169 Managing Knowledge In An Accounting Firm

Hey, you big smarty pants. Two things are happening right now. One, AI is making information more accessible than ever. And repurposing information and knowledge is getting easier. It's giving people access to that knowledge in brand new ways. And two, you're getting smarter every day. You're having, you're having client experience, you're having conversations, you're learning new stuff. You are becoming a more capable, more stronger, frankly, I'd even say more attractive version of yourself. But all that learning that you're doing, all that stuff in your noodle, if nobody can get access to it, what good is it gonna do? And if all we ever do is just dispense that expertise through the tiny, tiny coffee stick, like, you know, those little red coffee straws that are like a straw but nobody really drinks from them, they just use it for stirring. If all we're doing is dispensing our expertise through these itty bitty little red coffee stirring stick straws, surely that is not the most effective way to put our essence and our expertise out into the universe. Right? And so let's talk about how we manage information, knowledge, whatever you want to call it, within an accounting firm. We're going to talk about three different types of information, different ways to document that, manage that, potentially give clients access to that. And it's going to give you a bottomless fount of wisdom for you and your team to draw on. All made even better by AI. I framed it this way the other day, that you have in your head everything that you need to write a book. The problem is it's not in the right format. It's in your brain, it's in your mind meat. It's not in like a written, cohesive format and a book. But you know everything that you need to know. And if you think you don't, you're probably wrong because we sit there answering questions from clients all day, have all these conversations and a super low leverage medium, one on one, a client at a time, when for every single one of those questions you answer, there's probably a thousand other people that would benefit from hearing that answer. And it's why, like, I don't do consulting. I don't do one on one consulting. I do this podcast, I do the YouTube channel, I do social media. I take questions from people, but only answer them in a setting where tons of people can benefit from them because I'm squeezing the ever living heck out of leverage. Like, and it's not to say like the, the extreme there is like not having one on One relationships and always being too surface level. And so it's not to say that there aren't downsides to just doing one to many. But in a world where we are all distribution limited, like that's our bottleneck. It's the fact that you haven't found a hundred people in the world who most value you for what you do and you never will. Like we are all just distribution limited in that reality, then it almost always makes sense to lean into the higher leverage version of giving that answer away to as many people as possible. But we struggle with a lot of things from being visible online to like even knowing what you're going to say and all that. And so let's talk about how we manage that information in a way that will make it easier to produce content, to answer clients questions, to delegate technical things to your team. So three types of information and how we manage it. One. Internal information. So SOPs, you know, really easy examples like HR docs, your, your employee manual. I am such a fan right now of plugging your employee manual into any other HR docs, any transcripts from meetings where you like clarified those HR docs. I am such an advocate of plugging that into a chat assistant right now and then connecting that chatbot to your teams, your Slack, your discord, whatever you use to communicate with your team members. We all know the pain of like finding something inside an organization. Like what is the policy for I don't know, vacation accrual on a holiday or, or like all these little fiddly nuance things. And most companies like internal management system for those things are just like kind of dumb and they never get any attention and it's. You've just got a bunch of different things that you got to look through. Especially if there are like meetings that have supplemented those things, which is oftentimes the case. You've got these undocumented meetings where they filled in a bunch of blanks, having transcripts for those having all those things plugged into a single chat assistant that can only pull from that authoritative information like excerpts and, and point you to the answer. Not give you an answer necessarily, but point you to the source for the correct answer. That is a million times better than people having to go out and find the answer for themselves. To just have that chat assistant inside of your Slack and your teams, whatever man, that is such a big win and we can all build that right now. The thing, the what I've shown on YouTube a bunch has been tool called Chat Thing, Chat Thing IO you can upload any document you want to it and then connect that chatbot with whatever team, whatever messaging service that you use. And so where does this knowledge live for most people right now? I think for a lot of firms it's living in like a Microsoft SharePoint or even just a file directory. It's like a bunch of word files or PDFs or something like that. It's not living in a place that's very accessible. I know more people are putting their internal firm stuff on Notion, which is pretty cool, you little hipsters and Notions. Couple cool developments from Notion One, their search functionality is actually vector search or semantic search. So it will find things based on their semantic similarity, not necessarily the exact words matching. So we've all had the experience in a PDF doing the old control F, trying to find something and you don't know the exact word that you're supposed to use to find that thing. Like you can't remember the vernacular, but you know, the concept. That's the difference between semantic search and text matching search. So Notion now uses semantic search, meaning if you don't quite know the exact phrasing to use, it ought to still give you results even if those words don't match. Exactly. And then second Notion recently launched this Q and A session sort of tool that's like an, an AI Q&A where you can ask it a question and using all