020 Put THIS AI To Work Today

I have got that frisky Monday feeling today. Apologies in advance, but there's a lot of stuff happening, a lot of fantastic stuff. So today we're gonna talk about the what is my favorite new AI use case that you can put to use today? I shared it in my newsletter last Thursday. What I shared, though, is like the tip of the iceberg. There's a lot more we can do around that that I think's really powerful. And that stuff's all ready to go. Today. We just got the biggest announcement AI related yet from inside the accounting space. And last Thursday I was on AICPA's town hall live show. I'll share what that experience was like. Let's do it in a spicy installment of Jason Daly. So a bit of context from what I shared last week and why I think it's actually just the tip of the iceberg. So what I shared was basically how to build a custom ChatGPT experience. So a ChatGPT experience that is limited in scope, it's not going to pull random stuff from the web, and it's limited just to the context that you provide it. And. And a while back, one of the most interesting use cases I heard for ChatGPT was as a resource for junior accountants when they are first getting started. There's a lot of those questions that they're afraid to ask, questions that maybe Google can't quite answer for them. And I love that idea initially because I think we all know the feeling of that, kind of like flailing of being new and getting stuck on things, not knowing how long to spin your wheels. I love the idea of ChatGPT being a helper for those folks, but obviously it lacks any sort of specificity and context because just ChatGPT, it's pulling from the entire web. And so the use case I talked about with this bot was a chatgpt like experience that's built just on your own documentation, whatever you want to throw at it. So the first example I shared was a chatbot for your team to basically pick your brain personally, like a chatbot for a specific manager, where it's up to you what documentation you want that chatbot to pull from. It's not a replacement for talking with your team, but imagine for you being able to build up that bot as an asset that gets smarter and knows more over time, and it can serve as that first line of defense to be helpful for your team in a way that's not going to disrupt you, because they can work with it asynchronously and you can build on it over time. Now, this is a thing that is absolutely doable. Today it took me less than five minutes to build a little bit of nuts and bolts about how it works. It's called myaskai. I'll put a link to it in the show notes. Basically, all you do, you create an account and you chuck whatever source material you want at it. That could be a website, could be Google Docs, could be plain text, could be PDFs that you upload to it. But think about things. Let's just start through the lens of, like, that bot that will be helpful for the manager. Think about the sources of context that new people could learn from. That could be transcripts from training meetings, it could be internal company docs, could be Q&As that you've done with the team Internet. Any context you can provide will know. And it can then regurgitate and synthesize when people ask it questions. So where your mind immediately goes is, ooh, what context do I have? All of a sudden, context just got a lot more valuable, right? So one of the bigger picture AI things I've been thinking about for quite a while is it seemed like, well down the line, like, what is the moat anybody's going to have in AI? Ultimately, if you can slap a chat experience on top of anything, then what is proprietary anymore? And what I keep coming back to is context, personal context, the stuff that you have made, and professional context. This Daily show as an example, I basically started it for three reasons. One, to go deeper with the folks who this stuff is most resonating with. I think there's an entirely different level of trust you build with people that you have a daily habit with, rather than, you know, weekly or something else. And when you create that space, it's not just you, it's that whole community that's coming together in the comments. And just the value of that, that was reason number one. Reason number two was for me to exercise the absolute heck out of the muscle of speaking from a prompt. And like a logical, progressive way that doesn't go down a whole bunch of tangents. And like, that is a learned skill, but it's a skill I'm developing right now at a. At a fantastic grade that will always serve me right. But then the third reason for this show, for me, it's a context play. You think about who has context around running firms and mindfulness and now AI and stuff like this. This show itself is. Is generating a ton of context. If you think about the transcripts we have from Every single show. And people watch my content and like they'll go back through the back catalogs and stuff like that. But in the same way that, you know, when you run a training for your team, it's most relevant to the people who are there on the day, and then it kind of goes away over time. That's kind of how producing content works too. But what if instead of those things kind of being ephemeral, they were assets? Assets that weren't just a recorded meeting. Because let's say you record that 60 minute training and you're like, oh no, we've got to record a training. You can go back and watch all that, that's okay. But the way better version of that is the transcript that's searchable, that anybody can go back and look for that. When somebody searches across everything, that little snippet from that training meeting comes up. That's the power these days of documented context. And when we went through, like what your practice management system looks like on AI, on AI, much of that was simply doing more meaningful things with the context that you have