Loading summary
Transcript63 lines
- [00:03]
Tyson Mutrix
This is Maximum Lawyer with your host, Tyson Mutrix.
- [00:12]
Dan Schnerbush
Dan, I want to jump right in and let's rewind to a time, to the time when you realize that your current tools really weren't cutting it from a case management standpoint, from a matters management standpoint, from an automation standpoint. Go back to that time and you're like, okay, this just isn't cutting it for me.
- [00:34]
Tyson Mutrix
Yeah, yeah. So that point has, has occurred many times in my over the last several years, at least twice. So we started originally in. I really liked Llamatics and Locus early on. I was an early adopter of Locus and they've grown a lot since we use them. Very good, like introductory all in one kind of case management tool and the automation's there. It does a lot of cool stuff. But I was just kind of wanting to do some unnecessarily ridiculous things and at the time I was learning Zapier and figuring out how to do these kind of inter platform automated sequences. And I fell in with Lawyer on the beach and Regina Edwards and was enamored with Monday.com and I was like, look at how you can see everything in this nice like one display of all your cases and what the status is. And nothing I was using could really do anything like that. So we messed around with Monday.com for a couple of years and just never got it into a position that made a lot of sense. Like you could see kind of everything that was going on. But there was no data model. Like it was just here's a table of your cases and little status updates. But there was no data model. So there was no contacts connected to the matters. There was no tasks or anything like that connected to the matters. So it was just very decentralized.
- [01:56]
Dan Schnerbush
Yeah, that's wild. To me that is wild.
- [02:01]
Tyson Mutrix
You can set it up that way with mirrored properties and LinkedIn stuff. But I didn't realize that until we were already like 3ft out the door. And it's kind of not a super well developed function of the system. And at one point I found this airtable platform. I was like, what the heck is this? Why do I keep hearing this name come up? Why is everybody talking about it at least where I was looking online, different YouTube channels and stuff. They launched Airtable, launched a. You know, for the longest time I thought, oh, it's just this back end database builder. I don't really care about that. It seems like a kludgier version of Monday.com and then they came out with this interface system for building your own interface on your database backend. And I started over time gradually understanding like the importance of relational databases and data model and how you tie everything together. Airtable offered a lot more flexibility. So if you'd like think about like Lawmatics and Lockus and Clio in my case and all that, what they've done is they've taken a database, a relational database and they've put a pretty face on it, right? So they've abstracted away all that complication. They build the front end, the pretty face for you, right? They build all the buttons for you, they build all the features for you. But with a tool like airtable you're taking it one layer down so you're building both the database and the front end the way of looking at it and I realized once we kind of switched over to that and I started building some of my first automations and I was like oh my God, this is what I did, this is what I've needed. So that's kind of when we realized that, you know, that's kind of when I realized I wanted more flexibility, more capacity to move data around, do what I wanted with it without being constrained by the limitations of oh, did they launch this new feature? I think Clio just came out with automations like a couple months ago and it's like a, what it is like 15 year old company or something like that. Not to throw shade, there's a fine platform but anyway that's kind of where I, where I realized it and the kind of transition that took place there.
- [03:53]
Dan Schnerbush
I had made a post in our, in the, in the Guild a couple weeks ago and I think you've got, you and I think have opposite perspectives on this. I think, and actually I don't think we have opposite perspectives. I, I, I think that you and I have the same perspective but we, I think my expectations are maybe a little bit higher of software companies where it was, the one where I was, was talking about how why isn't there like an all in one platform? It doesn't make any sense to me why there isn't one at and, and I understand it does come down to cost and all that but it is, it is, it's, it's one of the things, one of those things where whenever I first became a lawyer like I, it just never made sense to me why there wasn't some sort of all in one solution. And by, I would make sure people understand what I mean by that. I'm talking about like everything from like bookkeeping, automation. So the CRM Part of it, the case management part of it I'm talking about because I think it, the fact that you've got to go through there's. So the fact that we have a phrase called tech stack, to me is frustrating. You know, the fact that you have a tech stack where we have all these softwares talking to each other, to me is ridiculous. But that's the world we're in, so. But I want you to hear your thoughts on it.
- [05:04]
Tyson Mutrix
Well, there's two things to it. I mean, one is like, like I mentioned, I think it's kind of inevitable. Like software development is really expensive. Like, and the difficulty is like whenever you're trying to build an all in one, like any Joe Schmo that builds a single purpose tool, like you could have a crack team of three developers that writes a, a tool that does that one thing way better than your omnipurpose tool, you know what I mean? And they're gonna peel away customers like if they attract. And as long as the APIs are open and you can automate between the tools pretty easily, like a certain degree of the, you know, the consumer market is just gonna be like, well, why don't I just use this better single purpose tool? I know how to use Zapier, I know how to use make. Like, I know how to combine these tools, so why don't I just use the single purpose? So I think some of it's just the economics that it's hard to wield a behemoth like that. Some companies can't. Like Zoho. They work very hard at creating this gigantic ecosystem, but they also benefit from armies of, of developers at non American wage rates.
