103 AI Chat As A Competitive Advantage for Accountants

Team, check this out. We've been talking about this a little bit. The whole notion of having AI chat in your practice management system, in your portal, specifically a client facing version because that is where all of your goodies live that their ChatGPT doesn't have access to and all that. What is the moat that we ultimately have left when general AI tools get really, really good? It's all the client data that we have access to that nobody else has access to. So how are we going to put all this data to work behind a helpful AI interface? Uh, let's do a deeper dive on this. This whole notion of having AI chat within your practice management system, is that the right place to do it? What sort of functionality would that enable? Come on in, let's talk about it. Okay, so we've been doing this kind of sort of series of AI plugged into xyz. We started with AI, your practice management system on AI. Now we've done meetings, we've done email, and I want go deeper specifically on what AI client facing AI chat within your practice management system could be. And if you think about all the forms that client facing AI chat could take, particularly through the lens of people who do tax, people who do accounting, we've already seen like what early client facing ledger chat looks like. So the ability to talk with your accounting ledger digits is doing that now. I think we're kind of getting a taste of what that'll look like. But if I think about the whole relationship that I have with my client, from the emails that we have and the meetings we have and the projects we're working on, past tax filings and all of that, if I look at how chat gets layered over the top of that, who's going to do it? Like I think it needs to exist in some form or another. There's going to be this element of, and I think we're already kind of there, people have this expectation of being able to self serve. It's like it's to me it's the one non negotiable aspect of portals right now is you don't arbitrarily need to have a human that is a blocker to them being able to pull past tax filings or something like that. But when all the other like modernized systems they use have a chat experience, it will pull helpful stuff for them while leaning into some sort of domain knowledge. Is the accounting firm experience always just going to be dumb? Right? Like is it always going to lack that when all the other services that we use have it? I'D like to say no, but there's still a big question of who's going to do it. Right now where all that information lives is your practice management system, hopefully. Or at least that's where the most of it lives. And practice management systems already have these client facing portals where we're trying to get them to go out and address requests and that sort of thing. So this seems like the most practical place for a client facing AI chat experience to live. Something that is holistic of like that entire client relationship. Now there's a lot of like question marks around what that will do and what that won't do. That's part of why I want to talk through this and kind of flesh it out a little further. I want to get your feedback of a additional cool things that it could do, of things that it could do that could be problematic. Because what I don't want is for the client to forever have this interaction with firms that feels sort of archaic. And to be totally honest, like I think we've been struggling with this for the past decade. It's taken a really long time to get like modern tooling for an accounting firm to work very well and give clients a modernized experience. Cloud practice management systems have come absolutely light years in the last three years. Every time I go moderately deep into one, I'm just dazzled at how far. Even in the last 18 months they have all come. That is a recent development. But for the longest time it's been really hard to offer that modern experience. Honestly, I would argue even that still today if you run a firm beyond around a hundred employees, where most of the cloud practice management systems like they're just not there yet, some of them are kind of trying to get there. But if you run a 300 employee firm like good luck building a software stack around like a modern client facing experience, like that's still virtually impossible to do. So to bring this back to AI chat, I want to continue this conversation because I think it is an important factor in like continuing to modernize how we work with our clients. Especially if this is a trend that's being built into the other tools that we use. I selfishly know there's a lot of product people that listen to this also and I want to be planting these seeds for them if they're not already working down this path now to like zoom out. There's a huge opportunity here in terms of addressing sort of a, an existential crisis of the larger profession that is to say, when somebody makes Mega Advisor. So we talked about this A couple few episodes ago, the whole avatar of your expertise sort of thing went into it, or whoever it's going to be does what, you know, the company behind PI, the voice assistant, does, where they invested $1.3 billion in this personal AI assistant for you. If somebody goes out and does this around tax and accounting and having this technical expertise and it's really, really good, ultimately, then what's left for me? Like, where are my scraps? Like, where can I continue to support my clients? And I think the answer is in the greater level of nuance that I have and the greater specificity that I'm knowledgeable around in a specific domain. The general tools will be the general tools and they'll be really good, but they won't go as deep into specific domain things. So for me, the solution is specificity, but it's in the information that I have access to that that general service does not have access to. Right. And for me, with many of my clients, that's a decade