in your firm, from organizational email to the file system and all that. But through the lens of this chatbot, this assistant for you, kind of as a manager, this is now an asset. And so it makes you start thinking about what is the context we have as a firm and what is the context that I have personally and how can I start investing in that context? You think about all the stuff that you have right now. We don't generally document generic stuff in a very helpful way. The vast majority of that stuff that we're doing, we're doing it through the lens of a specific client or a specific situation. And so oftentimes it's not documented in a helpful general way. But building your own personal context can be as simple as like recording a voice note for five minutes on some thoughts you have on a thing and then you've got a transcript. And that is now part of your personal context. That's now something that's queryable by your team or by your clients, if it should be. I think where we're headed is AIs. Ultimately, AIs helpfulness is dictated by the value of the underlying context of it. ChatGPT is great, right? But ChatGPT can't see my own tax return. ChatGPT can't see my accounting file. AI's helpfulness will ultimately be gated by the underlying context. So the currency of the future becomes context. And it's interesting to think about where you draw the lines between different types of context. You know, all of the information in the world that is all context. All of the information in your firm that is organizational context, and that's email, it's files, it's tasks, it's team chat. Smaller than that. Your clients, your clients have their own context. You as a person, you have your own context. Business and professional. The stuff you want members of your team to be able to see versus like a personal journal, that's gonna be helpful to you. Context is helpful, but also how we divide up those different sources of context will be interesting. So I shared the example of something a manager could use as sort of a gatekeeper. And that's just one really small example that I think is helpful. I was talking with somebody else the other day. They're like putting together internal training for their team to do advisory. And if you think about what that normally looks like, like, usually you're going through some sort of journey, some sort of learning process over some duration of time. And it's a great investment to make, but it's ultimately, ultimately an investment in the people whose butts are in the seats at that time who can turn in live, who are engaged in that process when it happens. And when half of those people are gone in five years, or you hire a new person in a couple of months, oftentimes you lose that. And you can try to document it and say like, oh, we're gonna record all this stuff and they're gonna go back and watch it. But it's not quite the same as going through that experience live. But that's how we've traditionally approached training. And it's still probably the highest fidelity version of training, like being able to talk live about it and all that. But what does an AI enabled version of that look like? Is it a advisory bot? Something that's trained on all of the context of all the trainings that you've ever done? Something that will help you suss out maybe things to be aware of in financial statements and think through how you communicate things to clients. That context that sits behind that training bot, that's an asset that you can continue to build over time, that's more abstract, but even very nuts and bolts things like where do I input this thing in my tax software? Context can be big things, it can be bite sized things, it can be a single sentence. This can be a great way to build a more collective intelligence within your firm that anybody can leverage. Especially when you can interface with this stuff anywhere in Microsoft Teams and Slack and all the places that you Already live, not going out to a new thing, you're talking with it in the spaces where you already live. If you've ever had a staff person leave who you've done a bunch of training with, who was a subject matter on a specific thing, you know, just the pain of losing that person and seeing all of that wisdom and expertise go out the door. And we say things like, oh, we gotta document all this stuff, and the output ends up being like a word document or maybe like a page in Notion or a page in your ClickUp or something like that. But those things generally aren't discoverable as part of like the bigger set of all of the knowledge that your firm has. I think through this paradigm that expertise becomes context that is then built into chat experiences that the rest of your team has access to. If you're a product company, you think of somebody like LivePlan who has like a advisory product, and they help accountants be trained on how to do advisory in a more scalable way with their clients. If you've got a ton of materials already, if you've got all that context, what is productized context look like? What does it look like to license access to your context? Because ultimately, if chat experiences are easy to create and like this thing, you know by example, less than five minutes and you've got it set up, then the limiting factor is what is the context? It's the context that's the true value. If context is the currency of the future, how do you license context? How can I make my chat experience smarter? I went, I talked about this on a, I don't know, some AI talk I did recently, but there was somebody else on the panel who runs a podcast for accountants, and they've interviewed a ton of super smart people. And if I love that podcast, how can I license that context to use within my personal chatbot or something like that, right? You've got all these transcripts of these incredible super smart people. And if I'm building sort of this personal