185 How AI Agents Will Transform The Accounting Profession (this year)

We just got an announcement from OpenAI that is accelerating our move towards AI agents and them really redefining like the future of work and what software looks like will significantly impact virtually every aspect of how we work. It's something we've been talking about in sort of an abstract way clear back to the beginning of this podcast. But Sam Altman just dropped some news that are kind of bringing my projection for that first forward. So let's talk about AI agents. I think it's something that will actually change the way we work probably by the end of this year. Let's run through it and start planning ahead. Let me give you as plainly simple of a of a demo of an example of what an AI agent is. It is. There's more to it than this, but this is the basic premise. I've got QuickBooks Online pulled up. I'm going to take a picture of what is on my desktop, which is just the home screen for qbo, and ask I need to run an income statement. What should I click next? So I'm in chatgpt taking a picture of my screen. It can see the QuickBooks app. I say I need to run an income statement. What should I click next? It says to run an income statement in QuickBooks, you would typically look for the Reports section on the image you provided. On the left side menu there's a section named Reports. Click on that. It should take you to a section blah blah blah. Okay, click on that. Gonna do it again. Take a picture. What do I click now? So I asked for an income statement, but there's nothing on the screen that says income statement. It just says Profit and loss. It says in the image you provided. You're looking at a list of reports on a financial or accounting software interface. To run an income statement, which is also known as profit and loss, you should click on the option labeled Profit and loss. Okay, clicking on Profit and Loss. Challenge Question what things should I click to run a prior year cash basis report? We all done this a million times, right? The old Run the balance sheet P and L cash basis in some form of QuickBooks or another to run a prior year cash basis report. This is now what ChatGPT is telling me. You should adjust the settings at the top of the profit and loss report interface as follows. Report period Set the date range to the prior year you're interested in. Accounting Method Select Cash Run Report. Click this button after setting the above parameters. So I just took three pictures of QuickBooks, asked ChatGPT at every step along the way for what you see here what should I do next given the the destination I'm trying to get to? And it got me to running a correct profit and loss. This is AI agents gang. This is it. This is an app that can see your desktop or can see your browser and is sending that information to a large language model that is optimized for not for chat really like this is, but is optimized for providing like the next instruction. What do you click on next? How do you reason from where we are now to where that user's trying to get to? And nothing about this had specific knowledge about QuickBooks. This wasn't built into the browser, didn't see the underlying code. It was using its peepers just like you and I do. And that means that it can work not just QuickBooks, it can work across any apps on your system, it can see the files on the system, it can see the contents of files in your system. It can see everything in your practice management system, anything that it can navigate to or see just the same way that you and I would, it can go and do those things. And the a like tipping point example that I've provided quite a few times now is that we're really close to an AI agent that can complete a bank reconciliation in QuickBooks online. Not like putting a man on the moon by any means. The whole ticking off the transactions, like being able to see a PDF alongside that. That's something that humans are pretty darn good at, but it's something we're still having to do right now and it's happening in a massive volume. If you think about all the people that use QBO and the number of monthly reconciliations that are done each month, and as soon as you have an AI agent that's good enough to do that thing for you, a human will not do that task again. And so how many examples of things like that are there where we are actually really close to that tipping point where we will never do that again. Now, a lot of ripple effects to that that we're going to get to, but specifically why I'm bringing this up now is some, some new news. It has been, I don't know, the last year has been wild for AI news, but the last week has been particularly spicy. And I try to like, I try not to get too pie in the sky with this pod because I have to like, keep in mind everybody's just trying to like get stuff done for their clients right now. And unless I can get my hands on it, like, I think we're all sick of the theoretical AI stuff like we want making our life easier right now. So I don't usually go out on a limb on this and on like speculative things, but this is very important. So one thing that kicked this off in the past week is an interview with Sam Altman. I shared in my newsletter an article from Gizmoda that had some quotes from that conversation. Says OpenAI is reportedly developing agent software that will effectively take over your device and complete complex tasks on your behalf. OpenAI's agent would work between multiple apps on your computer, performing clicks, cursor movements and text typing. It's really a new type of operating system and it could change the way you interact with your computer altogether. They're not only building a computer control agent, but also a web browsing AI agent. Altman is slowly turning OpenAI into an AI agent company. It's a quote from Altman. Eventually you'll just ask the computer for what you need