182 Delivering Monthly Financials with an AI Assistant

Okay gang, hear me out. What if the next time you delivered a set of monthly financial statements to a client, you included not only the PDF of the balance sheet and the profit and loss that they won't look at, but a link to a chat assistant that has seen all, like the month, all the financials, the general ledger detail. They can chitty chat with it, ask questions, a lot of the same questions they may ask you. What, what if we did that? Because you can. And it's actually really easy to build. So I'm going to talk you through how it works. Really like cool example of actually using speech like, like verbally chatting with it and you'll see how good and how compelling it is, but then also zooming out like, is this a good thing? Is this a bad thing? Is this something I should actually consider doing? Come on in, let's do. So. You may have seen this one on my newsletter last week. In fact, in the show notes, I'll put a link to the specific newsletter where I showed how to build this in greater detail. You don't have to subscribe to the newsletter. This is just a public link where you can see all the goodies. Swipe the prompts that I developed to make this work and like perform well and you'll be surprised like just how easy this is to build. So before I get into like a demo of how it works, uh, it is worth just acknowledging the why of this. Why would I send a chat assistant to my client? Is this like actively pushing them away from asking me questions or wanting to chat with me? It's possible some clients like may think that. And so people are gonna be across the board on how this is received. But my take on this and kind of where things are trending is people appreciate self service if you give them a really easy, low friction way to get the answer themselves. In general, most folks are going to appreciate this and that. It's the same discussion that comes up with the portal conversation. I don't want one more login to have to manage. Gang. It's a website. We have websites for every single service that we use where I can log in and download my utility bill. Imagine having to email somebody or having to call somebody to get a copy of utility bill. It is not like anti client service to stand up a portal where a client can log in and download a copy of their tax journey, something like that. And I see an AI assistant as almost like a self service path. And we've gone deeper on this in the past on this podcast talking about AI avatars and how those are about to get way, way better. And the reality is our clients are going to have access to free or nearly free advisor avatars that you can upload a tax return to or connect your accounting file to. And like, these are going to be fairly high level conversations. And so it is kind of an example of one of those things where people are building this stuff. It's going to exist. I mean, Intuit has been kind of teasing their chat assistant that I don't think is live yet and will not in the demos I've seen, will not do some of the things we're going to see today that this does. But if it's me, I would rather like be out there and transparent with my clients about like, hey, here's some early stuff that if you want we can kick the tires on. Because what I don't want is for the whole world around the client to be developing and they're seeing all these new assistants and all this cool new functionality, but then my advisor is just like blind to it or pretending that it doesn't exist, right? Like when Intuit launches Assist the Chat Assistant and QuickBooks Online, I need to show my clients how to use it. Otherwise I am the old timey advisor saying, oh, that thing doesn't know anything. You need to use me for everything still. And I don't want to be that guy. Like, I don't want to complicate things where it doesn't need to be complicated. And I think that's why, like, this is exciting to me is it's a version of expertise that isn't going to be as good as me and isn't going to be as specific as me. But between the client's expertise and my expertise is a big old void. And where Google and ChatGPT just solve general problems, this is actually answering very contextual problems down to the exact stuff that's in their financials. And that's pretty darn cool. And the fact that this is as bad as it will ever be, that's really exciting. That's cool to me now, I think more so than some. I definitely have an abundance mindset of. I'm really not afraid of this displacing work that we do because there's still so many things around it and so many more people and more problems and more advising that we would all love to be doing if we had the hours in the day. The reality is the stuff that we're stuck, like the problems we're stuck solving for clients right now are not fun problems are not big, big brain problems. They are, gosh, this regulation is way harder than it needs to be. Like they're not. There's better things we can be doing. So I, I definitely do not get spooked by this stuff because I think there's just so much more that we're capable of doing for clients that I'm happy to see this menial stuff going away. Okay, I'm going to give you a actually a voice demo. So this is built in ChatGPT so you can chat with it. But what we're going to do here is a speech to speech demo on ChatGPT. You can actually talk with ChatGPT via the mobile app and it's pretty darn good. Sometimes it's a little slow, but. But couldn't you say that about most of us, you know? So I'm in my ChatGPT mobile app. I pulled up the custom GPT that I made here and I'll get more into how we set that up after we do this little demo. And it just looks like any other ChatGPT conversation in the mobile app. There's like a little headphones button that will let you have a speech conversation with it. So let's ask it a few questions about the January 2024 financials. Did we make any money last month?

