494 Do Accounting Firms Need to Disclose AI Use to Clients?

Today, do we need to disclose AI use to our clients? You may be surprised to learn there is one situation where we absolutely do need to disclose it to our clients, but then some others where we may actually want to like, may go out of our way to do so. Some really interesting conversations on this subject came up in a recent sort of series of AI summits that I did. I want to share some of those learnings with you because I think it may change our mental framework for how we're approaching AI in our firms. The scenario where we absolutely need to disclose this to our clients, it is when your prompts are being trained into the model free AI tools right now, by and large, anything that you put into it, the questions you ask it, the files you upload to it largely are retained by the company to be trained into future models. And you'd actually have a whole debate on whether that's problematic or not because LLMs are not built in a way to be able to retrieve specific things back out of them. But let's be conservative here. I would say I don't want under any circumstances for my client's information to be retained by the company for that to even be an option for it to be trained into future models. So this is why I sent people to the ChatGPT team plan where this is never the case. Those prompts are never trained into the models. I appreciate all the folks like spreading the word on this online because there's still so much information misinformation about AI security spookies and so many, I mean the majority of people have still just not got their head around how there's completely secure ways that we can use AI. But I would tell you if you were using a AI tool that did train your clients information into the model, that's something you gotta disclose. I don't recommend doing that. That is, I don't recommend using those tools. But this is yet another reason why we need to be proactive in AI adoption in our firms. Otherwise a lot of your staff are going to be using their stuff without permission. Especially the, the chillins coming out of school. That man just spent the last three or four years of their school career leaning hard on AI tools, right? When they enter the workforce, you better believe they're going to keep leaning hard on AI. And frankly that's a great thing because so many of us, and I'm, I'm coming off a, a run of in person conferences with a whole bunch of accounting AI nerds. And the problem they all have is they're like, oh, we're excited, but how am I ever going to make the time to use it? Like, we have to somehow change our habits to remember to even use AI. Whereas these zoomers, I don't even know. Are they zoomers coming out of school still or is it something else? Whoever they are, they don't know how to be productive without AI. I'm generalizing, and there's some things about that that are awesome and some things about that that are problematic. But the very best, very best thing is it's this. For audio listeners. I'm taking my left hand, I'm taking my right hand, and my fingers are spread and I'm interlocking them together. The best is the combination of the old farts being like, how are these kids ever going to have success in this profession if they didn't have to fill out tax forms on paper like me? And then you got the zoomers that are like, chill, Granddad, I can do this now. And then granddad's like, not as well as me. You know, that whole back and forth, like, the best version of that is the one that is. That is in the middle. And for all the fears right now around juniors and, and, you know, hiring being down for new grads and all that, like, that's a big opportunity in our accounting firms is to pull new hires in that have a totally different comfort level with AI Then when you run that monthly meetup with your team, that's like sharing cool use cases. The youths, they can come in with an entirely different level of energy and inspiration and bring the old farts around. Lot of hyperbole in that whole runner, but you know. You know what I mean, right? So, yes, there is one scenario where I think we have to disclose AI use to our clients, and that's if their information is being trained into the models. But generally we ought to avoid this. So let's say that our prompts are not being trained in the model. Do we still have to disclose the use of AI tools responsibly, securely, to our clients? I will say this. I think this kind of has to be a personal decision for you. I can tell you my thoughts on this is at this point, I kind of just see and see AI as a continuous of all of the tech that we use. There's all sorts of AI and things like historically and tools that we use every single day from the accounting ledger and all sorts of other examples, generative AI. Personally, I don't see that being any different. As we talked about, as long as the information is not being trained into the model. Now there could be very specific to use scenarios where the answer ought to be different. So let's say I work with creative professionals, voice actors, actor actors, people that have a very different relationship to AI digital artists. If your clients are a bunch of never a and I'm not even going to offer commentary on like whether that is right or wrong, that might change what is the right answer for your business. Right. So that's why I say this. I think this has to be a personal to you decision is not to say that AI is not problematic. There's just kind of this