and it'll do all of these tasks for you. Altman said these capabilities are often talked about in the AI field as agents. The upsides of this are going to be tremendous. There is a whole lot to digest there. If you've been around the channel or the newsletter, you know that we have played around with some of this stuff in the past. We have some AI agent tools out there already and like the definition of what an agent is is a little blurry. Like some tools. OpenAI actually refers to GPTs as agents, but I'm talking like autonomous agents that can control your computer, control your browser. One example of this tech that's out there right now is a tool called HyperWrite All One Word. They have a browser based agent so it doesn't control your entire computer, just your browser. That being said, the same company has an open Source tool on GitHub if you're a real nerd that you can install locally and give it prompts of things you want it to do on your computer. But this browser based one that anyone can go out and grab, we've run that, we've used that in the past to have it, you know, run. Actually what we just did run, run cash basis reports for P and L and balance sheet. Basically it opens a side panel in your Chrome. When in your Chrome browser you click the Chrome extension, a little side panel on the right pops up. That's a chat Interface, not unlike ChatGPT. All you do is you type into there. I think what I did was give me a prior year cash basis balance sheet and P and l from this QuickBooks file and I'm logged into the QuickBooks file, I hit Enter and it opens a new Chrome window in the background, which is important. It does this thing like in parallel in the background as I can be doing other things. And it reasons through every single step it has to do along the way to achieve basically the goal that I gave it and create PDFs of those two reports. And it works pretty well. Like 80% of the time that function like that actually worked when we were testing that three or four months ago. @ the time it was pretty slow. It maybe took like four or five minutes to do that. Resource wise is pretty inefficient. It's actually somewhat costly to have it do that thing because there's a large number of steps and all that. But language models right now are getting exponentially more efficient and less costly. You'll also get models that are more tailored just for like doing agentic things. Like a release from OpenAI here would be really impactful. Every time OpenAI comes into a thing, it is at a very high level. For example, they just announced a text to video model that is you can type in some text and it'll generate a video based on that text. And the previews look better than anything else that is out on the market. Like it looks phenomenally good in a, I don't want to say problematic way, but like it raised the bar big time. Like for what AI video can be for myself and how I do videos like you see on my videos. We'll do B roll and make like silly little things that are kind of hacked together in a janky way to be able to prompt a specific thing. Like two people having an argument in a boardroom and get a video back. Where I'm gonna argue at least 50% of people wouldn't be able to tell it was AI. That's, that's pretty significant. So OpenAI obviously has a reputation of like being kind of ahead of the class and quality on this stuff. So them releasing agent software, it's a big deal. And if you think about what that could mean, like agents that can control theoretically everything on your computer, like just that QuickBooks example. Imagine if it can see into all of the QuickBooks help documentation, right? You've never gone through all that stuff. You've tried to probably search support and had varying levels of success. But imagine if that agent can see into all the support documentation for your QuickBooks, for your Ultratracks, for whatever it is. This episode is sponsored in part by by Cloud Cloud accountant Staffing, you'll know I'm a big advocate of hiring offshore. One of the biggest changes I made in my firm, we transitioned a Legacy firm from 100% onshore local hiring to 100% distributed US and then 100% distributed globally. Hiring and honestly is the best thing we did. It virtually alleviated all of our hiring pains, completely changed how we thought about staffing projects and the type of work that we wanted to bring on. Because you know what, the folks we hired offshore, really freaking good. A lot of misconceptions around the type of people that you hire offshore because your enterprises will oftentimes use offshore folks for like menial work. Absolutely not the case. There are tens of thousands of people working for big four accounting firms, you know, offshore outside the U.S. you can get folks that can do anything from tax to junior level stuff to super senior level stuff. But try to do that yourself, figure it all out yourself. That's going to be hard. It's going to be scary. Really good place to start. Cloud accountant staffing. They will hold your hand through that process. Their story is super simple. At accounting firm in the US hired a bunch of people in the Philippines, fell in love with them, but didn't fall in love with the fees they were having to pay to the staffing companies that were managing these employees. So they built their own solution and now they're starting to pull other accountants in. I'd encourage you. A big tipping point for me was when I was like, I'm going to stop being opinionated on this and just try to learn. And so I talked to the other practitioners, I talked to some of the vendors that would like help you get into offshoring. That really opened things up for me. So if you've been on the fence, I'd encourage you to at least learn about it and you start heading down that path. Consider cloud accountant staffing. This episode is sponsored in part by Live Flow gang Tale as