of communications, of meetings, of tax filings, of month end closes. It's all of these things that I have access to. So for me to put forward like this modern client experience, and I don't know if this is a 12 months from now question, a three years from now question, a five years from now question, but this chat experience that I can put up in front of my client is actually a super help to defend against the improving quality of general tools. If my client has a helpful assistant that can see into that stuff and somebody puts out this really cool sort of general assistant, in my mind, the client's like, well, I guess that's cool, but why would I use that when I have this version over here that has access into, you know, all of my historical stuff and my filings and all that? Hey, I wanted to tell you, this episode is sponsored in part by the lovely folks at Copilot. Oh, you know what keeps me up at night? This is probably going to trigger some of y'. All. What if you move your whole firm to a practice management system, you train the whole team, you move all your organizational data in there, you train all of your clients on how to work with you in the client portal, and then you don't like it or they stop developing it or something like that. You know how sticky these things things are and what a mountain it will be to retrain your whole staff, that's one thing. But to retrain all of your clients and say, hey, just kidding, we actually want you to come use this system over here now. Yeah, Sorry about that. Let me tell you, you know, a solution to that is to separate your client experience from all the systems that you use to manage your firm. Right? Think about it. So Copilot, one of the things they do really well is they are just laser focused on the client experience. They just want to be the client communication layer. And why I can appreciate that the value of that is it gives me the power to change all the paperclips and bubble gum and strings behind the scenes that are holding everything else together. So if there's a super cool new AI this or that, that's going to do my work better for me, I can pull that in without having to change the client experience. Pretty darn good argument for the apps that are just that. Client communications layer, right, Copilot, You've probably heard me talk about it before. A super flexible client portal layer where you can like customize what each of the clients see and even embed stuff from other apps into that portal experience. To learn more about Copilot, check out the link in the show Notes. Hey. This episode, it's sponsored in part by the fine folks at Client Hub. Thank you, Client Hub. Hey. Last week on Tales from the Hub, super smart accounting firm was super pumped about Client Hub's new vision for how AI gets pulled into the product. They're calling it your firm on GPT and let me tell you, super smart accounting firm. They just joined the beta program for the new email functionality. In a word, they're blown away. Client Hub now connects to email. Platforms like Microsoft and Gmail will pull those emails into the system. They'll take long email threads and summarize them in one click and even take a reply and change the tone of that reply to make it just right before you hit send. Pretty cool. And that is just the email integration feature. Client Hub is working on a bunch of cool AI stuff. You want to be part of that pilot program? Taste the rainbow. Get on board. Check out the link to Client Hub in the show Notes. So to get more in the weeds, specifically of what this AI assistant can do, I think so this ought to live where our clients already hop into portals and do stuff. Some practice management systems have good native mobile apps. Now, some of them don't have that yet, but virtually all of them have like a web portal where you log in and you've got files. Some of them will show status of outstanding projects. Some of them you can do like invoicing, billing, that sort of thing. Some will handle, you know, uncategorized transactions and handling Questions like that. And the biggest point of resistance that we face as advisors is oftentimes there's value to us in putting a client portal out there, but not value to the client. And so we have this situation where we really, really want the client to use the portal because it makes our life easier and it doesn't really make anything easier for the client. Nobody wants one more place to message, like they don't want to communicate with you via it. Some of them may go out and like self service and fetch files themselves because that's faster than having to ask somebody. But we're always looking for honestly like more compelling reasons for clients to use this app. What is it a solution to? Because there's very obvious benefits to like them handling all of the requests that I send to them within that portal. Like it's really good for me if they will become best friends with this portal. So I'm always looking for what is that next really compelling rationale to get them there. And chat can certainly be that. So what are things that it could do right now? Fetching files. Most practice management systems now you have a file system for a client and you have kind of all of the files and then you tag files to be client facing like the client, they're visible to the client. That's something that's not exceptionally hard to do right now. The ability to pull up files either by the similar similarity of their name or even according to the contents of the files. You know the kind of super basic example in the US is like we're always getting people asking for their prior year tax return. That's something that the AI can absolutely do like fetch for them. And in the context of how AI is improving email, we talk about how hey great, my assistant now can like suggest a reply that has that prior tax return attached. But in the context of chat, like that's just the client chatting with an AI assistant and it going