original sin aspect of AI where the premise of the whole thing is kind of messed up. Like we're just going to gobble up everything that humans have ever made and build it all into this really powerful tool and not pay anybody any royalties. But it's also like the notion of ever going back is like would be the biggest toothpaste back in the tubing of all time. So it's hard to imagine that as well. Right. So it's personal to you. My take, if it's me right now we're just learning AI as a part of doing work. And like personally I don't even in my firm wouldn't even have an AI usage policy because I don't want it to be any different than the rest of the tech that we use. I think AI usage policies were necessary in an era when you couldn't put client sensitive information into AI. And before the ChatGPT team plan specifically that was the case. There was no way to use CHAT GPT where your prompts weren't trained into the model. Now we have completely secure ways to do that stuff. And honestly I'm only going to pull a tool into my tech stack if I know it can only be used securely. So for me right now, I don't, I don't. That's not something I feel like I need to disclose, but I can give you examples of where disclosing it is almost a business imperative. And someone at one of these AI events shared that when they go through the process of like hiring a marketing agency to work with their firm, they will ask them, show us the ways that you're using AI. And if they're not using AI, they won't hire them because there's loads of ways to use AI and marketing to make you more productive. Right now I can tell you in my media business, like, you know, we do loads of generative AI video stuff for B roll. Like we use AI in Tons of ways. Not just for the sake of using AI, but like to raise the level of what it is or like offer a different, more visual way of doing the storytelling or to better land a joke or something like that. Or even just to make me look more muscular than I really am. Really sorry for all the folks that have been turning out to live events the last month or so. Uh, I. It's two things I get. One, you're taller than I thought you were. And there isn't any way to hear that that isn't insulting, believe me. Cause I've tried. Are you taller than I thought you were? Wow, that's interesting. When I. When I see you on the Internet, you. You just look like a. A small man. Today's episode is sponsored in part by cloud accountant staffing, who helps you find talented offshore accountants, but now also nearshore accountants. So cloud account staffing, it's actually run by former accounting firm owners. They started hiring in the Philippines, had great luck, ended up hiring some local recruiters and the rest is history. But now they're expanding to Latin America and South Africa. And the main benefit of this, if you've worked with folks in the Philippines, is having folks in different time zones makes it easier to align with your work schedule. Particularly if these are folks that are going to be either part of a pod that like needs to really be able to work like live shoulder to shoulder and also if you're going to have these folks like serving clients directly in client facing roles. Now cloud account staffing has helped hundreds of accounting firms build teams in the Philippines, but now I've got more flexibility than ever to learn more. Check out the link down in the show Notes gang, we're coming down to the wire on a limited series of live events that I'm running in partnership with Canopy. By the way, this episode sponsored in part by Canopy. We're running the events and I Austin and Dallas, in Minneapolis and in Boston, super fun half day events. You're gonna get a pile of cpe, some free lunch, fill that tum tum. I'm gonna be doing a keynote, I'm gonna be doing a Q and A at those. We're gonna have something like 70 firms out hanging out. You're gonna meet some other smart folks, we're gonna get to hang and you're gonna leave with some new local friends. Nothing better than in person stuff. Like online stuff is great, but man, like in person stuff just lands totally different. If you are anywhere near there, I would love to see you check out the links down in the show notes for all of those events and let me know if you're going to make it. Second thing people say is, okay, so the muscles weren't real then. And I fear that the whole joke of like, hey, let's do funny AI generated images. It's really taking a toll on my ego right now because I find myself just perpetually letting people down in person. Maybe we swing it the other way and just make them, like, real, sort of weasel, just kind of frail. And then people meet you in real life and, like, you see that guy? That's a fit boy. I'm sorry. For the last 90 seconds or so. So you're not going to hire a marketing agency that doesn't use AI in the same way that right now, in my creative team, if you don't have skills with AI and if you can't use AI to be more productive, buddy, I'm not, like, I'm not going to pay you to work for me. That's costing me money. There are a lot of professions where, like, AI is a prerequisite to even being able to be productive these days. This is another reason why it almost feels like a moral imperative to be bringing our accounting teams along on AI. Like, if you cut somebody loose, you lay somebody off three years from now, and you never, you know, pulled AI into your firm and got them using AI assistants like ChatGPT. That person's gonna have a really hard time finding