the information in your notion, it will then try to give you an answer to that question along with citations. It's still early days for it. It's pretty cool. There's unfortunately right now no way to limit what it looks into. So like, for my notion, I've got like everything under the sun in my notion because it's like my personal life operating system, but I also have a lot of business stuff in it. So if I, if I ask it a question and maybe it's a question of like, you know, a technical question, like how would I answer this on a podcast or in a, in a tweet or something like that. It's like looking into the gift list I'm keeping for gift ideas to give my wife in addition to like a transcript from a podcast. So the way I use Notion right now is a little tricky that I can't limit it to just specific bodies of information. But I imagine that'll get better over time. Obviously Microsoft has a copilot for, for literally everything that at some point we'll hopefully get our hands on. So maybe the easiest path here, if your internal information is currently in a file directory or in sharepoint or, you know, OneNote or some, some travesty like that, hopefully at some point we get a, an AI chat assistant that can see into that stuff. Honestly, the biggest blocker for people right now is we just don't document stuff very much. Like we are under documented and hyper documented. People have the most to gain when AI makes all that information super accessible. So I've, I've shared this. This is like one of the reasons why I was like, yeah, let's make a daily podcast, obviously to build relationships with new people, to help people. All that hopefully show a generation of accountant that there is a different way to think about accounting firm running than I was ever shown. But also it's a massive mountain of context that AI can access as well. And eventually when we've got tools that like do this really reliably, like we'll, we'll stand up a way for people to search these episodes and all that. And right now, right now the best way to do that is probably Google Bard into a specific episode. Like Google Bard you can give a YouTube link to and it will like search through the, the transcript and stuff like that. But like, even for me right now, what's everything Jason's ever said about X, you know, on the podcast or on the main channel, that's super helpful. Or across Twitter, across LinkedIn, like everywhere, like across the Jason verse, like what is every reference there's ever been to this thing? That's tremendously helpful because like for me, let's say I want to write a tweet about email inbox management. If I have a system that has access to all that information, I can be like, write a 500 word thing on this or give me 10, the 10 most, 10 biggest references to this. And like exactly what I said. That's incredibly useful. Right? But if you don't have that stuff documented to begin with, those tools aren't really going to help you out at all. So on the internal example, when you have that meeting, that's like filling in the blanks on maybe some things the employee manual was ambiguous about or something like that. Take a transcript. I mean I think we should be recording like all of our meetings now, but chuck that transcript into the same documentation. And now the chat assistant for that employee that starts in two weeks can say, oh yeah, three months ago we had a, we had a meeting and Susan clarified that, that you could do this, but you could not do that. And now the chat assistant knows that and that's just going to save your team a ton of time. So first type of information, internal information. You know, people often ask me, Jason, who is this episode sponsored in part by? Well, today this episode sponsored in part by liveflow. Did you hear the news? Live Flow just launched a consolidation product. You actually might have seen it on the main channel recently. We did a whole demo day of it. Liveflow's automated multi entity consolidations. It's beyond simple to use. You can easily map multiple unmatching chart of accounts from multiple QuickBooks Online companies into one standardized report. And once it's set up, LifeFlow's gonna get to work updating the consolidations automatically in real time, the realest of time, so you can focus on analysis using instantly updating data across entities. LifeFlow can even consolidate financials that are in different currencies. That sounds disgusting. Yikes. And it doesn't stop there. LifeFlow offers flexible, powerful reporting tools. Create customized dashboards that meet your specific needs, you little snowflake. Build executive presentations, cashflow forecasts and and more with just a few clicks. The consolidation thing is actually super cool. If you haven't seen that yet, check it out on the main YouTube channel. And thanks to Liffel for sponsoring the pod second type of information, client information. And this is probably the biggest time saving opportunity, but maybe the one that we individually have the least agency in impacting how this progresses. And so if you heard in our, in our AI predictions yesterday, the biggest thing I'm most excited for is practice management systems more meaningfully pulling all that data together in a helpful way, being able to query that. You think about Notion's Q and A feature that they just launched. I would love a Q and A feature into all the client information that I have inside my practice management system, right? Like we are entrusting them with a huge amount of data from documents to projects to email history. Like so much stuff is there. The clock is ticking frankly to do something useful with it or somebody else will. And that might be Microsoft copilot, who knows, like maybe it'll be killer once we actually get our hands on it. But unless the practice management systems give us a really good version of that, like somebody will and that's going to be such a big change for us. And the systems like AI is making information so much better accessible and, and maybe the, the examples we're seeing of this now is like when you upload a piece of legislation and you're like, does this say anything about 1099s or pull five examples of this from like this huge thing and it'll do it like instantly and you're like, wow, that just saved me a lot of time. Oftentimes transcripts of this podcast, I'll check it into cloud and be like, give me the 10 most impactful quotes from this that I can use to like