out and fetching that file for them. Now notice as we talk about this, like the things that make this assistant helpful for the client, many, many of those things are the same things that would make that assistant helpful for me, the advisor user. So as we're thinking through this, I think what it gives back to the client is oftentimes what it also suggests in like a generative email reply. So if client sends me an email asking for a thing, AI assistant goes out to past projects, past emails, meeting transcripts, file contents, pulls the most relevant info and then tries to generate a helpful and correct response that it suggests to me it doesn't just send it out. We've talked about that through the lens of email in a past episode. But that's by and large the same thing that we're going to be asking of this AI assistant to send that information back to the client. So stuff it'll do. Fetching files for the client, recalling past communications. Man, I would love it if I have. I always had clients with just comedically bad memory and they would ask me the same things over and over again and even come in and say, I know, we've already talked about this. What did you say about xyz man, that would be gold. If I've got meeting transcripts, past emails for chat, to be able to either show them the transcripts that's most relevant or summarize those transcripts. One of the two, I'm not sure what would be best. So fetching files, recalling past communications, status updates where projects are at. By the way, I think you probably leave it up to the firm admin to toggle off and on what all these data sources are. Right? Like what is that chat assistant able to talk about and what is it not able to talk about? Ledger connection would be interesting. More and more practice management systems are integrating with the ledger. When you do it makes some really helpful things possible, like managing uncategorized transactions from the portal, that sort of thing. I selfishly know there's a lot of product people that listen to this also and I want to be planting these seeds. But the ability to then have a meaningful conversation about stuff that's happening in the ledger, like why was office expense up last month? That's kind of going another level of depth further. And admittedly like from an engineering standpoint that's more involved to build. But I also think that you could see a service like digits or you could see, you know, somebody else building something similar where that ability to interact with the ledger is actually more of a service. So developers can go out, plug into this service and it will help their AI assistant intelligently chat with that ledger. Like, I think there's a, an intermediary opportunity there for creating a really, really good version of that ledger chat. Same thing with being able to make sense of tax returns. So you've talked about, you know, chatting with documents with AI right now. That's one of the first like razzle dazzle AI use cases, I feel like. So you can upload a document and then you can chat with it. Well, man, that doesn't work for tax returns because if you look at a tax return Semantically, like it doesn't make any sense. You've got all these boilerplate lines and then just a number. And so that stuff like if you're chatting with a transcript, Cool, that's easy. That's all semantic. If you're chatting with a tax return, that's like not going to mean anything to anybody. There's a really helpful service to be made right now that semantically makes sense of a tax return and that's by literally like conditioning the text to be like box X, you know, a number and then actually an explanation of what that box is. If you think about a tax return and if you think about how like a tax bro could create a transcript of a tax return, like reading through it line by line and explaining what every single one of those things is, a service that just does, that would actually be really helpful right now. If I have a practice management system and I want the client to be able to chat with this AI assistant and pose a question about a past tax filing and I have this like semantic version of a tax return, right? Like an explained tax return, not just this boilerplate form with a bunch of numbers, it doesn't make any sense. If I have a semantic version of that, I have a much greater likelihood that the assistant can give something helpful back to the client who is querying that tax return or looking for a specific thing. Same thing really with month end closes. And I does like a so so okay job of comprehending what is happening on a balance sheet and financial statement. But it could definitely be improved. But ultimately if I want to myself have a reasonable defense against Robo Advisor, kind of already means something, right in the space of financial advising. But this sort of AI assistant for tax and accounting questions and all that, if I want to have a great defense to that, then being able to give my client a client facing version of that that can see into all their historical stuff, all the history, the past filings that we have together, that would be really cool. And I think our practice management systems are probably best suited to enable that for us. Now there's an interesting version of chat. And so like an obvious concern here is, well, what if it tells it that maybe I wouldn't agree with or isn't correct, or what if it hallucinates? Honestly these days with GPT4 and with some additional tooling that kind of puts this stuff on rails and makes it real explicit what sort of things it's able to chat about, what it's not able to chat about. So if you ask it about a presidential election. It's just, it's going to say, sorry, I don't talk about that stuff. Like we have tooling that will reliably do that for us now. But there's still a risk of what if that thing just leads to an insurance claim? Like, what if it says something that I wouldn't say? Like, I'm pretty sure that at this