Yes. Last month the business had a net income of negative $1,677.19. This means the company spent more than it earned, resulting in a loss for the month of January 2024.

Why'd I lose money?

The loss occurred because the total expenses and other expenses exceeded the income for January 2024. Here's a total 3,081 38 which includes design income, landscaping services, pest control services, sales of product income and services. Total expenses $2,663.82 covering advertising automobile including fuel, equipment rental Job expenses, legal and professional fees maintenance and repair meals and entertainment Office expenses rent or lease utilities, gas, electric and telephone Other expenses 2,666 listed as miscellaneous expenses after subtracting the total expenses and other expenses from the total income, the net income came to a loss of 1,600 70,000 dol7.19 this.

Means the 2,600 of which was miscellaneous.

And other expenses than it earned in income.

Tell me what was in the miscellaneous expense account.

The Miscellaneous expenses totaled $2,666 and comprised two items. A bill from Brosnahan Insurance Agency for $2,000. A bill from Tim Philip Masonry for $666. These expenses contributed to the overall loss for the month.

What were the largest largest expenses out of the checking account last month? Trying to use like Normie language, like not all the perfect vernacular, the largest.

Expenses out of the checking account last month were a payment to Brosnahan Insurance Agency for 2000 a dollar a bill payment to haul properties for 900 Al Qaeda a payment to Tim Philip Masonry for 66 DARA a bill payment to PG&E for 114, 09 dollars. These expenses significantly contributed to the overall expenditures for the month.