assistant for myself, what would it look like to license all that context, to pull it into the AI stuff that I use? We talked recently about how do we approach the accounting profession in a more open source way? Open source context, like, what can we build that then can be taken and pulled into anybody's chat experience, right? Whether that's technical guidance or the point of view of a specific expert that you look up to, who are those people online that you really look up to? Especially those that are well documented. Maybe they've got a blog already. What would it look like to license that context, to be able to pull it in to the assistance that you use each day to make you more productive. I'm convinced context is the future when you can put a chat experience or a voice experience on top of anything. I guess we're just talking chat voice is really compelling too. It's pretty much the same thing as chat. But like, I get really excited about voice. It's becoming more important to just be a well documented human being because that context is ultimately like what is specific to us and what is different from everyone else. And this is. Here's the kicker. Here's where it gets really exciting. We're talking about doing this through the lens of like, maybe the people who work for you or your peers within your firm. And the more you document what you know and your expertise and all that, the more it'll be a resource for them. But that same context, that same information from here's how to do this or that from a nuts and bolts standpoint. Here's how to enter this thing into tax software. Here's how to draft this email to a client. That same context that will be an asset to your team or community is just as big an asset to AI, to automated agents who will do that work for you. So you can teach humans with it. You can also teach AI how to do things up to your level of expectation as well. So let's talk about AI agents. You may have seen Auto GPT. It was like the number one trending thing on Twitter for like a week last week. And it's basically a spin on GPT where it's this agent that will browse the web and you give it a task and a set of goals to complete. And it will work on that until the end of time. It will go out and try to figure that thing out or do that thing that you've asked it to do by basically prompting itself and saying, okay, I've been given this task. What are the things that I should go out and do? And then it actually goes out and does it. And if it gets stuck, it comes back to you. If it needs to take an action, you can have it ask you, should I take this next action or not? These are AI agents that can go out and do their own stuff. They can see the browser, they can go out and actually execute commands and do things. And in their current state, they're pretty bad. They're pretty broken and fiddly. But it's a peek into like, what the future of like, fully autonomous stuff will look like. And with them it's going to come down to context too. It's gonna come down to the specificity of how you do your work when that thing can see your desktop and the programs that it uses. How would you train a junior person to do stuff with all the context that we're talking about building right now? So if you build that context for your team today, it's also gonna be an asset for AI ultimately to do work to your expectations. This all sounds like pretty sci fi stuff, and I get that, but we're actually pretty darn close. If you go out and look at baby AGI at Auto GPT, these agents are already out there. It's probably the fastest thing in AI right now is agents. But if you've got a project management system with a bunch of tasks, I think the future looks like rather than assigning that to Tina, you're assigning that to Tina GPT or Tina Bot. And it's referencing maybe organizational context, maybe Tina specific context about how she does her job so that when you reassign that task to the agent, it actually goes out and does the work. Now on this subject of context, uh, I thought this stuff was actually long ways out. If I told somebody, oh no, this Daily show is a context play. On the day that I started the day the Daily Show, I think everybody would have thought I was nuts. And I still wasn't like a hundred percent, but like that seemed, that made sense to me, right? When information becomes like infinitely accessible, which ChatGPT has kind of already done, then your value as a human being comes down to what is the specificity of information that you have or the proprietary information of like your clients that's not general information, like, that's what's left. And so that becomes really valuable. Back at QuickBooks Connect last November, that conference kind of came at the perfect time because it was in peak ChatGPT excitement time. A week before ChatGPT was released, GPT 3.5 came out, which was a huge upgrade over previous iterations of GPT. I gave a talk last summer in NIO's taking your firm virtual conference where I gave a whole live demo of Jarvis, which is a tool built on top of GPT3 that generates like copy and drafts writing for you and all that. And that was like old now archaic versions of GPT that were frankly already helpful. But my talk was basically like, is this gonna put copywriters out of work? Probably not, but it's a great assistant for copywriters. And then we kind of Frame that through the lens of accountants. Like, is AI going to put us out of work? But no, it will be a phenomenal assistant for accountants. And so this GPT thing had kind of been building, but then QuickBooks Connect happened, like, right when ChatGPT hit a million users. Like, there were just so many exciting things happening at that time. And I was there hanging with a bunch of smart people. You know, all the developers are there, and there were people that were plugged into this that were tasting the rainbow, and there were people