a job at that point because it's just part of professional work. It is like your typing speed. It is like your Microsoft Office proficiency. We don't even put those things on the job description anymore. And AI is kind of becoming a default aspect of being a knowledge worker. And I would argue that. That this might be an argument to go out of your way to disclose AI use to your clients. Right? And so oftentimes the whole conversation around, like, should we disclose AI use or not? We think we'll scare our clients. And so we feel this initial pang of like, ooh, let's find a reason why we don't have to disclose. But then if you go down your client list and you think about the entrepreneurs that you work with and the ones that would object to that versus the ones that are trying to run a better business with the latest technology that's available, AI or not, don't you kind of want those good clients to know that you're ahead of the curve, that you're like, and you're doing this stuff responsibly? This is a great example of where we can let C and D clients get in the way of where our A and B clients need us to go. And this can even extend to like, hey, like here's where we've learned to work with AI responsibly and it's probably something that you could be doing in your business as well. And as soon as we start thinking about that, we're like, oh my gosh, I'm not an AI expert. I can't, I can't, I'm not, I'm not comfortable having those conversations. Meanwhile, it's like, I mean accountants got tax tasked with cloud tech adoption, like Covid relief. None of these things were accountants jobs, but we got stuck with it. And if we throw our hands up and we're like, listen buddy, I'm not, I'm not an AI expert, so I can't give you answers here. The trouble is AI impacts the accounting department. Like AI impacts small business approach to finance. And if we're not opinionated here and if we don't like take their hand and lead them into this, who are we waiting to do it? Like who's going to do it? Software engineers. Because that's the only other option. The AI experts that are not accounting experts that don't know anything about what we do. I don't want those people like building out their accounting tech stack and telling them how to do the accounting finance and all that stuff. So if not us, then who? And there's a difference between holding yourself out as an expert and sharing your journey. And we should probably be doing the latter. Here's what we're learning. Here are the responsible ways that we're putting it to work. And it is more documenting than holding yourself out as an expert. Lord knows there are enough people holding themselves out as AI experts right now that don't know the first thing about AI. My goodness, so much fear mongering. Still, every single news story, even from like this news story last couple weeks around Chat GPT conversation shares. So in Chat GPT you can share a conversation. And these were being indexed in Google so that certain Google search results could show Chat GPT conversations. Everybody's freaking out. They're like, oh my gosh, all my Chat GPT conversations are showing up at Google. It's like, nope, that only ever happens if you share it. It's also exceptionally unlikely that it ever shows up in the results because, because Google results are a product of like, oh, how many other pages have backlinked to this conversation? And so how many websites are really backlinking to your ChatGPT conversation. But also you can't share conversations in ChatGPT team. So if you're using it responsibly, it's not an issue to begin with. It's one of the great features of ChatGPT team. But the folks that don't understand this are the ones that are like behind the curve and not showing people how to proactively do this stuff responsibly. And so every single time this, these news stories come out. The unfortunate thing is oftentimes AI thought leaders will use this fear mongering to drum up business. And if you have to buy their course and go through their series and all that to understand if it's okay to even use that sucks. I mean, 2025 don't pay all that crap. But this is like how they kind of stoke the flames of that, that voice that's sitting on his shoulder that's like, oh, maybe it isn't time yet. Maybe we shouldn't, maybe we do need to invest in one more thing. Just gotta start. Like we're just, we're just accountants. We are wired to be chronic overthinkers and it comes from a good place. Like, we do need to be very mindful of data security. But ChatGPT teams, it's almost a year and a half old now. There is a newness that wears off with the passage of time. And I don't want people three years from now to finally give in and be like, okay, it feels like it's time now when three years from now it's the exact same as it is today. Now our clients, they're wrestling with their own version of this as well. But I can tell you they want you to be as productive as you can possibly be. Right? They want to understand that their information is not being like trained into an AI model. We can show them that. And as simple as going to the ChatGPT pricing page and understanding like, here are the features to look out for when you're using an AI tool to ensure that your information isn't trained into the model. ChatGPT is using the exact same API on the back end that all the big mainstream services that are using AI are using as well. Notion AI, your ClickUp AI. Microsoft, choose your flavor of AI. They're all pointing to the same