that prompt that we just tested in Chat GPT saying, is this a sales email? Just respond yes or no. Then in Zapier, after that you have branching logic that says, do this if it's yes, do this if it's no. Or in this case, maybe just a filter that's like, only proceed if it's a yes, and if it is a yes, move the email to this folder. Boom. You now have AI classifying every single email that comes into your inbox without you having to lift a finger. It's intelligence in a box. Now that's a very, very basic example. What you would probably do is say, like, classify this email into one of these eight buckets and then those each have their own paths. Like, this is a whole rabbit hole of something you could build. It would be very, very complex. You could try to do it yourself. I'd probably recommend pulling in like an automation agency to help bring to life that idea that you have. They'll probably do it in a more robust way than you will. And if you're like me, you listen to a podcast and you're like, well, this is a very good idea, I'm going to go do this. And that's just kind of where it goes to die, right? This is what AI is unlocking. That is the next step after ChatGPT. And we could do this today. Like, this is all tech that we have that can be deployed. You can literally build an AI agent to manage your email inbox today. Today's episode is sponsored in part by the folks at Cloud Accountant Staffing who help you find talented accountants in the Philippines. We get a tremendous amount of inbound from anyone and everyone that's got like offshore staff and wants to come and work for us. Right. So how do you cut through the noise and know who's legit? Well, for one, I'm going to let two offshoring groups sponsor my content. And I now know tens tons of firms that have used each of them to great success. Anytime I hear about them using them, I always ask, how did it go? How was the experience? Have heard nothing but positives as why we've. We've stuck with them. So, so how's that for a filter? Because I have talked with a lot of them, that ought to put them ahead of most. But the other thing to look out for in any of your vendors is are these folks turning up at accounting conferences and having real conversations with real accountants to understand what we need? I can tell you Cloud Accountant Staffing like they are making the rounds this year. If you are out on the circuit at all like I am, you're going to See them, and that is a great signal that they care that they're not some like random offshoring group that are just throwing work at people with all sorts of various skill sets. Like, they are very focused on the accounting firm and exactly what your needs are. They're actually run by folks that used to run an accounting firm. Fancy that. That's cloud accountant staffing. To learn more about them, check out the link down in the show notes. Hey, real quick. This episode is sponsored in part by clienthub. Who's ready to turn your busy season into you ready for it? Into your wow. Wow season. Actually your whole year into your wow season. How are they going to do that? Jason, I'm glad you asked. Three bullet points. First, it's all about the client experience. Make it easy for your clients to communicate with you. And half the battle's already won. 2. Simplicity. ClientHub. It is not any harder than it needs to be in a lot of practice management systems these days. Let's just say it's getting a little sweaty. You step into Client Hub, you try it and you're like, oh, this, this actually feels like a nice place to be. And number three, AI stands for artificial. What was it stands for artificial intelligence. You know, I like that. Client up. It'll help you draft emails with AI. They got a new like automated book review thing that pulls in some AI. Say goodbye to busy season, man. Make the commitment. We're not doing it again. They've got a limited time promo right now. 50 bucks off your first three months. They give you a personalized demo because you want it to go better next year. Right? I want that for you. Make this the last stressful busy season you'll ever have with Client Hub. Link down in the show notes. And so that's email, the example of email. But we could do the same thing with AI reviewing a set of financial statements. So Megan here is talking about like, what is the standard? What is the bar? Right now we could totally build a step by step workflow that's like, look into this, look into that, look into that. And it's separate prompts basically going to OpenAI saying look into all these different things individually. Which it works much better than going to ChatGPT and asking it to do 20 things at once. It can't really do that. It can only do so much in a single request. So if you are building a custom workflow, say you've got a 20 item checklist, then each request to OpenAI kind of each step in the process is a Single item in that checklist. It's not all 20 items. It does the first one, then it does the second one, then it does the third one. And there's a version of this that we can do with AI right now. Like, by the way, like AI quality assurance tools, like this is how they work in the back end is just a whole bunch of prompting and logic to like, sequence through what's the right way