write a social media post about. Boom, five seconds. And it gives me that. And this is a 30 minute transcript. Like that is really valuable, but it doesn't stop there. Like the power of in ve of embeddings and vector search databases. I'll reference this episode again that we did clear back in June, the path to fully automated bookkeeping. I was sitting in a bed in a Las Vegas hotel room. We geeked out on like some benchmarking there where it was like, how big, you know, how much information can you search and how quickly. And there was some benchmarks where it was like they had like 6 million documents that they could query in 6 milliseconds or something like that. So like we can query astronomically large volumes of information to pull the most relevant stuff back. And then right now we wait for the large language model to write us write a response. Like that's 99% of the wait time. But because information is getting so much more accessible, I think what we will actually see is these systems begin to ingest more information than they ever have before. Meeting transcripts, a great example. What would you do with a pile of meeting transcripts if you didn't have a meaningful way to search into it? Right. That's probably why there hasn't been much value to meeting transcripts in the past, is if you talk to somebody for an hour. That's a lot of bloody text. I mean the issue in the past has also been the quality of what it transcribes has not been good enough. Now quality with like OpenAI's whisper model, like is phenomenal. You know, like the auto captions on YouTube are still not that great. Otter is not super great, but. But like the best. Like the frontier models that do this best, really, really good in like a huge number of languages, no less. Which is also really interesting, the whole eliminating language barriers part of this. But when everything's searchable, not just text match search, but semantic search, those transcripts just got way more useful. And don't you wish you had awesome transcripts for your last decade of client meetings? I know that I do. But these systems are gonna start pulling in more different types of information. Transcripts are an obvious one. I think we'll also see pract practice management systems start prioritizing, connecting to your client's accounting ledgers because you think of something as simple as replying to an email, they need a copy of P and L or they have a question about something that happened in the financial statements. If the practice management system is trying to build the very best like generative AI email reply functionality that it can, that general ledger is a pretty big source of information, right? I mean it's. You've already got all of your documents and your files inside of the practice management system. But I could see that, I could see it then prioritizing, connecting to other services to get as much of that information as possible. Because an unfortunate side effect of the current AI stuff that's happening is it is moving us a little back towards all in one. Which is too bad because like the standalone best of breed and then integrate all your different tools, I feel like that creates more innovation. Like you have these small companies that can figure out how to do this one thing really well, like your uncats where it's like oh my gosh, this is so novel, this is awesome. Or I don't know companies like that where they go and solve a single problem tally for and they do that by integrating with other tools. I think AI kind of brings us the other way where we're trending back towards all in one. Because the power of search all in a single place is super valuable. If accountants spend a third of their time on email, the tool that is the best email assistant for you, obviously that's going to be huge, really big business in our space, right? And the email tool that is a standalone email tool like your Outlook, that can't see in anything else, not that great email tool that can see into meeting transcripts, into projects, into files, into into the accounting ledger. The more that it can see into, the better it's going to be. And this is what has me thinking we're all trending back towards all in one. It's also what has me thinking that tools outside the accounting space that do not incorporate email are ultimately going to hold you back here. So click up notion. We have to be thinking about where all of this information lives now. Is it in a place that's going to be accessible to my email clients, like to all of the tools that I use that could benefit from having that information, bridging this gap with integrations right now? Not super possible. Like not. There's not a great way to have a massive body of information in one system and create an interface for an entirely different system to like search that maybe the closest like Analogy you have is, you know, a general ledger, like an accounting platform where there's an agreed upon interface of like accounts and balance sheet and P and L. And like we have a vernacular for what all these things are. But if you take something like notion, which is just completely unstructured data, are they going to integrate with any other tool that I use or my email client to enable to be, to like make an API call to get some sort of search results back or something like that? Just not super practical as opposed to having all that information in a single system. And I've always been a best of breed and integrate guy. Like I haven't really liked all in one systems because they just, they're always a compromise. Right. But unfortunately I think AI is taking us back there. The biggest time saving opportunity with this current wave of AI stuff is greater access to your client information all in one place to the transcript. The meeting that your colleague had with the client yesterday that you didn't even know about. That outstanding request that's in the practice management system, you know, it ingesting the transcript from that meeting and then ticking off those requests, the things that you ask them about in the meeting and pulling in the quotations from the meeting. I have so many examples of how having all this stuff together in one place are going to make us more productive. So that's the second type of information, client information. Reality is that's, that's probably happening in our practice management systems. So internal stuff, it's probably happening in, you know, internal