point we can engineer our way around that. But let's just say it does happen. Like we have to have, I don't know, it's something that we have to have a conversation about and be mindful of. This episode is sponsored in part by the fine folks at Cloud Accountant Staffing. Do you hire accountants? Bless your little heart. Not the best part of the job, in my opinion. Not something I ever enjoyed. Well, listen, you can build your accounting dream team with talented offshore accountants in the Philippines that work 100% full time for your firm. Their accountants aren't freelancing or contracting for multiple firms. They're all your they work exclusively for you and are incentivized to stay with you and your team long term. They're not gonna get swiped. Cloud account staffing is 100% dedicated to the accounting industry and founded by a former accounting firm owner that understands your business, knows your pain points. They had to hire some accountants and they said, you know what, we're gonna build our own pipeline in the Philippines, gonna pull in some super talented people and then open that up to other firms. Basically that's the story I've been talking about a lot about staffing, building more resilient staffing pipelines for your firms. I had staff in the Philippines like totally red pilled me to like, oh geez, like we need to globalize the way that we get our work done. Check these folks out Link in the show description Cloud accountantstaffing.com Back in January of this year I put a couple of polls out on Twitter and it was around what the very first AI chat assistant for your ledger could look like. Back kind of before we had digits working on it, we also had genius sheets doing some work on it. We're just kind of thinking through what that could look like in the current sort of GPT AI era. And there were people in the past that did like natural language processing ledger stuff. Chattah AI did it. QuickBooks had their own sort of assistant that they built and I don't know that anybody used, but I specifically was thinking about like what is an advisor in the loop version of this look like? Because what I want is for this AI chat assistant to be like my first defense for client questions. Not because I don't want to talk to the client, but because most of the questions my clients ask me are really dumb questions and questions that they've even asked me before. And so like most of them would be perfectly happy. Just like chucking that into a chatbot and being like, hey, is there an easy answer to this question? If not, great, it's still gotta go to me. But I gotta think that first line of defense could knock out like 75% of those questions. And oh, buddy, if you want a way to crank up your practice, practice management monthly fees, like something that would save me a mountain of time like that I would be first in line. But we did kind of a temperature check of where people were at and this was January. And maybe things have changed since then on the notion of an AI just like answering my client's question without me maybe even knowing, like, is that uncomfortable? So I posed the question, if you could give your clients access to a general ledger connected chatbot as like a first line of defense for clients to ask questions of it, why is office expense more this month, stuff like that, would you turn it on for them? Like would you flip the switch and let them have access to that? 40% said absolutely, 46% said maybe for some. And 14% said nah. Which surprised me. So only 14% are just ruling it out. But then I asked a follow up question. What would your appetite for risk be on the types of questions it would answer without your insight? So let's imagine this variable sort of confidence level where, where if it's super confident in the answer, great, it'll give it to the client. But I'm able to set exactly how much confidence it needs to just give answer versus saying I got to defer to, you know, your big brain advisor and a human will actually answer this question for you. Only 11% of people said they would crank that all the way up Mad max setting. The AI could respond to absolutely anything every time. Only 11% of accountants said they would let the AI just have free rein. 53% said something in the middle, 35% said they would be pretty cautious with it. And I worry that, that what people will say and try to get to is like it's not worthwhile until it's perfect. Until it's like this thing that will answer any of the questions. Because if it's just going to ask the advisor, then what's the point? Let me tell you first of all, like I will repeat man, so many of the questions I get are so basic. Like I know an AI chat assistant can handle them just fine, or like pull the ledger detail or the transaction in question. Like I can totally do that for them without my intervention. But here's the other thing is even in situations where the AI assistant says, sorry, I can't give you a complete answer for this, let me pull the advisor in which this is like the fail state, right? Like this is the worst case scenario, they went to ask it and it's going to say it doesn't know or it has to defer to me. Even in that situation when it pings me and says, hey, I need help with this when I come in, that AI assistant can for me present everything that it thinks I'm going to need to answer that question. So even in the worst case scenario, I'm a whole lot better off than having to go out from scratch, log into the ledger, like figure all that stuff out myself from scratch. And that's the worst case scenario when it won't give that answer to the client. Man, that seems like a win to me. And obviously this stuff's going to keep getting better. This is the worst that it ever is now. But I love the idea of this sort of really intelligent AI assistant built into my practice management system that the client has a degree of access to. But otherwise this is still being helpful to me and, and even, I don't know, like another