to do that. We can do a fully AI version of this right now. That honestly would probably be pretty good. I don't think it would be as good as, say, the seniors in your organization, but I've seen enough to tell you that it's better than the junior bookkeeper who sends opening balance equity and all sorts of crazy stuff across your desk. And so that's, to me, an example of, of where we already have a fully automated version that could be done that's not as good as the biggest brains in your firm, but is light years ahead of your clients and even better than some of the people on your team. Which just brings you back to the question again of what is good enough for most of us. I think what that looks like is it's as simple as this is how good is it done by the best person on your team? Like, like, that's it. And that's, that's not an absolute answer in any way. But if you look into any firm, the answer to what is good enough is what is the. The best, like, most senior technical person in my firm capable of? And. And we kind of just go about our day and say, like, yep, that's the absolute answer. When it's like, no, that. That really isn't. That's just a product of what you have in your team today. And sometimes you, like, you discover things where you're like, oh, shoot, like, we could have been doing that a whole lot better than we've been doing it. R and D tax credits. We had this in my tax firm. We didn't have our antennas up for R and D really in the way that we should have. As that developed like more during the Obama administration when it, when it kind of was opened up for more folks and there was all sorts of applications around, like agriculture and. And we, we had this perception that it was just for software companies. We didn't work with software companies. And we ended up going back and amending a whole bunch of old stuff. And we were totally blind to that going back even further. Domestic production activities deduction. Yeah, I'm an OG man. That was something that Actually way more people were eligible for than we originally realized. Went back and captured a whole bunch of that. That was early in my career. And so like whatever the highest technical level our firm is capable of, it's not the right answer, it's just the answer that we have today. And I will say as like more of these product solutions come on, come online and we build more of our, our own sort of AI driven solutions. In fact, I was on a call yesterday in my accountant community, I do basically weekly coaching calls in my private community. Realize that's always linked down in the show notes on a call yesterday, there's a guy that's basically built a whole sequence where they, they gave it just what I described their checklist to review their financial statements. And they run this thing through. It's not only reviewing it, it's doing like some of the actual work of putting them together. It's a process that took their team about four hours per month, then close to due and it's like shrunk that down to about a quarter of that size. So people are like actively doing this right now. And if you're worried about the security stuff like for sure, think through that. We talked about Zapier. Zapier for example, comes with multi factor authentication, SOC2 type 2 security, even on their free plan. So like that is, that is the bar for me generally that's good enough. It's a lot of very commonly used tools in our space that do not come with SOC 2 type 2 certifications. But for sure, like do your due diligence on each step in that process. If you're working with OpenAI via API connection, which is what Zapier is doing, none of that stuff is ever trained into the model, so we don't have that issue. But this notion of getting to like an absolute correct answer, it's never been a thing. But if anything like AI models may actually get us closer to a standard. Because when everybody's using the same tools and the same AI models to get to get a month end closed up to a certain level of accuracy, like to me that's actually better than, than no standard, especially right now as I actually believe that what it will give you and what it can identify is above the level of the lowest level people in your firm. And it kind of brings me around to the notion that I don't know that humans are capable of creating that standard. But the closest thing to a standard that we may have anytime soon is whatever AI is capable of right now there's going to Be a hundred different ways of, like, how you use AI to do this, and the good ways and the bad ways. And they'll get productized. And increasingly we'll be relying on our software to do this for us. The folks more at the bleeding edge that want to take it further, like, they will for sure build their own stuff, just like what's happening right now. But because I struggle to see humans being able to create that standard, because it is admittedly a hard thing to sort of put in a box, the standard almost becomes whatever AI can do. And we're kind of living in an age where I would argue the biggest existential crisis to the accounting profession is accountants not being able to turn out quarterly and annual reports for publicly traded companies, which we're increasingly seeing they just can't find good