that weren't. But, like, my memory of that conference will always just be like, me and Chad Davis being like. Like this. Like, what does software look like after all of this? My mind is blown. And I remember we went to dinner one night, it was me and Chad, Jake Kasenich and Ryan Lazanis, and I was talking to Ryan, and I told Ryan we could go out tomorrow and train an AI model on how Ryan would respond to sticky client situations. Ryan is phenomenal at mindfully articulating sticky things way better than I am. Like, he. Like, he is so good at that stuff, especially through, like, the sales and scope lens. Like, he is so good at that. And so we're talking about, like, what does AI look like in the future? And it seemed like a logical extension of, you know, the work he's already doing is to train a model that people could use. Like, how would you feel about that if people could. Could, like, leverage that resource library to be able to communicate with clients in a more mindful way, or for it to point out potential flaws in that reply to say, like, hey, consider this, or that. It's kind of like the AI art stuff where there's an element of that maybe being dehumanizing. Is it dehumanizing or is it like a higher, higher leverage use of our expertise? And so we're having those conversations back in November around, like, specific training data and, like, where would you draw that context from? And for people who do things like what Ryan does, and honestly, what I do, I think the future is going to be driven so much by context and access to context and whose context you can lean into. And for you, what's the right mix of different types of context that will help you to do your job better? I've toyed with, you know, I. That bot I did last week I called kind of tongue in cheek, Jason GPT. But ultimately that's going to be a thing at some point. Like the notion that we have all of these transcripts, we have all of my YouTube video history. We have these daily shows. There will be a day when there is a chatbot that is trained on all of that context. And where we're at right now, to me is interesting because today there's part of me that thinks like, oh, that's like a helpful sawdust, like byproduct of a thing that I'm already doing. So just go set that thing up tomorrow and then if people want to use it, they can use it. So part of me thinks that's sawdust. Part of me thinks that is like all we are long term. Part of me thinks that, like, that is our, the value in our expertise is that specific context. So if I put that thing out there for free tomorrow, like, then, are you giving this away? And this applies to you too, through the lens of things you put in front of your clients. If you put that thing out tomorrow, are you giving away the very thing that is kind of like the last frontier of our expertise, the stuff that AI will not take over? So I'm stuck between, is that a thing you just give away because it's a byproduct? Is it like a cheap little utility you roll out and you see people want it? Or is that something you charge hundreds of dollars a month for, maybe thousands of dollars a month for? Because it is your essence. Like, it is you, it is all of your experience funneled into this context. That's interesting, right? What, like, what end of the spectrum is that on? Is it the essence of you that you should maybe never even give away or you can't even put a price to? Or is it like, no, it's a byproduct of this stuff that I'm already doing. I think right now, if you have line up a hundred people on the street, 80 of them haven't heard of ChatGPT, and 95 of them would tell you, like, no, that's sawdust. Like, just give it away. Like, sure, maybe that's helpful to somebody. But five years from now, 10 years from now, when you've got AI agents that are doing really helpful stuff for us, I think that actually might be like the really valuable thing that we have to hang on to somehow. Like I said, this all seemed super sci fi about a month ago, but the fact that anybody can build a bot today on any context, plug it into Slack, plug it into teams, do it in less than five minutes. I think today context just became extremely valuable. Interesting couple. Big AI updates. Canopy just launched. Generative email replies. I call I said last November, this is like the biggest Opportunity Slam dunk. Generative email replies to make accountants life easier. Let's do it. We're six months down the road. I wish it happened earlier but also credit to Canada because they made it happen faster than any of the other practice management systems. Big thing that this does which I really tried to point out in that what does your AI look like on practice management systems Post this is a big shot against the PMs that don't look lean into email. We spend so much time emailing, messaging, whatever it is, however we communicate with our clients, if that experience is disassociated from all of the context in your firm, I mean I think you're on borrowed time. So I have a hard time seeing any practice management system now going forward that does not include like an email experience. And I don't know if that's anything anybody could predicted five years ago but generative email replies are just going to be that helpful to where that I think that's kind of a non negotiable aspect of a practice management system these days. Uh, ClickUp just launched their AI assistant as well. I'll be honest, like I'm so so like I've always loved ClickUp but it basically looks like what notion had like two months ago uh, and I'm sure it will get better and do more meaningful things with all the information that's in there but and I have been a ClickUp fanboy like ever since their launch. Like I put my entire cast practice on ClickUp early days still won't do email. So if I'm building up that asset of all this documentation, all this context, all these things happening