wiki or your notion or your SharePoint or even just a file directory. Second, your client information needs to be parked somewhere that will long term be valuable and searchable. And that's probably a practice management system. Third and last, and this is the one, frankly, I'm most excited about domain expertise. And this is the one that we're probably the worst about documenting. This is the expertise within which lies a book that has not been realized. You know, you do a podcast interview, heck, and get the transcript for the podcast interview. That's 30 minutes of you talking, you know, talking somebody's ear off about like sharing your domain expertise. We don't all run a daily podcast, but what is like your version of that? How can you start getting closer to that to have a better documented, I don't know, reflection of that expertise that you've built up over years and years. In a perfect world, we would all sit down and spend half the day writing or recording a podcast or making a YouTube video or something like that. And the Result would be after a matter of time, you're gonna be super well documented. But most of us don't have the time for that. In fact, most of us go as far as to not spend any time on that because we're too busy like solving problems for our clients one on one. And that is unfortunate because a big, a big factor in how you manage each of these types of information is who you can give access to once you have it organized. So internal information, give access to your team in a better way. In a chatbot, that's way better. Access to client information, I want access to that in a chat assistant. I want that to inform the generative AI like email reply drafts that it gives me for responding to client emails. That is so gold for me. But when it comes to my domain expertise, I want anyone honestly and this, like, we'll see how this shakes out. But I think the right answer is probably giving just about anyone and everyone access to a version of it. 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 first 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 freaked out 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 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 start heading down that path. Consider cloud accountant staffing. So Monday, we built out this, this database of client problems, right? I think that's actually a killer exercise. Let's say you work with beekeepers and you make a database of the 20 biggest problems that beekeepers have. And those problems don't have to be related to tax or accounting. They can be related just to anything about the business of beekeeping. And that may feel like a trivial exercise for you, who has all the answers. But if you think about all the ways that that can be used, once you have documented those 20 problems for the rest of your team, for one, they may not have good, concise answers for, like, how to help clients navigate all those problems. And this is where we run into your team not being able to support the clients that only you can support. And when you're trying to scale beyond yourself, your team's capabilities are not the same as your own capabilities. So sometimes you have to go backwards on, like, client complexity so that your team can support it. But this is a great way to upskill your team on your domain expertise. And so if you have this database of 20 problems clients have, how can you start building assets that tie directly back to those problems? Could be blog posts, could be LinkedIn posts, could be going and doing a podcast interview on that specific problem on a beekeeping podcast. But then when you do that in your database of problems, link out to these assets, grab a transcript of the podcast, grab the text of that, that LinkedIn post. You want to keep pulling that stuff in. Because what you're doing as you're amassing all that information is you're actually building a knowledge base of your domain expertise. And virtually none of us are doing this. Maybe the closest thing we have to this is those blog posts that you wrote for a couple of years. And now the last time something was posted to the blog was like 2021. But nowhere else are we documenting the domain expertise that we have. And when we have super useful tools for querying that and pulling information and using it, even for ideation, the people who are super well documented will have a big advantage over those who don't. And it can be as simple as we talked about, you know, notions Q and A thing that they just launched. I could spin up a notion account and in fact I may try this just to see how good it is because it has the limitations right now where you can't specifically tell it where you want it to search and not search. I'm thinking about just spinning up a notion account, getting the Q and A license and just chucking all my podcast transcripts into that and start asking it questions and have it pull out citations from transcripts of the podcast around certain topics. So for example, let's say I'm going to tweet about offshore hiring. Pick, pick anything. And I've got kind of like a, or maybe it's like a LinkedIn post, 500 word LinkedIn post. And I think I pretty much got it cooked in there. But I drop a, I just drop a question into that notion that has the transcript of, of all those past podcasts and it's gonna give me all the things that I've said about this stuff in the past. And maybe I'm not even gonna give it a. Maybe there's going to be stuff in there that I didn't think about in the moment. Like the fact that that is documented is gonna be tremendously helpful for me. It will help me build a better article when I do that post today because I've given more complete thought to all of the thoughts that I've ever had on this subject matter that I've documented. And this is a really practical step towards that. Book. Writing a book is inevitably work, but like step zero, like where most of us are is all that stuff is just in our head. Like we aren't documenting our domain expertise to give access to folks. And the starting point is maybe social media and like little bite sized stuff like that, maybe your newsletter. But anytime you publish these things, you got to capture them. Every time you publish that thing and you capture it like that's a valuable asset for you, that then is searchable for you and searchable for your team. You could tie it back to those client problems. You can repackage