help anymore. They can't get that stuff out. And that like, becomes a risk to entire financial markets. And if the accounting profession, like can't put enough butts in seats to get that stuff done, financial markets will find a way around the accounting profession. I think to me, that's a bigger existential risk to what we do, or at least what, what some of us do, the folks serving publicly traded companies, right? And in situations like that, good enough, good enough changes. It really does. And it starts looking more and more like, how far can I get us right now? It's weird, weird time, honestly. And this isn't just an accounting thing. Like there's a version of this in every single industry where there's this like, rising tide of how far can a pure product, can pure AI get us without human intervention? And that, like, that line is coming up every single month now. We've talked about this on the podcast a lot. In fact, this leads right into the next question that I got. Because where does that, where does that leave us, right? Like when, when it keeps getting better all the time, where does that leave us? Here's the question. I got it over dm, so I won't say who it is. Something occurred to me today. Chat GPT passed the CPA exam. Intuit is scooping up clients like a vacuum and selling their, their live full service option to have the return filed by a real person. With AI, it's possible to create avatars that can have a conversation with you. And when you put all that together, it seems highly plausible that Intuit and ChatGPT will eventually sync up to create AI tax pros with CPA level competency to complete, to completely eliminate 1040 shops. From there, even progressive firms like Mine are under threat of becoming obsolete. What do you think? So ChatGPT can pass the CPA examination. That's, that's actually been the case for a long time. And this kind of comes back to the question of is what we do deterministic or not? Because the deterministic stuff that has a concrete right answer, like math and science. This is why all the benchmarks for AI models are around like, oh, how smart is, is it like at the level of a PhD scientist? Or this math benchmark. How, how, how many questions can it successfully answer on like the Math Olympiad? Those are all things that have concrete right answers. I would say that, that while this is an inevitability for sure, like Intuit will continue building AI into their offerings and it'll get better and better. But what's left over one is the subjective side of this. When it's not 100 deterministic. A lot of very simple, you know, tax returns, for example, they are deterministic. It's like it's a W2, it's a 1099 interest. There's not a lot of wiggle room to interpret that any different way. And so that's an example of a deterministic return. And we already refer those out. We already tell those people, you don't need me to do this. Like go plug it into this software tool because the answer will be the same and you'll save a bunch of money on professional fees. So to me, where, where the, the space remains for us is on the subjective stuff. The stuff that's not totally black and white. And, and as we talked about AI, it is getting better at handling subjective stuff. But we're, we're, we're still quite a ways from it, like weaving all the things together that can potentially happen within a tax return. Especially when you consider like the wildly various formats that clients give you, the information that you're then supposed to interpret, figure out the follow up questions to ask, and somehow pull that together into a structured format. Hey, real quick. This episode is sponsored in part by the folks at Team Up. Team Up Up. They help you find talented accountants in the Philippines. Team Up's got a killer PDF guide to how to do this stuff. They call it the 2025 salary guide, but it covers way more stuff than just salaries. And one of the things that it digs into is I think the biggest misconception about hiring folks offshore is the difference between outsourcing versus hiring direct. A lot of folks that have had negative offshore hiring experiences, what they're actually doing is working with an outsourcing group where they're buying month end closes or buying tax returns. And it's up to that outsourcing group to manage how it gets done and the training of the team members and all that. That was not the model of offshoring that I did. I did what Team up does. Direct hiring. All it is, you're working with human beings in another country and they're capable of all of the things that my onshore people are. At the end of the day, it's up to me to train them and kind of develop them from A to Z. But there's no like middle person. Oftentimes in these outsourcing groups you're paying. And this was the case for me because I'd done both. You're paying that group pretty much the same amount of that person's salary so like doubles the cost. I just want to work with the person directly. Team up will help you do that. Check out their guide covers a lot of just like kind of the questions you're kind of too embarrassed to ask. You can download that one down in the show notes. Hey, real quick. I write a weekly newsletter, goes out every single Thursday. And let me tell you, it