030 Design The Robot Assistant OF YOUR DREAMS

If you could pick up a bot to do anything for you, no limits, dream bot, take any work off your plate, what would it be? What would it be? Could be anything. Rounded up like 30 answers from folks. I would say a lot of these things are actually pretty close. Now let's do it. On this very self indulgent installment of Jason Daly. Okay, so Nio Carter Gray kicked off this discussion on Twitter. The exact tweet was, if you could hire the bot, you know, the one they say is coming for your job, what's the first task you'd want to get rid of? Essentially, what would you want the bot to do for you? Of all the things it could do, what's the first thing you would want it to do? Obviously, because we're accountants, we, we all threw in like very nuanced, very specific to us types of problems. About 50% of them that way were that way. About the other 50% or so I think were relatively similar. So let's cruise through these. Interesting to me, a lot of these things we already have like some version of or it's 12 months or less away. Like, it's just. As I went through all these responses, it was a reminder to me of just like how much of what we're doing right now is like actively being disrupted in a positive way. In a way that's like, awesome. I would love, like, I, I cannot stop doing that thing soon enough. I would love to not have to do that thing anymore. Okay, let's cruise through these. Mike Valenti. This was kind of an interesting one. He says, eliminate the 60 newsletters I get and curate a morning brief of news articles that actually apply to me and my clients. Mike's handle is Ria Tax Nerd. I suspect this is like tax newsletters and like news updates. And that, honestly, that was like, I can relate to that. That was a pain when it came to like keeping up on current tax legislation and case law and all that stuff. Because you have all these clients with very specific situations that apply to them. And then you have all of this fiddly, these fiddly updates coming through. And how do you apply those updates to specifically your client base? Or when you read that thing, if you've got a thousand clients, what's the quickest way to surface, like, the seven clients that, that may be relevant to where you're like, hey, maybe I should send this thing out to them? I think this, not just a tax thing, probably applies to like any sort of compliance you provide for clients in the US if we had More intelligent tax software where we could like on a more real time basis get access to all of the things that your clients have, like client properties, like numbers on forms, stuff like that. Right now the only way to get a a summary of what is present in all of your tax reporting is like a manual export that you run. As far as Mike's wish list item here, I actually think we have the tech to ingest a whole bunch of newsletters and classify those items based on who they would be relevant to. I actually don't think that's too hard. The harder part of that right now is then like marrying that to the client specific data, wherever that system lives. Tad Johnson Technical Review of Tax Returns and Financials so we're getting some movement on this around in quality assurance around accounting and like month end closes. A number of tools doing this now scrutinize, verify, queue. Seems like there's some others I'm forgetting where the idea is you can just plug them into your accounting ledger and they will do a technical review for you. The upside is they will do this technical review in a standardized way across all of your clients, which is really helpful. If you've ever manage an accounting practice across a team, getting all of those books to the same standard, that's really hard, especially if you have multiple technical reviewers and just doing that in a standardized way is difficult. Those tools are getting better really fast. The problem I've run into with those tools today is the level of things they look at is a little simplistic, not quite to the level that I would want them to be reviewed before going out the door. And so even though it can identify some things, it's not gonna identify as much as like a seasoned reviewer will identify. So if you still have to do both of those things, did it help? The way I see those quality assurance tools today is through the lens of them being like a helpful self review for a bookkeeper. If you ever manage bookkeepers, even if you do the work yourself, you know how hard it is to get your brain to then self review after you've done all of the work to stop to like set it down, to come back the next day and then go through all of your own work again like with a different perspective, the perspective of a reviewer. That's just a hard thing to do for whatever reason. A really helpful application of these QA tools today I think as is as like a self review kind of buddy so that it can shine a light on maybe any obvious things that you missed. It's better than Nothing, right? Like an automated self check. I love that. Building that into like a AI bot seems like something we can get, right? I don't know if that's like a next 12 months thing or like a 24 to 36 months thing, but I will say the ability of GPT to take a whole bunch of unstructured stuff and then find the associations between those things and discrepancies. It's really good. One GPT specific thing we're seeing right now is there is a plugin for ChatGPT called the Code Interpreter and it gives you the ability to upload files to