What was that? 110. Okay, so that you got an idea of how the voice chat works there. This episode is sponsored in part by Cloud Cloud accountant staffing. Y' all 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 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 US 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. An 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 gonna stop being opinionated on this and just try to learn. And so I talked with other practitioners, I talked with 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. Been talking a lot lately about the new consolidation stuff that they rolled out. Honestly a killer solution for rolling up a large number of companies. But a couple good use cases for liveflow may not be thinking about Liffel Syncs accounting data from your QBO file out to Google Sheets, which can be super helpful in situations where you don't want to give somebody access to the QuickBooks file itself, but they need access to specific data from the accounting file. That could be your client who you don't want in there messing stuff up. Could be some other stakeholder who may need access to metrics, but you don't want them in the file. Like actually in the file. Could be you're up like bumping up against user limits within QuickBooks where maybe you have to upgrade to another level to let more users into the file. Some interesting use cases there for lifelow around just giving granular access to folks who can't be into the file themselves. Kind of wild if you think about the whole ecosystem of all the things that can be automated around spreadsheets as well. Stuff you can build with, you know, zapier and make and the scripting built into Google Sheets and all that. Use Live Flow to get the accounting data there and you kind of do whatever you want with it. So if you're looking for a way to liberate that accounting data from the accounting file, check out the link to LifeFlow in the show Notes. Really cool to me about what this does. And whenever we talk about chat assistance for accounting things, people always say we had these five years ago with like chat AI and intuit, like had like a very basic assistant. I don't know that they ever actually released. But the beauty here is it is your information but also all of chatgpt. So like I can tell it Give me a table of all the stuff in miscellaneous expense or a table of all the stuff in this account, this account and that account and then give me an Excel download for it or give me a CSV download for it. So you can either via chat or over voice, like semantically work through, you know, show me all the ledger detail for this or that. Here we're just working with a single month, but behind the scenes here it can actually handle a pretty large volume of information because what it's doing is it's taking all the documents that you give it, in this case a balance sheet, a profit and loss and a general ledger, and anytime you prompt it, it uses a different process called, you know, embeddings and vector search that actually pulls just the most relevant bits from those documents. So you can have massive documents in here and not be up against. You know, you've probably heard about context limits with chat assistants or how much a chat assist remember at any given time, it used to be that was very little information. Now it's actually hundreds of pages of information. But even if you got a massive ledger report, you can chuck that in there and it'll just pull out the relevant stuff for the prompts that you're giving it. And the beauty here to me is just how easy this is to build. So I'm in chat GPT. We've talked a bit about custom GPTs. These GPTs are kind of like special purpose assistance where you can give them instructions to only do certain things and kind of put them on rails. But then most notably they give you the ability to upload knowledge, they call it to them. And this knowledge can be text files, PDFs, that sort of thing. And so this is a chat assistant that knows all of the ledger activity for January and the balance sheet and P and L results. And it actually does a very like coherent job of talking through those things. If you haven't made a custom GPT yet. Couple, couple weeks ago on the main YouTube channel, we did a video on how to make a custom custom GPT using the example of building an HR assistant. So like uploading your employee manual and any other kind of organizational docs so that your team can just ask an assistant, hey, what's the rule for this or that? And it won't just give you an answer, it will go out to you, to the authoritative source and say, here is what I've gotten from the author. Authoritative source. That probably answers your question, which is what we want. We don't want like a black box AI that just gives us answers. Because just like that, that intern, I'm not going to blindly trust what they tell me. I need a work paper, I need some kind of proof to verify that. Like yes, this is the answer. At that point when AI is giving me proof, then it's saving me time. If it's just giving me an answer, is it really saving me time if I have to go back and re research it? Maybe some, I don't know, but we can get it to give us proof. So if you haven't made a custom GPT before, maybe check out that YouTube video, I'll put that in the show notes also. But this is genuinely something that takes less than five minutes to create. If there's a little bit of nuance here, it is in the instructions that are given to the GPT. So just for perspective, I'll read through the instructions that I gave it here. From the uploaded knowledge specific to this company, retrieve answers for all user questions only using the provided knowledge and explain where in the documents you got got the answer from. If an answer is unclear from the balance sheet and Profit and loss, use the general ledger report to look further into the detail. The balance sheet and profit loss reports contain sections and subsections with the most indented rows being the actual accounts from which you can find detail in the