is a banger if you ever struggle to keep up with like a high number news updates. Things are happening so fast, Jason, where do I learn more? What is the most up to date tippy top hot goss in the industry and tech advice and all that? It's the newsletter, buddy. Straight to your inbox Every single Thursday, 2pm Eastern. Thousands of accountants read it. I'll put a link to it down in the show notes. If you aren't on that list yet, you better get on it now. We talk about taste a lot here. We've talked about this on the podcast quite a bit the last few few weeks. The reality is it will continue to get better and be more and more capable of doing all of these things that we are talking about. And so it becomes more important than ever for you to be able to apply your lens like your worldview, your thoughts on the right way to run a specific type of business through this, through accounting and through compliance. And the analogy I always come back to is if I tell chat GPT to go buy me a living room set right now, by definition, without any other prompting, it will be the most basic, the most vanilla, like the Stanley mug of living room sets. Because large language models, they're, they're probabilistic, they're literally the average of everything. And so that will be the default. Now I can say give me like a cool like 60s kind of chic vibe or something more Scandinavian and modern and it'll go out and it'll do that. But there will still be a day when you walk into somebody's living room and you're like, oh dang, this is. Nope. I didn't even realize that this was a thing. Because the person who put that together has a taste that, that you weren't aware of. And on the day that AI can do all of the things that we're talking about here and it can take that mess of information and file a tax return and get the books done at a good enough level because that is probably an inevitability. And I think you're probably just fighting over is that three years from now or is that 50 years from now? When that day comes, it's, it's worth thinking through that timeline because if you're, if you're my age, it's going to happen, it's going to happen while we're still doing this stuff. If you think through that timeline taste and like what your, your living room furniture set looks like, your, your opinion on the best way to run that type of business, it still matters. In fact, there might even be a greater appetite for it because at that point more of the things that we work with and we do day to day are totally automated. They're machines. And there's like a craving then for the human perspective. I mean, remember coming out of COVID how much we all craved these in person experiences. And I don't know about you, I, I feel like that hasn't really gone away. Like, I feel like there's a greater hunger for meaningful experiential stuff, at least in a way that I wasn't aware of pre Covid. And I think the, the more things are automated, the more people are working from home, that machines are solving things. There's like this itch inside us all to make connection with other humans, to learn and to connect around very specific things. Even when it comes to work. Like, you know what it feels like to not have any friends that run an accounting firm and then walk into a room with 10 people that run accounting firms and you're like, that was the best 90 minutes of my life. And if you haven't experienced that yet, like man, man, do I want that for you. Because it's transformative, that's still going to exist. I don't care how good AI is, like people will still hunger for that. And so when it comes to running a firm that helps entrepreneurs navigate all things accounting and compliance, I think we'll, we will still be doing accounting and compliance things that would probably just look different than what we're doing today. But people will still want to know your take. They will still say like, with all the tools that we have today, like, what do you add to that? Like, what's your spin and your take on the things that I need to be aware of? While the bar for all the software that we're using continues to rise and gets more and more powerful, the differentiator. On top of that, when we're all using the same tools, it will still be you in the same way that it is today. Right? We're all rolling QuickBooks zero whatever your tax platform is. We're all using largely the same tool sets. And what differentiates my firm from all the others is me is, is the humans that I bring into the process is on the opinions we have on what the right and wrong way to do this stuff is. And that will remain regardless of, of how powerful those tools get. And so, so, so what do you do about it today, man? You don't be afraid to like have a, have an opinion and put that out there. You have conversations with other entrepreneurs. If I look at this through the lens of what I do today, I have conversations every day with a whole bunch of other firm owners. And that is what informs my worldview of what's the right way to run a firm. And as a result, I'm very opinionated about it. I'm a super irritating know it all. But that's because I've had all those conversations and seen what people have said anecdotally worked for them and didn't work for them. And