ChatGPT and then perform analyses on those files. I'm really excited to get access to that plugin because we've already seen it do some cool stuff with CSVs of transactions. Taylor Hartman's been showing that off on Twitter. I shared some of that in my weekly newsletter newsletter, JSON cpa. If you're not on that, that code interpreter plugin will let you upload files to it and then you can basically instruct it to do different things with that file. So for example, a technical review of financials, maybe that is uploading a set of financials balance sheet and P and L uploading all the source documents and then having Chat, talking with ChatGPT to go through like, you know, does everything tie out the way that it looks like it should? Are there any obvious discrepancies here between the financial statements and the underlying supporting documents? Are there variations in the month over month financials that maybe we ought to have more supporting documentation for? I think the first step, similar to QA tools, I think the first step to that being helpful for us is, is like this self review level where it's like if I can in virtually no time and no cost bounce this file off of this thing to like check for any obvious errors like why not? Right. As a kind of self review tool, I think that's probably the first step. The holy grail, like the destination that you hope that you get to someday is the technical review, like for it to be able to do the final sign off. But that's like, that's like the hardest thing, right? That's like the last thing. If you're solo and you do all this stuff yourself, boy, I would sure love having this service to bounce that stuff off of as like a second set of eyes kind of. Right? So it wasn't just me. So I think the first version of that is probably like a self review kind of support tool. Sean Marley Doom Scroll Twitter for Me Chris Lazeri Time Entry I feel like we've made some progress on like automated time management apps in the last couple few years. If you do want to track your time, there's stuff like what is it? Rise, R I Z E that will now track the apps that you're using and like the specifically the titles on the Chrome tabs that you're in. And so for an accountant, like that would mean if you're in a QBO or a zero file for a specific client for an hour, it'll actually detail out. Here is how you spent 9:30 to 10:30 yesterday morning and it will show the applications that are like active and focused on your computer. And I it where it was helpful for me, it was really helpful to retroactively go back and be reminded, all right, that's what I was doing at this time. But it doesn't go as far as classifying like what client that was. And so an interesting version of that that actually somebody could go out and build today. There's so many. There's so much new technology that hasn't been productized in helpful ways yet. This is like something I keep coming back to is like, oh, we actually have all the tools to do these things. It's just nobody's productizing it. Like the email generation thing. Canopy launched. We talked about how you could have done that back over two years ago now. And there's another example of those things. So if you take something like Rise, Logan Graff has been sharing a lot about Rise on Twitter this year. It's kind of like automated time tracking. If you marry that with a client list and, you know, say you bounce the title of that application that you were working on and maybe that's something like Google Chrome 0 Jim's Car Wash. You bounce that off GPT or something and say, hey, here's my client list. Like, what client is this to automatically make that classification? I think you could totally have like largely automated time tracking that probably got you 90% of the way there. Maybe there's a few things that you'd work on where maybe it wasn't quite sure, but if it could see what your clients were called and maybe if it could see your time history that you've already completed, that's that would probably be a big driver of accuracy. If you've been working on this thing all week, that makes it a little more likely that maybe that's the thing that you were working on. I feel like that time entry could be largely automated. Blake Oliver Managing the Team Automate those weekly or daily check in messages, request status updates, send reminders. There's nuance to it, but if the bot could draft those messages and I just hit send each morning, that would be amazing. Check it. Like check in stuff I think is like suited for bots and there are tools to do that now, like daily standup sort of things. So I use that with my team for several years. It was like a daily check in bot in the morning that was like, are there any blockers? You know, is there anything that you need today? And then how are you feeling on an urgency standpoint? From like totally swamped. I need help to. I'm pretty good and I could help somebody else. We did that stuff in Slack, so we, we went through a couple of check in tools that we used in Slack and those worked totally fine. They. But they didn't. And maybe this is more what Blake is getting at. They didn't request those updates with any sort of contextual nuance of what was going on that day. Or like, you know, is there. Would it a better version of that be something that could see into like your ClickUp project management or something like that that we were using at the time? I don't know that like that nuance and that like how to do that interaction, interaction with the