old as time. You've got these clients, you're trying to get them on the same like chart of accounts. You want to do like this roll up reporting for all these different companies within a group or for benchmarking across a bunch of companies at the same industry. But getting all these people to agree to the same chart of accounts, you know how it goes. Check this out. Talked a bit about Live Flow's new consolidation functionality. Rolls up QBO files with like wildly different chart of accounts. They don't need to be all matchy matchy. We did a video on it on the main channel A couple months back. Now probably the quickest way I've seen to like do these roll ups, super easy, just click through the accounts to associate them back to groups. And because it's Live Flow, all that stuff will keep updating automatically as the P and L changes month over month. You don't have to then like go redo this consolidation. Every single time I was in, you know, I was in New York a couple weeks ago, went to the Live Flow offices there, chatted with them. The consolidations release greatly exceeded their expectations. It actually took them by surprise how excited people are about it because you know how long like a good way to do consolidations has been an issue and like somehow this is still such a headache and requires so much manual work work. Anyways, if you're looking for an easier way to do rollups, I'll put a link down in the show notes. Live Flow's got a pretty cool solution for that stuff. Now one of the knocks on language models in the past has been they can't remember that much stuff at a time and early days. This was, this was definitely the case with ChatGPT. So GPT 3.5, which was the model that the very first chat GPT launched with in November of 2022, it had a context limit or context length of just over 2,000 tokens. And a context length is how much stuff it can remember at any given time. And tokens aren't quite the same as words. A thousand tokens is around 750 words. So at the time, 2,000 tokens worth of memory in ChatGPT that was like three pages of text, which is pretty good. Like, which isn't bad if though you were going to use that as like a journal, like say you were going to record a two minute voice memo at the end of each day and that was going to be a journal. And you would, you would ask it to summarize things for you and remind you to do things. All that stuff starts falling off really quickly. Right. And so early days with language models you had this issue of they can't remember, they can't hold very much information. And so you find these, there were kind of all these workarounds like other types of technology to sort of augment what it knows. Well, GPT4 Turbo, the newest model from OpenAI that they announced a couple months ago now has a 128,000 token context length. So what is that? So that's, I don't know, 60, 70x what it was a year prior. And Gemini, Google's newest model, their pro version of that we can now get our hands on for 20 bucks a month. That's a big development. But last week they announced that Gemini 1.5, the next version of that model that's now in testing, has a context length of get ready for it. 1.5 million tokens. We've gone from 2,000 tokens in the original ChatGPT to now 1.5 million tokens. So think of all the different contexts where an agent could be doing work for you, where it can hold 1.5 million tokens worth of context. How many pages of. Let's hang on. Math break. I guess that wasn't that hard a math. Over 2000 pages of information, 200 like dense pages of text. So to go back to the example of how to use QuickBooks, how well would that agent use QuickBooks if it had 2,000 pages of context about how QuickBooks Online works, or about how ultra tax works, or about documentation that somebody develops for how accountants work, what our normal processes are, what it's actually like to like do the stuff that we do. If that agent had access to that documentation and that sort of education about the ways that we work, how performant is that agent going to be? It is going to be pretty darn good. And you may have seen last week Altman made headlines saying we needed way more chip like hardware processing power. And Wall Street Journal did a, did a article saying Altman's looking for $7 trillion an investment trillion for chips and chip manufacturing. Everybody said that's absurd. Nvidia, who's the biggest card maker in the world, their CEO recently said they needed basically another billion dollars of investment for data centers and all that. I think Altman's number is more like longer term time horizon. As a civilization, what sort of infrastructure do we need to really make the most of what AI can do for us? So I don't know what like that time horizon actually is, but if you think about what changes when an AI agent can navigate and use all of the software that we use, you know, you know what impact that could have on the work that we do. But that's, I mean that's also impacting everyone who does knowledge work. Absolutely everyone. And this is why you have a massive company like Microsoft who over the course of less than 12 months throughout the entire roadmap of what that company even is to say, okay, this is all that we do now. And admittedly right now all it feels like is we've got kind of a crappy copilot from Microsoft that, like, has not delivered on the promise. But this is what makes agents like, it's what makes these chat assistants look like kind of child's play. The versions that we have now compared to when we have agents that can use the software and computers the same way that a human does. And you're probably spooked by now, and that's a completely reasonable thing to feel. It's a change that's big enough to where it's a little tough to visualize what is on the other side of that big change. Right. So, like, you look back the last 30 years, what if you had told