general ledger report. For example, there isn't an account called Utilities. This is a section presented in the report. But under that there are two accounts. Gas and Electric and Telephone. When you mention a specific transaction, always include the dollar amount of the transaction. You'll be chatting with the business owner. Answer concisely in 100 words or less. Always include a dollar amount with any specific transactions. So that's the whole instruction prompt and it took a little bit of iterating and you may have noticed I called out a few very specific things there and it were. It was a few different a few things that the assistant was doing differently than what I wanted. And so I added a prompt or I reiterated a specific thing to say never do this. So a couple examples in this prompt here only Ever use the knowledge that I've given you. So I don't want it going out to the web. I don't want it. I don't want it looking up Craig's Design and Landscaping services online and trying to find additional information about it to answer a question. All I want it to do is use the knowledge that I've given it. I've also got a bit there explaining that the balance sheet and the P and L have like sections and subsections. And that was because a couple times it would see like a utilities total and it would think there was a utilities account, when in reality there were actually there were two accounts inside of like a utility category or a utility group. And when I added that stipulation, it says the balance sheet and profit loss report contains sections and subsections with the most indented rows being the actual accounts from which you can find detail in the general ledger. When I added that stipulation, it never made that error. Again, a great example of how it's really easy to chuck something in AI and be like know how to do this? And then you like add A sentence. And it's like, oh, oh, no, it doesn't know how to do this. I called out to always include a dollar amount if, if you're talking about a transaction a couple of times I asked for the, the transaction detail and it would give me like a list of transactions and the date and the vendor. But no, no dollar amount. Which I thought was kind of weird, but I told it always give me the dollar amount. And it did that. And then I said, answer concisely in a hundred words or less. Originally, when I was like, tell me how my, how the month went last month. It gives me like this full exhaustive executive summary, which is fine and might actually be totally fine for a normie user. They may like that. But in my case, as we're doing this demo and as I'm generally impatient, I didn't want that. So I said keep it under a hundred words so that whole prompt, you can swipe that whole prompt so you don't have to rebuild that thing from scratch. But you can also augment it to add other stuff. Like if there isn't adequate information to answer the question, always say, get in touch with your accounting manager, Tina, and she can give you more information about this one. You could also have it like more explicitly outline where in the financials or where it's getting this information. If you're kind of wanting to educate the client on where could I have gotten this information in the materials that were sent to me so you could augment this to do a bunch of additional things. I was just really impressed with actually how coherent and accurate the responses were. But then there's. There's this element with chat GPT of like, the utility of it and just how, how generally helpful it is because it can do things like make tables and sort and filter these tables, like just semantically, like, say, okay, give me a table of all the general ledger transactions or give me a table of all the transactions in these few accounts. Now sort it by this. Now filter it by that. Now give me an Excel file download that has all that stuff in it, or a CSV. And now, really interestingly so this is a GPT. Really interestingly, you can now tag other GPTs mid conversation to get that GPT to do something in the context of that conversation. So, for example, let's say you got a GPT that'll create a work paper for a conversation or something like that and spin out a word file. We've made GPTs to extract transactions from bank statements. Maybe it's more like an Advisory GPT and it will, it will give a specific type of insight like according to the information that it sees. So this is a single GPT that will chat with you about your month end close. But you can also augment this with any other GPT you want to develop that could like improve the user experience. Now the main rub in its current state is getting a client access to this. So when you create a GPT, you can share a GPT at a public link. It won't show up in the GPT store or anywhere else. The only people who can use it are just the ones that have that unique link. It's just like you know, a spreadsheet or a Google Doc or something like that. Sharing it at a public link. It's actually generally fairly secure because it actually it takes years to like guess a URL. The bigger issue there is like how that stuff normally gets exposed is if you email a public link to a client and then somebody gets access to the email, then they can see it. Like people don't just stumble into these URLs. It only becomes a problem if they get access to something else that person has where they can then see the link. Maybe that's good enough for you. If the financials are redacted, you could certainly redact the financials. This episode is brought to you in part by Team up, helping you recruit top Filipino accountants without any ongoing monthly fees. The difference between Team up and all the other offshoring options 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 can source accountants with experience working at US or Australian firms familiar with tools like Xero, QBO and Dex. Also recruit specialist roles, team leaders, tax specialists, administrative assistants. Thought experiment. What if you had an executive assistant for the first time this tax