human feels like a thing that's harder to automate and maybe ultimately one of those things that even if you could automate it, to me that would maybe be one of the last things I would try to automate. Just because like that kind of comes down to the, the squishier, softer touch points with team members where maybe if they're being asked that. One thing I learned like using a robot for daily standups is people will interact with a robot differently than they will with a human. And it's like they just know they're supposed to do that every day, so they just kind of do it and over time they put less thought into it as opposed to. If I had sent them that same message, there would be a different level of like thought and humanity and how you interact with it. It'll be interesting to see as we're capable of automating more communications and like email generation is a good example. It'll be interesting to see if we're capable of automating more of that stuff long term. What do we actually settle into automating versus like what are the things that need to remain human just for like relationship building and stuff like that. Right. I do think there's a ton of communication that's actually just like approvals and requests and stuff like that that do not need to be human. But there's also stuff that I think will probably never automate because it plays an important role in kind of how we develop relationships. Summarize, plan with pathway, summarize and categorize a client's communication with everyone in the firm. Basically sort through the emails, text messages, calls, practice management messages, and boil it down to the relevant parts for the task or conversation at hand. I like that. One of the things that surprised me about Carbon's AI launch last week, we did a whole show on it last week. If you haven't heard about that. They had a. They shipped a feature that I wasn't expecting that I actually really like, and it was summarizing email communications with the client. So oftentimes, if you have, like a tool that has kind of a CRM sort of component, when you go into that client, it'll show a feed of, like, what's all the. What are all the things that have happened recently with that client? And that could be project status is changing. It could be. Could be anything, just kind of a little concise feed. And in the past, they would have had links out to emails. So, like, if somebody on the team got an email from that client, you would want to know about that and you'd have to click out to that email. But now in that little feed and in the context there, they actually put a summarization of that email so that you can see at a high level. 43 days ago, we got this email from Steve, and it was about his daughter's college tuition. And you wouldn't have to drill completely into that email to get the idea of what that email was about and get kind of a summarized view of it. So I actually think we're pretty close to what they're talking about here. That just works with emails. A really hard thing in all of our practices is when you communicate with clients across a ton of different channels. Like, wrangling all that back into a single place is really difficult. Some people take the position that's why clients should only be allowed to talk with you about stuff in certain channels. Like, everything has to go through the portal, and that solves the problem. I don't think I'm quite there because you still get on Zoom Calls, you still do emails with clients. Even if you use a system for requests, there's still communication that happens outside those requests. So it is like, no matter what, there's like. It's a sliding spectrum where if you will let clients text you and Facebook message you and you have all of these channels, yes, it does get worse. But, but even, even in a more locked down communication strategy, man, it's just hard to get all that stuff back to a single place. But to this person's point here, yeah, the, that's one thing that GPT is actually really helpful for is summarizing the most relevant parts of a message. Um, but. And then ultimately distributing it or parking it in a spot like in the Carbon example, where if anybody is then going to go do something with that client, it's really obvious that that communication just happened. Carbon. Carbon does a. It's one of the better practice management apps for managing communication, obviously because it's so email centric. But having a centralized home for like all of the interactions that you have with a client, that is, man, that's so important. Especially when you're getting to the stage and firm running where you're trying to delegate more of that work and, and create greater transparency in anything that maybe you've done with the client, where you don't want to then be a blocker to your team being able to work with the client because it's something that they just told you, right? So like that transparency gets really important. Sabrina said email read, respond, delete, delegate. We're getting baby steps there. Honestly, in the past few weeks we've gotten some big steps in the right direction in using AI to better manage email. Laurelin also said keeping up with emails first steps that I think are really valuable that we've made just in the last couple of weeks. I like what Carbon's doing with email importance. Without having to open the email, it will assess like without me, the user having to open the email. Carbon will assess the urgency based on the tone and the contents of the email. And