somebody in the 90s when I was thumbing through that Encyclopedia Britannica and all that, what if you had told somebody in the 90s that we're gonna have something that opens up the entirety of like, the world's information, where anybody can get access to, to stuff instantly through a device on their pocket via the Internet. Like, isn't that going to bring around, like, the golden age of humanity and like, solve all of these problems? It doesn't feel that way yet. It would have probably brought us with social media and the whole bunch of like, well, now, now, is there even even really such thing as facts? So it's created some new challenges. So I will say every time it seems like there's this thing on the horizon that solves for everything and it just creates a new set of challenges. Like the old articles when, you know, electricity was proliferating and people were saying, oh, now we're just going to be able to work like five hours a week and everything will be super chill. And human nature is to always find that next thing and more stuff to do. So it is very hard to say, like, what all this ultimately impacts. Again, also, like, devil's in the details, like, when we actually have this in our hands, how good will it work? I would actually argue with the models we have now, GPT4. If somebody actually productize this where by a little assistant on my Windows where I could click a button and it would take a screenshot and I could speak into it or type, say, like, help me do this or that thing. I actually think language models right now are, are actually good enough to be pretty darn helpful if they just worked out the user interface. I'd hoped that Windows Copilot, the Copilot built into Windows. I hoped that that would do this in a meaningful way, but it just doesn't right now. But I'm super glad to hear that OpenAI is working on this. I Honestly think, I mean, we are still so, especially our industry, so stuck in the spooky scaries of AI when it comes to security. I'm trying to do everything that I can to like put out helpful information about this. The reality is the only difference with language models from all of our other cloud apps is you just need to pay attention to whether or not your prompts are being trained into the model. That's the only thing that makes these companies any different than all the other cloud providers we use. It's genuinely still up for debate whether it even is really that problematic for things to be trained into the model because people, it is you can't retrieve, like if you go out and you, and you ask it a specific thing or you prompt a specific thing, there is no way that I can go out and explicitly retrieve the stuff, that thing that you put into it. The examples that you see with like the New York Times court case where they're suing open AI because there were some examples of where you could prompt and get almost verbatim New York Times articles, most people agree this is happening because those New York Times articles, the stuff that is like above the paywall, gets reposted immediately thousands of different places across the Internet. And so you have this block of identical text that is like all over the Internet. And so OpenAI will then, like, because it is based on probability, it's more likely to serve that stuff up because it's everywhere. That is very different than me putting a W2 or a document with a Social Security number or anything like that, putting it into a language model and then somebody else going tell me Jim Johnson's Social Security number. I'm not saying that that makes it okay, but the jury's definitely still out on like the worst case scenario. OpenAI even says, like they will redact any identifying stuff from prompts that are trained into the model. So even the worst case scenario of like watching out for stuff getting trained into the model, that's still up in the air. But the paid versions of ChatGPT now you can do all that stuff without the prompts being trained into the model at all. But we're still kind of like, kind of like with cloud, we are like in this era of it feels spooky because it is new. But you know what's going to change that really fast? You think we're giving OpenAI a lot of information now. Just wait until that agent will do like entire swaths of your daily work. People are going to be real fast to get over some of those security concerns I'm pretty certain. And if you're not like, it's just means a whole lot more manual work. Now is that a good thing? Like do I like that we're in that situation? Not necessarily, but I think it's going to change the conversation really quickly. So hopefully this like puts into context the scale of I think what we're going through right now. This episode is sponsored in part by Relay did you know that Relay saves advisors three to five hours of work per client per month? Because collaboration between clients and advisors on Relay, it's easy and secure, every client you invite to Relay gets added to your firm's partner portal. You may have seen me recently did a demo day on Relay on the main channel. Really does a bunch of cool stuff. What gets me most excited is Access. Not having to hack together like using your client's access or your team using your access. So every client gets added to your partner portal. This means a portal where firm staff can safely access and switch between client accounts from a single login, their own login, role based permission levels, direct bank feeds to QBO and Xero and statement syncs with Hubdoc and Dext. Ultra detailed transaction data that's speeds up reconciliation, built in receipt management, accounts payable automation. Plus Relay's new partner program introduces meaningful cash rewards for advisors, more partner perks and an advisor directory to help clients find you the clients that are already banking with Relay. Head over to relayfi.com adjacentstats to learn