models. Microsoft, that's not the case. They've got their own, like Azure version of it. So it's not OpenAI's API that it's pointing to into it. I mean, go down the list. The Folks that are like, I can't use chat GPT, that's not secure. I, I would almost guarantee you you're using a collection of other tools that use the exact same API to work with the AI models that ChatGPT is. And the sooner we can bring people around to this, the sooner we can sort of unlock these cool and really exciting new use cases and be the ones to kind of decide like, what does the future of accounting with AI look like, right? There's all this sort of fear around and it's not misplaced. Like, especially man, the testing I've been doing around tax stuff and it's taking a pile of source documents and putting it in Excel in a structured workbook. Like it's already so good at that. I think I've shared like, I've got like just kind of these seven sort of source stocks, brokerage statements, W2s, Social Security statements, all that stuff. And I can flawlessly put that into a workbook that I give it like pulling hundreds of numbers, doing that correctly in about 15 minutes time. If I gave it 50 source docs, it wouldn't do that. But I would argue the only reason it wouldn't do that right now is that OpenAI won't let the agent work long enough to do it correctly. Because if it can work indefinitely, like if it worked 12 hours, that's extremely expensive to open AI. We're talking about a 30amonth tool. So I'd argue the models are actually already good enough. We just don't have features that are allowed to think or reason or work for long enough to do this job well. But that's just a matter of time, right? As these things get more efficient and we're supposed to get GPT5 this week. As you're listening to this, a brand new model, we'll see, you know, how much time, like how long is that model allowed to think? How efficient is that model? That'll dictate, like if it is a total resource hog, then when it gets announced, you're only going to have so many conversations you can have with it per week and there's going to be all these constraints around it. But the fear generally around like AI coming in and doing stuff that we do right now, it's completely justified. But don't you want to be the one riding the wave? Like don't you want to be the one putting that to work first? Because I don't see that like overriding the trusted relationship. There's so many aspects of change that should have happened in accounting right now, that's taking eons. Like, I mean, people even using cloud technology. On the one hand, it seems like it's all going to happen incredibly fast, but in the past always seems to take longer than you think it will. And in the middle, in this transition, it is the people who are testing stuff, having these conversations with their team every single week. You're gonna be the winners. Like, there's gonna be winners and losers in this transition. And if you're even thinking about it and having these conversations and using it now and testing it and trying to be really first principles about what are our reasons for being okay using it versus not being okay using it, like doing the work of working through those blockers. You are the pioneer. Like, you're the one figuring out what it looks like next. And that's really exciting to me because there is still so much tedium in what we do. If I could pile up the hundreds of hours I spent just annotating PDF documents in my life, I'm not proud of that. That's a thing that I think we could all agree could go away. Organizing documents into like bookmark PDF structures, land o lakes. I'd be perfectly happy if that went away. Chasing clients for documents for a month and close, doing two factor authentic authentication Olympics just to be able to get access to a system we need to get into that is the client system. I think we could all agree that can go away. And we do have to think about, like, how do we make our juniors productive? Because there's a risk that the change happens so fast that we don't know what we want juniors to do then. But what are our options? Like, this is where the world's going. You either ride the wave and be part of the conversation of defining what's next, or you, I don't know, hide under a rock and get surprised by it. We're not going to do that though, right? I honestly think the people listening to this podcast, I've said this before, like, this is arguably among the most progressive folks in the entire profession. And so it's the folks tuning into this stuff. You will genuinely be the ones, like laying the roadmap. Like, we just did four AI events that folks live AI events. About half the people flew in for those events. These are the people dictating what the future of accounting AI looks like. Like a half to two thirds of the folks at those events were podcast listeners, gang. It's y'. All. Like, you're the ones figuring it out. And I would encourage you to figure it out together. Like turn up for events, meet the other people figuring this stuff out. Share something you figured out in the comments. The fastest way to learn is to learn with other people, but I don't think we're going to be surprised by anything. We're going to be the ones riding the wave. Okay, if any of you wave riders are in the Minneapolis or Boston area, I've got events going on there later this week. If you're listening to this as this comes out, want to see their jasononfirms.com events and I'll see you in the next one.