if it looks pretty spicy, it'll escalate that for me automatically. I like that. One ongoing issue that I, that I think people struggle with with email is like in our, in our right minds, we know that we shouldn't leave email open all the time, right, because it's distracting you from doing the work that you actually need to do. Worst thing you can do is leave that up on another monitor. But what people do is they leave it open due to the risk of something urgent coming through. So like we're monitoring it because there's an outside chance that something that is urgent could come through. And as a result we lose productivity all the time because of something that might happen. Like just to monitor it, the Way I mitigated this was I would basically email once a day and I had an assistant who, who three or four times a day would be in the inbox and if there was something that was actually urgent, it was then her responsibility to notify me that I had to hop in and tackle that thing. Or at least like, give me the heads up, like, hey, this happened. I know you're not gonna be back in the email until 11 tomorrow. FYI, if you do wanna do anything with this, you just got this message that gave me permission to close my email and not have to worry about the email for all but 40 minutes a day, which was great. I think what I'm seeing now with what Carbon just launched with Carbon AI last week is that could be something that actually your email client does for you. So if you currently keep that email open all day long, just in case something spicy happens, the much better version of that is maybe all you need to see is the stuff in your inbox that's flagged important, and the software will automatically flag the stuff that is important so that you may get 20 messages in a day. Maybe one in every 50 messages you ever get is quote, unquote important. And if all you monitor throughout the day is the important inbox, then there's a much lower likelihood that you're going to get disrupted. So, like, to me, that is a great first step toward taking a lot of stress out of email that Carbon's already working on. I love seeing like Canopy launch their version of this and Carbon launch their version of this. And the mere fact that they've launched this, I think makes all the other PMs that help you manage email also work on kind of their own similar things. And like those people are going to just keep driving innovation across those tools, which is fantastic for us. So I'm really excited, excited about that. In terms of responding to emails right now, I think the most rich AI enabled version of that is Carbon's reply generation, because that's not just composing an email from a prompt, that's composing an email taking into account the whole communication thread. And that's a, a huge step toward AI enabled email. So Carbon folks, if you used to absolutely get in on the AI beta, if you're not you yet, Carbon folks are getting to taste that rainbow already. But the mere fact that Carbon shipped it, I think will light a fire under everybody else to get us something similar. Email was the most common theme throughout these, like, if I could hire a bot to do anything for me, like that was the Most common recurring thing that came up. Kylie Parker, answer all my emails. What is the future? I mean, it's interesting to think, think ahead. There's a lot of applications we use that are built around a thing, a kind of core functionality, where that core functionality is kind of going away. Great example is accounting systems. Accounting systems historically have been built around the classification and reconciliation process. Like that is the meat of the product is you come in, you categorize, you reconcile, and then all the other stuff like reporting and invoicing, all that stuff is kind of built around that. If you caught the embeddings episode that we did, classification for accounting transactions is going to get vastly simplified and kind of go away ultimately, or at least be way less work. So what does an accounting system look like if classification and reconciliation isn't at the core of it? There's some similarities here with email, where email clients are built around like authorship from scratch, but the more intelligent generative email gets when it can see into all your projects and your documents and all that stuff. And Carbon AI is a really early example of this. If the starting point for all of our emails is something that AI suggested, which, I mean, 12 months from now, like, that's probably just the norm. Like, will we be writing any emails from scratch? When you write outbound emails to somebody, even that, like, I think is like, probably happening via a set of prompts. That may be more of a personal preference thing. Would I rather write that new email from scratch or would I rather put four bullet points into a prompt and have AI generate version one of that email? It's not, but it's not. All that is to say, it's not unreasonable to think 12 months from now we may not be composing any emails from a empty box. Like AI. Like, Carbon's already doing the suggested reply stuff. Canape and Carbon are already doing generative new email compositions where you just give it like a couple of bullet points and it gives you a starting point. So what does an email client look like if you're not composing stuff from scratch anymore? Does it start looking a little bit more like an approval system? Like think of a bill pay system where you've got a feed of emails, you're