them into white paper or you know, PDF asset or a video or something like that. There may even be a day where there's like an AI chatbot that you give clients access to that can see into all this information and that's, that'd be totally like legit, totally valuable thing. Right? And this might even be stuff that never sees the light of day. Like this can be as simple as voice memos. I'm really interested in how AI will will be like a sort of sounding board that remembers everything. And so we've, we've played around with PI, that assistant on this podcast before, and. And a big thing with PI is the notion that it never forgets anything. Literally everything you ever tell it, it will remember. Like Pepperidge Farm. And there's so much value in like, verbalizing things sometimes. Like, it helps us kind of reason through our thoughts. And oftentimes this is how, like, books come into existence. Especially if you're using, you know, a ghostwriter is you basically talk through these different prompts in the same way that you would with a client. So this is like a very natural approach to working through this for someone who is just used to handling these questions one on one. Because we can talk about this stuff until the end of time. Like we already do it with our clients, right? So it could be as simple as what are common questions that your clients answer or clients throw at you and then answer these questions and voice recorder or whatever, get a transcript and that becomes like the beginnings of your domain expertise and how you capture it. And that's not just gonna be an asset for your team, it's gonna be an asset for you too. Let's say you decide you're gonna write a book and it's gonna be 16 chapters, and here's the outline, and each, you know, chapter has four sections in it. When you are writing those chapters one by one, imagine if you had documented all that stuff already and you could query all of that, like all of those professional conversations you've ever had and every time you've ever shared your domain expertise, imagine if you could query that stuff, how helpful that would be to do everything from write a tweet to write a book. And when you're thinking about running teams and how people with expertise come and go, oftentimes we don't capture organizationally the contributions of that person and their expertise and what they're able to add to the team. That is a great way to make knowledge sticky so that it never escapes the firm. That's kind of one of the unfortunate aspects of, of what we do is you have a meeting and you don't record it and then you hire somebody the next week. Well, they weren't part of that meeting. Or like there's so many of these things that are kind of ephemeral and people that come and go. And as you're looking at what are the assets, the long term assets that you can build inside of an accounting firm, like knowledge. Domain knowledge should absolutely be one of them. But we're just not capturing it because we're not taking the time to like capture that transcript or to record your thoughts, to go on that podcast interview, to like, I don't know, write that 500 word LinkedIn post that one day. I'm super bullish obviously on doing a better job of capturing our domain expertise and that, that body of work building over time to be an asset in your firm, not just for you, but for the people around you and maybe even someday like directly for your clients. Worst case, in the short term it's going to be a killer like second brain for you for building out writing for, for posting on social media, for answering client questions, all that stuff. I'm actually in the process of like trying to build this out in a few different ways for transcripts for my content. And it's actually been a really fun exercise. The quality of what you can get with AI right now is phenomenal. It's so good. It is really, really good. Without getting too into the weeds. There's been some technical testing around like the quality of like how much better AI has gotten about answering very specific questions from a large body of context. So in the past the issue was always context limits that a large language model could only remember so much information at a time. We've like found workaways, workarounds to manage that not really being a problem. But also every language model now has much higher context than they did six months ago. But testing of like hundreds of pages of documents where within that document is like a one sentence statement that answers the question and you ask the question and like 99.5% of the time it will pick up that answer like that needle in the haystack style thing. So there's been a benchmarking of different ways of building this from a technical standpoint and how good different language models are like retrievals versus just all doing it with a language model. All nerdy stuff, I know, but like the best of the best ways of doing these things right now are unbelievably good, where you just have these massive amounts of text and it will pick out just like the most fiddly little thing that will answer that question incredibly reliably. It's what has me super excited about applications for tax research and ultimately about the value of capturing information. I think that's, that's our bottleneck today is we just aren't documented at all. So when we have that technology, it's not going to be useful for us because we didn't document it. But we can start doing that today? I don't know. I've been thinking about, I don't know, a way to give, like, the best way to get access to this out into the world. Once I put this together, maybe we'll just stand it up on, like a. On a webpage and people can just query and chat with it and, and see where it goes from there. But it is amazing how much information it can, like, sift through, like, near instantly. It's really cool. That's all I got for today. If you have examples of how you document your domain knowledge or make that a habit to do, like, every single day without, like, sitting down and being like, okay, I'm going to write a blog post. You have any ideas of how you can capture that stuff every single day so that that builds over time as, you know, part of a routine? I would love to hear it. I know, obviously, like, my podcast is kind of my forcing function for that, but what's the best way to do that inside of an accounting firm? Still kind of thinking about that. Thanks for coming and hanging.