so even in, in a, in an era when AI is running way more of accounting firms than it does today, you better believe I'm still going to have thoughts on like how you navigate everything that falls in between the cracks. Right? And I, I see the same for accounting firms. Last question. Is local AI the secure AI solution for accountants? There's a growing number of companies that are, that are pushing this narrative. The, the notion that the only way that you can work with AI securely is with a local solution. For me, this is just like patently false. It's, it's the, the cloud debate over again. The notion that things are more secure when they run on your local machine, which I, I think is totally false, honestly. The notion that something on your own office server is more secure than something running on, say Microsoft servers, that's probably just wrong. Like you're not unplugging the ethernet cord every single day. That thing is connected to the exact same Internet that everything else is. And you got a whole lot less in the way of resources than the other groups do to protect the goods. And you know, phishing is probably the biggest risk to firms as a member of your team, getting tricked into putting their credentials into a fake login screen or, or something like that. And the reality is like, at this day and age, I, I just fundamentally don't think it's any more secure to run your own server than to use Microsoft's servers. And so there's, there's a growing number of solutions providers now that will give you like a local language model that you can run. The main issue, and this isn't a hard thing to do, any of you can go out and run a, an LLM. The main issue is the models you talk with and say a chat GPT are running on unbelievably, unbelievably powerful servers. So the models that you can actually run locally are very, very small, very underperforming. And there's, it's, it's a big old headache to get it to run on Windows hardware versus Apple hardware, which is, can actually handle a lot more. And so on the one hand, it's, it's like not worth your time. Like, you can, you can run a really crappy AI model locally right now or literally for free. You can go to ChatGPT and use unbelievably powerful models. Just know that on the free model, on the free Chat GPT, they're, they're training it on your prompts. So I send people to the team plan $30 a month, totally fine to use client information with that. And so like, the only people that I think are attracted to the local alternative are maybe the ones that haven't gotten their heads around the fact that there are completely secure ways to use AI in the cloud. The only difference with AI apps is to look out for whether or not your prompts are trained into the model. On the ChatGPT team plan, that's not the case. So it's literally the same as every other cloud app that you use. If I'm using Practice Management System in the cloud, I've done my due diligence. If I'm going to put sensitive information in there, I feel good about the security setup. ChatGPT team, it is as secure as every practice management system on the market, if not more secure. So the notion of going offline to do AI stuff like that's like going back into the past. It's true that models you can run locally, they'll get better, they'll get more efficient. It'll be amazing the things that local AI models will be capable of doing in two years time. But you know what will be even more amazing is what models can do in the cloud on really, really powerful hardware in two years time. So you'll perpetually be living in the past to solve a problem that probably isn't a problem at all. Because I don't actually think that it is really any more secure to run that stuff on your local server versus OpenAI servers. Right? This was a nerdy one today. Sorry about that. We went deep on some some pretty geeky AI stuff, but if you still listen to this podcast, you're probably more than okay with that. Otherwise you would have left a long time ago, right gang? Thanks for being here. You got any questions, drop it down in the comments. We shared some nerdy stuff here from like building custom workflows. If you want to explore say building an AI agent for email, actually just pop that into YouTube, like do a YouTube search on it. You'll find some pretty interesting videos that will get your wheels turning on how to do that. Not from me, but just more from general like automation experts. If you want to connect with a bunch of other automation folks, consider joining my my community. Realize Chad Davis, big time nerd. He runs a monthly meetup where folks are kind of sharing the workflows they're building like this. The whole idea of that community is to like make some other friends, get connected in small groups, have kind of an abundance of connections with with folks who run other firms. And if you're just not sure where to start with chat GPT for your team, a couple months ago we did kind of a beginner soup Soup to nuts. How to use chat GPT for accountants. We'll put a link to that down in the show notes. I think it is. It is the most watched training chat GPT training for accountants, like worldwide. It's a great place to start with the team gang. Thanks for being here. We'll see you next week. It.