coming through and you have to prove them. What then becomes like the ultimate UI for approving a suggested email or making simple modifications to the email and then approving it and sending it out. That'll be interesting. I think that email client looks a little bit different. It starts becoming more about not just approving the message, but what are all the things that I may need to know about this conversation in the most concise way possible that got me to this message. So it's not just approving it, it's fetching all the relevant details and proving to me that this is the best reply I could make. So if they ask about a past document that could look something like, hey, here's, here's the email that AI has drafted for you. And the things that I think that you need to take into account when sending this email is we went out and we grabbed this detail from a file that was two years old to kind of answer the question. So heads up this, there's this detail, Heads up in a zoom meeting transcript from six months ago, you discussed this and that and it kind of gives you a few lines of like the most relevant context to inform that reply. That entire user interface, it looks very different than today's composing an email from scratch. But it's actually not super sci fi to me to think that we 12 months from now, like won't really be composing emails from scratch. So it'll be interesting to see like how the email client develops and who leans into, who leans into that email sending being like more similar to an approval process than like an empty word document, right? Where you're just creating from scratch. Kathy Williams Review simple personal tax returns. I actually think that could be like in the next 24 months. I think we're probably going to have that. And to go back to what we were talking about before, the Holy Grail is it does the final technical sign off, right? But that's always gonna be the hardest thing. And it's like there are a lot of steps short from your level of expertise. There's a lot of level of, there's a lot of steps short of that that are still helpful. Like it doesn't have to. And everybody goes to AI is not going to do my job. It doesn't have to do your job to still be really helpful. And so like that self review example, like something that will at least like look out for some things and it keeps getting better over time. That's still really helpful, right? Especially for solo folks if you don't have that review framework, I wouldn't be surprised if those simple personal tax return review tools are like maybe something that we even can start dabbling within the next 12 months. But like definitely the next 24 months. John Pitts Track and update daily progress of all projects. This is something that I think we'll absolutely have a much better snapshot of in the next 12 months. And maybe a couple examples of platforms that could do this well are Carbon and ClickUp. So ClickUp's investing big in AI. And if ClickUp is your home for how everything gets done, the work, the tasks in addition to how the work gets done, your documentation, SOPs and all that clickups like built in documentation is really good. Similar to notion, then managing the struggle of what do I do next? Could get easier. We all have a ton of different channels in our life that could, that we draw from to decide what to work on next. That could be email, project management system, the text messaging app on my phone, Slack. Like, there's all these different channels where we get things, you know, a voicemail where each of those different channels kind of have their own set of to dos. And you could go through that channel, you could go through your email, for example, and decide, okay, today this is the top priority to respond to. This is the second priority, this is the third priority. But then you go out to your project management system and you're like, woo, okay, this task is first priority, this task is second priority, this task is third. But then your mom just sent you this text and you're like, I gotta do this at some point. How do I slot this into those email priorities and those task priorities? And like all of the other channels that have their own priorities, I do think AI will ultimately be helpful in rolling up all of that stuff to a single place to at least give you a starting point. Because like, for example, AI meeting transcriptions, like when you record a meeting, it will give you action items to say, like, based on the meeting, it looks like these are the follow ups. So it can see a message and assess, like, is there something that now needs to be done to address this? You know, Carbon's email importance functionality now is an example of its ability to assess the tone and the content of a message to get an idea of how urgent something is. So if I can see across your projects, if you can see across your email, like different channels, I think that will be a step in a helpful direction where we can sit down at the beginning of the day and it's like, hey, here's kind of what I think you need to be working on. Obviously that's going to have blind spots. Is it going to see into the text in my phone? No. Like, is it going to see my voicemails? Probably not. This is something where Microsoft Copilot, I think, is actually probably going to be phenomenal at this because it can see into Microsoft Teams and your Email. And if you manage your whole firm in the Microsoft ecosystem, I suspect that's probably going to be something that it will do really