season? Just. Just throwing it out there. What would they do? Start at that email video I did on the main channel recently. Get help with that stanky ol 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 employee, 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. If your client is a ChatGPT user already, they have a Chat GPT account, Chat GPT team account, they can invite you to it. It's another, you know, $30 a month and you could create the GPT inside of their space. Otherwise you're probably having to create your own like, GPT team workspace and inviting the client to it so that they can then see it. And you're into that, like, for a client, you're into that 60 bucks a month. So it's a. It is depending on whether the client uses chat GPT already. It's not like an easy thing to create a hundred of right now. But I also don't really think that's necessary. Like small firms, like, we don't work on scale. People like to throw around the word scale. Like all it takes is 100 good clients who will pay you really good money to build a mega profitable. Pick a number million to $10 million accounting firm. And so much about AI right now, like it or not, is marketing. It just is. Like, companies are shipping stuff because they don't want to miss the wave. And the great news for us is in the world of accounting firm running buddy, that bar is not high. Like, we talk about this, you know, we talk about this a lot, get more specific and just don't look like every other like, accounting firm website out there. Like, we don't need another templated professional services website. And if you're looking, looking for a way to like, be different, cool, like give folks the option to get access to a chat assistant with their monthly financial statement delivery. I mean, how much you want to charge for that? Like, you could charge a lot for it. Much more than the cost of the software subscription. But even in like, how you hold your firm out and how you present your marketing and stuff like that, like, that's a really interesting thing that you could do. Last week on social media, I ran a list of 50 niche ideas and I actually got some nice messages from folks that are like, oh man, this just gave me a couple ideas for, for this and that. And it was all generated out of chat GPT based on some prompting we've done in this podcast in the past. And so on my newsletter, I put A new section that was niche of the week or specialization of the week or something like that. And it was one of my favorite like niches from that list. I'll pull it up here. Personal fitness trainers with online platform. So trainers who are doing like training online and you may see ads for this stuff on like Instagram, you know, like celebrity trainers who will sell you access to their content. A lot of fitness youtubers do this. Like this is how they make money is they sell training plans and work with like an ex, like kind of a short list of clients over some sort of collaboration platform where they can message and give them like training programs and meal planning and all that stuff. And so this niche idea is personal fitness trainers with online platforms target business size individual trainers or small teams with revenues 100k to 5 million. Key pain point managing diverse income streams including online subscriptions, in person sessions and digital product sales along with related expenses. How do you keep that stuff organized in your accounting file? Obviously tax compliance around that. If you do that. I shared a few ideas. Target Marketing Partners first software platforms listed a couple trainerize exercise.com these are platforms where coaches I promise I'm going to bring this around to the month end Close assistant. These are platforms that let coaches host content online for people to buy and to work directly with clients. I said if you can get in with those platforms, develop partner content, weasel your way into their, their podcasts. Both these platforms have online communities for fitness pros. Like that's a super valuable place for you to hang out and add value. Influencers. So not fitness influencers, but influencers for the influencers. The people who are showing these folks how to build a business. So they're probably targeting athletic trainers and saying here's how you can start building a business online. A good place to start for finding those people could be looking for folks doing affiliate marketing for those software platforms. They are resellers of the software and so they're targeting athletic trainers who are going to then put their content on these platforms. And then last podcasts. Listen to the podcasts in this, this space to find thought leaders. See if you can be a guest or sponsor these podcasts. And I found a few that were super relevant. Certified Personal trainer podcast the business advantage, the athletic business podcast. These are all about like the business of selling personal training. And so if you're thinking about how do I build a firm around a specific type of person and just look different, how do I not look like every single other firm out there? Specificity, that's a biggie. But also like, what's a novel way of delivering the service? Like, you know, delivering monthly financial statements over video. Like having that on a landing page to say, like you're going to get a video explaining what your financial statements actually mean every single month. For some of us, that's table stakes. Not super interesting for most SMBs, they're like, oh man, I would love that. Like, that is such a greater value than what I get now, which is maybe nothing or just some PDFs that I don't understand. This chat assistant, man, like this is so like straight down the middle of the stuff that people are excited about right now. So if you have a more, a more, I don't know, techy type of client to be able to say, yep, I'm an accounting firm who's going to give you a chat assistant with every single month end close. What other accounting firm is doing that? Like you want a way to look, make yourself look really