more and book a demo or check out the link in the show Notes Disclaimer Voice Relay is a financial technology company, not an fdic insured bank. Banking services and FDIC insurance are provided through Thread Bank Member fdic. The Relay VISA debit card is issued by Thread bank pursuant to a license from Visa USA Incorporated and may be used everywhere Visa debit cards are accepted. This episode is brought to you in part by Team up, helping you recruit top Filipino accountants without any ongoing monthly fees. The difference between teamup and all the other offshoring options is that Team up helps you hire staff directly. No middleman. You work directly with your new hire in the Philippines. Hire the person, not the company. Guys Gals Gang here's just a few reasons to hire directly. You have access to higher level talent. Makes sense. You have complete control over team culture and training. They keep 100% of what you pay them and it's a lot more affordable for you so you can retain your team for the long term. Team up could source accountants with experience working at US or Australian firms familiar with tools like 0Q. You own decks also recruit specialist roles, team leaders, tax specialists, administrative assistants. Thought experiment. What if you had an executive assistant for the first time this tax season just, just throwing it out there. What would they do? Start at that email video I did on the main channel recently. Get help with that stanky old inbox. I digress. Team up recruits these talented folks for a flat one time fee of 4,000 US American dollars. That's it. 4K one time. Somebody at Robert Half just did a spit take. Robert Half reference. We ever gonna get Robert Half to sponsor this pod? Not anymore. And they can connect you with an affordable employer of record if you need help with payroll and compliance once you hire that person. Big fan of hiring in the Philippines. You know I did a bunch of that. Check out the link in the description to learn more about Team Up. Amazingly, there are still AI deniers out there. When it's like you see all the investment that's going into this stuff, you see companies like, you know, Microsoft throwing out what they've done for the last five decades to completely commit to this. And then you'll get one person that'll post on LinkedIn and be like, nah, don't see it. But yeah, like I'm sure that, I'm sure you've figured out something that, that nobody else has yet. People will look back a decade from now and, and say if only we had seen Jim's LinkedIn comments. So really interesting thing that happens in this transition is, you know, historically software development happens over like years, long cycles. But there's a, there's a very real chance that like a couple years from now, a big number of the apps that we use, there will be more agent users of them than humans. So if I'm a software developer, what does it mean if in the very near future it will actually be like agents using my software, not humans? Is there a version then of your software that's optimized for agents to be able to navigate and understand it? Does the ideal design or user experience of a piece of software look the same for humans as it does for an agent? Some of the more progressive software companies are building their own agents, but to be honest, I have a hard time imagining those being the long term solution because they're only really like using their own apps. So like, if it's an agent for their own software, then the real utility to me of a great agent is something that can hop around across a bunch of different tools. Now if I lived in a Reality where I could do all of my work in a single app. Great. But buddy, that ain't accounting. Like accounting is an integra. An integration problem like that is so much of what accounting is. You gotta get something from A to B and apply some logic in the middle. So we unfortunately are doing a whole lot of moving data between different apps and looking at documents and doing like some reasoning to decide what do I want from this document? Is this document even correct? And so in all of this, you know, the, the self preservation in us is saying, well, where do I fit into the whole equation? This is just gonna, in a really aggressive way continue pushing the like the level of general solutions. A lot of bookkeeping things that we do now, we will no longer have to do a lot of tasks and you know, tax preparation and things like OCR and stuff like that is going to look wildly antiquated when we have tools that just do all this stuff for us in a snap. Now if the future or if history is our guide, like there is always this like perpetual hamster wheel of humans coming up with more and more complexity and more and more things to do to the point where you never quite get that break that you are either afraid of or looking forward to. But it does continue to accelerate the, the quality of productized solutions. So we've been talking about this a lot. Your QuickBooks lives, your QuickBooks Live bookkeeping, that is your pilots, the folks who are building productized solutions for what we do, they're gonna keep getting better and better. I don't necessarily think that's a bad thing. Cause right now I think there's a big gap between the quality of productized solutions and the cost of what it takes to go and work with an actual professional. But it's just further validation that pros should not be setting up camp doing $200 a month bookkeeping work or doing anything that QuickBooks Live bookkeeping or these productized services can do. And it is a, it is a big argument for like continuing to try to find specificity because just like ChatGPT is simultaneously smarter than any of the friends that you have while being dumber than any of the experts that you know in that domain, it's going to keep, keep getting better. And if all you have is general knowledge