well. The main, like the reason that I wouldn't go to Copilot, even though from a RAW AI standpoint is probably going to be the most powerful platform, the reason I wouldn't go to Copilot is there's a lower ceiling of what Microsoft Copilot can ultimately do for me because it doesn't solve accountant specific problems. So I do think accountant specific software platforms will, will inevitably be behind like Microsoft Copilot in terms of like RAW capability, at least in the short term. But, but longer term, what does the next version of that really helpful platform look like to me when it's like my clients are connected to my QuickBooks files, for example. So like take Client Hub, it's a practice management system. One of the things that it does is it connects with the QBO file of each of those clients. And in that client portal where you can exchange files and make requests and all that, it'll also manage uncategorized transactions. So you can point it to specific accounts in the quickbooks file and then the client has a single portal where you can exchange files and all that. But they can also address any transactions that you have questions about. So like thinking beyond what Microsoft Copilot will ever do, some really specific to us applications that will be helpful are things like extending that to the accounting ledger and other kind of accountant specific applications. That's why I wouldn't go to Copilot right now. But like that, that bigger picture, knowing what to work on next. Oh, when are we going to have a good version of that? I don't know. But I do feel like, like Carbon's email importance is an example of a step in a helpful direction there where like I think it did just get a little bit better for those users and we'll see what the other platforms roll out and, and we'll also see like what are the ways that we can get better about managing all those channels of communication in a more centralized place. Like for example, meetings, like Zoom meetings are a huge black box of context right now because most of us don't record them. And it hasn't quite been normalized enough yet for that to maybe not be an uncomfortable thing with clients. Recording meetings is one of those things where if you explain it to a client in the beginning when they become a new client and that's just the norm. Like kind of like when you call a customer service thing and they're like, this call may be recorded for quality assurance purposes. If you just set that meeting recording expectation from day one when those clients come in and you're like, here's the deal. There's gonna be this other, you know, fathom or whatever it is, things sitting in the call that records transcripts of our conversations so that my team can then see into the stuff that we discuss. And you won't get Tina asking you the same question that we discussed. Right. Like that's just better documentation for our team so that we're not wasting your time with sort of repeat questions. And they're probably like, cool, yeah, that's fine. Like, the people who have issues with you recording their meetings are probably the ones that are like trying to get you to do things you don't want to do anyways. But meeting transcripts are a great example of one of those channels of communications that we have right now that our systems generally can't see into. So going back to the example of what do I work on today? If something came up on a zoom call with a client yesterday and that's like, crap, I got to get that done tomorrow. Or it's this urgent thing that came up. If that transcript doesn't live anywhere, like if that, that documentation of the meeting and what was discussed doesn't live anywhere, doesn't matter how good the system is, it's never going to be able to see into that and then be helpful to you when it comes to deciding what to work on next. Right. So kind of like capturing those, those sources of TO dos that right now aren't documented in any of your systems. Um, that's something we could probably do a little better on too. We didn't get through like two thirds of these. I'll put a link to Nio's tweet in the. In the show notes. A lot of good stuff was discussed here. Curious to hear, like, what are the things you're most excited about? Like, what's that next? That next development that you think would be a huge time saver for you? Maybe it's just the thing that stresses you out most about your day to day. Like, maybe that's the place to start. What is the most energy sucking thing that you have to deal with day to day? I'd love to hear what that looks like for you. Last Friday, I teased a super, super exciting AI announcement that I thought was just around the corner and I might be able to talk about. I think I mentioned that Thursday and said I might be able to talk about it Friday. It's been kind of a slow burn. I may do an episode on that in the next couple few days. I know that was kind of a tease and then I didn't deliver on it. But I think we'll be able to get to it this week. It's like a pretty big deal AI launch that's just around the corner and kind of a big indicator of what's to come. Like bigger, even more useful applications, like beyond what we've talked about with canopy and carbon and all that stuff. So that's exciting. That's coming soon. Hey, thanks for coming and hanging today. It's Monday. This was like a low lift Monday show. Like, I didn't want to make it too sweaty or too technical or that sort of thing. But thanks for coming and hanging and I'll see you tomorrow. Bye.