smart behind the scenes. Is there a bit of duct tape and paperclips holding that stuff together? Yes, but that's the reality of like most of the things that we see that look very impressive. There is much more duct tape out in the world than we realize. So is this kind of a bodged together thing right now? Yes. Like are your clients going to be able to blindly trust all the outputs? Probably not. But I tell you what, makes you look smart, makes you look different. And even if this isn't super novel for us because we're the ones that know where to get all the answers already, our clients sure don't. I mean, if I had a dollar for like every email I ever got that was like, why is office expense up or this or that. And these people have access to the accounting file, we gave them the gld, we give them all that stuff and you still get those questions right. I think this is a really cool development. That is just the start because honestly, this may be the future of quality assurance and accounting firms building your own, you know, sequences of, of prompting, chucking in like the journal detail for a month and having a standardized process for how you have AI do like a secondary review for you. Especially the ability to chain different GPTs together. So a GPT is only really going to be able to like, you know, follow so many instructions really well. Like if I give it 10,000 words of instructions, there's a, I think a thing basically called prompt decay where especially the stuff earlier in those instructions, the longer they get it will kind of those will be less sticky and it won't follow that stuff as well. But now that we can, like pull another GPT into a conversation and have a conversation that's like 10 chat messages, all to different GPTs in the same conversation stream that they can all see, you think about quality assurance, like, you think about all the different things within your firm that you can build out into different steps to either, you know, augment a human or help a human do that thing, or to serve as a second set of eyes. How many checklists have we gone through in our lifetime? Can an AI check a box? Check that thing that was on a checklist? Hi, he. Oh, I sure hope so. It may take a little bit of prompting for it to understand what exactly it's looking for in that checkbox. But the hundredth time you've seen that checklist, I'd know just how many brain cells you are expending on looking into every single one of those things that you checked. So this is exciting to me, not only because it's frankly a cool thing you could do tomorrow with a client or put on a landing page and they would think that was cool. But it's probably the future of a whole bunch of stuff. And like, we are, like, we are just in the infancy, itty bitty baby stages of all this. This is the worst that it is ever going to be. As I'm recording this, Google just launched their new language model, which is supposed to be a little bit better than GPT4. And frankly, what excites me most about that is OpenAI finally has a version to ship a new, better model. They shipped GPT4 last March, and nobody has been able to even get remotely close to that in the last year. Which is wild if you think about how much investment, how many companies are working really hard on this, like, they are at least a year ahead of the pack. And what incentive have they had to release anything new when nobody else is even near them, right? Microsoft's just piggybacking their models and still managing to somehow not do as good of a job with CoPilot as the ChatGPT experience. So what really excites me about Google launching something better is now open OpenAI being like, okay, we've been sitting on GPT 4.5 for six months. Here it is. We'll give it to you now so that we're still like the frontier model, the best of the best. And every single time this stuff happens, these applications overnight just get way, way better and way more coherent at following instructions and being able to look into stuff Like a, you know, a ousand transaction ledger PDF. Super cool, super exciting stuff. Can you tell I'm excited? I get especially excited about, like, there's two sides to AI adoption to me. There is the side that our product builders will do where they will build this stuff into our products and make our products better. But that's like 50% of the equation. The other 50% is stuff that's very specific to me that no software platforms ever just going to do for me. They're like, specific to me. Applications where I can build this into my business and really, really helpful ways and make me wildly more productive. That side of it's really exciting. That's like kind of what I try to focus on in this podcast is not the stuff that your PM's gonna build into their system, but, like, what's a helpful, quick little thing that you could build that's gonna make your life easier, dazzle a client or save you some time? Exciting stuff, man. If you do take this for a try, let me know. I'd love to see how it works. Uh, the. There's a link to this actual assistant that I built in. That link in the show notes that you can check out if you want to throw some, I don't know, hard questions at it or something like that. I love figuring this stuff out on the fly and then like, sharing what you're learning so that we can all kind of, you know, steal the best of those ideas and actually build some really cool, practical AI applications into accounting firms, of all things, right? Gosh, sometimes you say it out loud and you just. There's no way around it. We're talking about AI accounting, right? I went to an eye doctor six months ago, and he's like, tell me what you do. And I'm like, well, I'm an accountant. And he's like, oh, so do you like taxes and stuff? And I'm like, no, it's actually worse than that. I don't like calling it this, but people would probably call it an accounting influencer. And he goes, oh, and we didn't say another word about it. But it could be worse, right? I don't know. Okay, thanks for coming and hanging today. I'll see you in the next one. Sam.