sort of version of the same thing. I was talking with an IT company who at one point had was using an auto reply. Anytime they would get a new ticket, it was a managed IT company. Anytime they received a new ticket, there would be an auto reply that would include what they thought the right answer probably was, but still said they would get back to them. So it was like that acknowledgement that like, hey, we got your question, we're going to get back to you as soon as we can. But also after that it said like, hey, based on our help center and what your question was, we think this is probably the answer. If it was, let us know and it'll auto resolve the ticket. And so we're talking about this AI assistant where clients can chat with it. And it's probably more or less the same technology that's like doing generative email replies and like suggesting to us how we should reply to that email. Would a version of that even make sense where it's like an auto reply saying, hey, I got your email based on the question you just asked me. Here's what the answer Might be. And here's where to find the information that you requested. If this answers your question, can we just skip, skip the conversation? Like there's like there's gotta be some version of that that the assistant could enable that could save you a whole pile of time. And I'll tell you what I actually like about that setup. The whole autoresponder is you don't necessarily have to force somebody through the app. That's not to say that they could use the app, but if somebody's a client, and that's an important distinction is it would have to be a client and identified as an email that's associated with your clients. Right. But any old client sends you an email. I love the idea of the autoresponder answering their question 50 of the time. And I know that we are going to be super mindful of this not being personal. This like we don't want to lose the personal touch. Particularly in the age of AI when people are doing more stuff, talking with more robots every single day. Like there's going to be a greater appetite for the human stuff, which is really where we fit in. So we're going to have to be really mindful of not making this stuff feel impersonal. But if somebody sends me an email and they get an auto reply trying to answer that question, are they going to be mad? Like, is the sender gonna be mad if it does answer the question? Or even if it just tried to be helpful? Like that feels like the most friendly, low stakes way to try to get them an immediate answer. Because at the very least you've acknowledged that like, hey, I got your email and maybe you've got a boilerplate kind of reply there of like, here's the timeframes when I try to get back to people. But if under that it actually tries to answer the question, I can't see anybody being upset about that. Right? And honestly we talk about this, this future timeline where like we're all using AI to reply to our emails and stuff like that and then you just end up with RO talking to each other. Maybe that's how it starts. Maybe that's how all of it starts, is like the instant responder. Here's what I think the answer to your question sort of thing is. Like maybe that just gets so good that that's going to knock out 80% of the questions. And that's like the first step to fully automated replies. Because that's a hard thing for me to get my head around, like just ever giving something agency to just Auto reply, I love the idea of it suggesting replies for me and pulling in stuff that I need to write my own reply. But auto reply, I don't know, but maybe that is like that sort of happy medium is the auto reply that says like, hey, this might possibly answer your question, but I'm still gonna get back to you right away. Unless they somehow, you know, opt out of a reply or say like, yes, this did answer my question. Imagine, whew, okay, somebody build this, put together a 60 second demo, sell the heck out of your practice management system and then shoot me a royalty. Imagine rolling up to work in the morning, you got six emails overnight and the client acknowledged that these four auto replies answered their question. Boom, you are four out of six emails ahead of of where you were otherwise. Right, man, how cool would that be? We got a bit off the rails from the whole like chat and clients having access to a chat assistant, but I think this is a good conversation to keep leaning into, keep developing. Our practice management systems are getting more and more mature. They're leaning into really good AI stuff right now, which is awesome. But what are like the big meaty needle movers? There's some, there's some big time saving needle movers for us internally. But what are the big meaty things for the client experience? I want to give them access to a version of instant answers and it's not going to be as good as me getting the real deal. But man, I'm asleep at 11 o' clock at night. I see that you're messaging me and emailing me, but I'm asleep right now. So if my practice management system can help me knock that stuff out, you're actually going to save me a ton of time. What do you think? What is like the best? I don't know, there's. There's a lot of really hard things here because you don't want to seem impersonal and you don't want to delegate your client relationship to AI and all this. Right? You have any good ideas though for how we could lean into this like helpful stuff that it could do for the client in a way that isn't going to make me look like I'm just putting you off to a robot or something? Because that's the best or the worst version of it, right? Is making somebody feel like disregarded or delegated to a machine. Like that's. We don't want to do that. That. What do you think? We'd love to hear about it in the comments. If you're a product company working on this. Please do tell as well. Feel free to get some free promo in them comments. Thanks for coming and hanging. We'll see you tomorrow.