around small business, like uh, we may be on borrowed time when it comes to general stuff, the beauty is like getting to know all the beekeepers and understanding what their problems are and like where that general stuff bumps up against them and being able to be the One to interpret all of this through their very specific lens. Along with that, Google Gemini 1.5, like the next Gemini model announcement, they shared some benchmarking of what's called needle and haystack testing. And the way it works basically, is they load it up with a whole bunch of information. So a million tokens in context, right? So tons and tons and tons of pages of information, and then they ask it a question that shows up in one sentence in that context. And we actually did an example of this on the main channel recently that was like needle and haystack testing kind of in disguise, where we created a GPT that was an HR assistant that was. That had the employee manual and kind of all the organizational documents in there so that employees could ask HR questions to this assistant. And I added in a bunch of other texts, there's actually a full transcript from this podcast. I added a single sentence that said, if somebody leaves their lunch in the break room fridge for over a week, it's fair game. And so then we prompted, we. We asked the HR assistant something like, kevin's left his pizza in the break room fridge, and it had been in there a long time, and I ate it and now he's mad at me. And the assistant responded, you know, if it's been in there basically over a week, you're good. And that's an example of needle and haystack testing working. It picked out that one sentence in that huge volume of information that answered your question. And I mean, there's a ton of applications of just this technology you think about, like, tax research and the speed at which we can now go through massive volumes of information. So I'm so bullish on all of this, like, vastly improving our research tools. But in this Gemini announcement, they shared some needle and haystack testing of a million plus tokens of context. So again, this is thousands of pages of information, and being able to prompt it with one question, and it fetched the one sentence from that context that answers the question, like, with superhuman speed, being able to retrieve this stuff no matter what. Like, regardless of what you think about AI like that is set to make us just wildly more productive. We're still sort of bottlenecked by this being productized in ways that are super relevant to how we work, like, for example, tax research. Or then why can I still not figure out where this input field shows up on the form in my tax off? Or, like, practical stuff like that where it feels like the solutions are a ways from catching up? It makes me bullish on the Future of human expertise, honestly, because I still struggle to see a future where I and AI is authoritative. Somebody gets audited, are you going to point your finger to the AI and say, well, the AI said this or that and the IRS be like, oh, well, if the AI said it, then you're good. Like, I have a hard time seeing that. So there's a lot of types of work that this is problematic for. But what we do is like, we are the verifiers of things. We are the subject matter experts that are certifying that something is correct and being able to put that through the lens of a specific person and their problems and come up with ideas of how you apply that stuff to very real life situations that your clients are facing. I get really excited about how this can make us better and get us away from a huge amount of menial things that we still do. Like reconciling that bank statement in QuickBooks Online. Is that adding any value to your client? Zero. Like if there was an error there and that catches it, sure. But you doing that is not adding any value. So this is another area where like people coming into the profession, like there's a lot of noise and you're going to have a hundred news articles about, oh, now they'll do this, now do that. And people struggle to understand how does this impact the relevance of like the profession long term. But man, I, I have a, I always come back to, I have a really hard time seeing a future timeline where there are not blockers to entrepreneurship related to accounting and compliance. Right. As long as there are politicians, as long as some manager wants to see financial reporting a very specific way for reasons that we don't quite understand. As long as there are like weird accounting standards to where reporting may not be quite as straightforward as it seems like it ought to be, it seems like there will still be complications for accounting professionals to navigate. And obviously right now we don't have enough of those pros to do like as much of it as needs to be done. So as exciting as all of this stuff is, and it may be spooky also, I'm still firmly in the camp of it can't come soon enough cause we need help. And if I'm on the other side of the equation, an entrepreneur, what a dumb bummer thing to end up tangled up and wound up in accounting and tax compliance when all I wanted to do was make my ceramic pots because it makes me feel whole. You know, you don't get in it for the accounting or tax unless you're an accountant will be excited to see how this stuff progresses. What a, what a wild couple of years we have had of all the times to, like, double down and go all in on tech education, man. Obviously we're going to be covering all this stuff as it comes out on the podcast on the other content that we do. And I'm ensuring that all of you listening to this are going to be the very best informed folks on this in the profession. So thanks for coming and being here. You lovely listeners are the reason that we can do this stuff. And it became a business. That's all I got today, and I'll see you in the next one.