inside of ClickUp if it ultimately won't inform how I communicate with my clients, then is it the right place to have all that information? I know if you're a ClickUp user that's like a big old punch in the gut context man. It like so much of this comes down to context. What context is proprietary and what systems can and can't see that context. That's what ultimately is going to drive the usefulness of AI straight down to here's how to do this task and then you assign it to an AI agent. And because you had that context and now knows how to do it in a more intelligent way, there's going to be a firm that has none of that context that will then try an AI agent and say this stuff sucks, it doesn't work right. But that firm that has that context very explicitly outlines expectations for how things should be done. They flip the Switch on that exact same AI agent that can draw from that context, it's 10 times more helpful. I told you this was going to be a sweaty one. Okay. Last Thursday I was on AICPA's town hall. So AICPA does this weekly town hall live stream. Like it's like a one hour produced show. And they started doing it when Covid hit in April of 2020. And it was like a way of getting timely information out to everybody when they like super needed it. They're still doing those 11 hour shows and I came on to talk AI. It was fun experience. Like some behind the scenes they do like they kind of go through like a pre planning process for each of the shows. They have an alignment meeting before they do the show. The CEO of CPA.com does the like he is like the main host of the show. They shoot it at, they've got some studios in New York City at their headquarters there, which I've seen, I got to see when I did the startup accelerator thing with the AICPA super like legit studio production setup. But then they pipe, pipe in various people virtually to like have several segments throughout the show. And so we ended the show with it was about 12 minutes on AI and and a lot of it was chat GPT because that's I think most people's first experience with AI. And it was good, it was exciting. Like I, I think people got excited about it. Bunch of questions, but also questions like what is chat GPT? And it's a great reminder of like what an echo chamber, not only like this show lives in, but that we all live in. It's one of those things I have to always keep telling myself is our brains treat the world as being much, much smaller than it actually is. And what that looks like in practice is like you being afraid that someone has already heard the thing that you've said. Or are you being afraid of repeating yourself? Because in our heads everybody hears all the things that we always say when in reality 99% of the time nobody's watching. Even if you send out a tweet, the vast majority of the folks that you know, it won't even show up in their algorithm. You know, me and the stuff that I produce for an example, like by example, a lot of the folks that are here have been here for a long time, like back to the beginning or something near that. But about half the people who are here right now have found me in the last 90 days. And so their perception of me and the things that they've heard from me and what I do. And they may not even know that I ran a firm for years and years. Those people don't know any of that other old stuff. Right. And I think this is especially relevant through the lens of, like, just doing a little bit of content creation, which, by the way, great way to build your personal context. We worry about coming up with, like, new and novel things to say, or we worry about repeating ourselves when the reality is nobody's watching like that. Like, our brains just can't comprehend how many people are out there, how many new people you stumble into every day and the level of specificity that is out there. We think, like, I don't know if I niche down in this way or that way. There's not going to be enough people out there for me. Well, there's. If there's not enough people that you found yet, it's because you're not putting yourself out there to find those people. There's so much specificity out there. There are people for whom I'm just. I'm a broken record, but maybe not, because I just said people don't hear what you say. There's so many people out there for whom the thing you do is 10 times as painful as the people who are on your client list today who will pay five times as much and love you for it. And the only way you find those people is by going out producing content, being out there talking about what you do. So that aspect of, like, our brain not being able to understand how big the world is, I think the main takeaway there for me is, is find the thing you believe in, find the thing you're passionate about, and just keep hammering on it. You can hammer on that until the end of time, and every day you're gonna find more people that. That resonates with, and that personal network is gonna get even more powerful. Is it ironic that I say this on a day where I'm like, so far in the weeds on AI stuff that probably only makes sense if you've been here from the beginning of this show. Maybe, but that's what's nice about running the show yourself, is you don't have to follow your own rules. Okay. Hey, thanks for coming and hanging. I would love to know your thoughts on the whole context thing. I think it's a really interesting discussion, especially around, like, what are some applications of context that we may have today that we could build helpful bots for right away, or ideas for bots where maybe you don't have that context yet, but it's worth then investing in creating that context to bring that bot to life, either yourself or creating that context through like the lens of a community. Would love to know your thoughts on all that. Obviously I'm very excited about it. Thanks for coming and hanging and I'll see you tomorrow.