- [06:09]
Dan Schnerbush
Yeah, I love, I love Zoho. You know, you know this. I use it, but I do. Part of me is like there's something they do I don't like and that's, they kind of just like rip off what other software are doing. Like we use Bigin for elites. It's a ripoff of pipedrive is what it is. It's, it's a, it's pretty much a knockoff. And the, the trade off is, is that the beginning version of whatever it is is not super great. And then it becomes great eventually. So it becomes, becomes the same thing or better. It's just like the trade off is, is that, you know that whatever they, they, they release really early, they come out, they create like a version and then they, it's not even a beta version, they release it and you're kind of, they, whatever the new thing is. It's, it's not great and then it get. That's something I don't like. But you know the CEO of Zo or the former CEO, I don't think he's a CEO anymore. But his whole view on it, he was like the founder where he, he, he wanted to make sure everyone had access to it. That's why it's so cheap. Like that's, that's. He wanted to make sure that everyone has access to it. And I think that that's, I like that. As opposed to some of the legal softwares that charge us an arm and leg because we're lawyers.
- [07:18]
Tyson Mutrix
Yeah, I think it's mostly the legal, the legal tax is just, it's coming from the fact that like when you're developing for a niche that's really small, like think about how many people use Zoho. Like companies is probably in the tens of millions. There aren't even, there's only a hand. What is it like a few million lawyers in the entire western world, like English speaking West. Like assuming software dev and English and all that. And even globally there's probably Only less than 10 million if I had to bet individuals and most of whom are working at corporation or at least so like your total addressable market or your TAM is pretty small in law firm tech dev. But you've got to compete with the airtables, the Monday.coms, the Zoho, the massive multi industry vertical players with infinite software dev budgets who can service tens or hundreds of millions of customers. It's hard to compete. It's hard to compete. And so they charge higher fees to offset the fact that they have to build a software that's at least as good as, or at least comparable or at least not as so bad that you don't want to use it.
- [08:20]
Dan Schnerbush
Don't you think that the time, their time is running out because of, because of AI and what you can build now? Because you were talking about how you'd have like your team of three people that could build out this great niche product or whatever. But the fact is that you no longer need that team of three. You can do this much faster and at scale for much cheaper.
- [08:40]
Tyson Mutrix
Yeah, it's wild. Like there's this whole like category of coding now called vibe coding which is basically just where devs sit with a microphone and build a custom agent on, on premises through like Claude or something like that. And then they just sit there and curse or whatever and just kind of mumble into the microphone. I need you to edit this and Then like, it just, the agent just kind of switches stuff around. And you know, if you're truly a novice, you don't know anything about coding. Like, especially if you're trying to launch like an enterprise level software, you're probably gonna have like security problems. And you gotta be careful what you're doing. You still need to know what you're doing. But like somebody who does know what they're doing equipped with these tools, like there used to be this concept of a 10x engineer, or an engineer who's just like 10 times more efficient than other people. They could just spit out code and write stuff. And that's just. Some people are just that way. Like Those people with AI are now like 100x engineers or more. So. Absolutely. But what's even crazier, I follow this guy, Nate Jones on TikTok. Highly recommend people who are interested in AI to follow him. He's fantastic. But he was talking about how one of the things he predicts, and I think this may be true in the next few years, is just kind of a deterioration of the interface layer of software development in general. So like we're in this world, we've got the Clios and the myspaces and all, or my cases and all the stuff, and we interact with these interfaces, we log in and we look at our case system and whatnot. And the prediction is that we're going to see kind of a decline in this whole interface layer and we're going to be relying predominantly on agents to do a lot of the interfacing. And what's going to matter over the next five to eight years is going to be who's got the cleanest, easiest API that is equipped with a model context protocol server so that LLMs can interact with it easily, that has good data, like, who's got the best, most accessible data for agents. And like I'm, I'm sitting here thinking like these companies, like, you know, Westlaw and LexisNexis and stuff, they've got these huge repository data, their APIs are just crap. So like if somebody comes out there and builds this beautiful case management database, if anybody wants to help me build this. But like, could you, could you, could you do this? But that's kind of my prediction. And one of the things he was talking about too is that he was spinning up an app. Like, he's like, he had an agent, he's like, hey, I need you to build me an app for my day today. I just want to know what's going on. What are the Top emails, like, what's. What's on my to do list. Like, what are the action items I should prioritize? And in like two minutes, the agent had spun up a custom app for his day, never be used again. And then the next day he did again. He's like, now I just have an agent spin up a new app for my day every day that tells me everything I need to do and has all these buttons to do. All the action items that I need to do is like, this is going to be what's happening in the next five to eight years.
- [11:20]
Dan Schnerbush
Yeah, it's. It's a. To me, it's like a massive arms race and they're all, they're all fighting. And I'll, I'll get into Thompson Reuters in a second. And what they did with Copilot and Case Text pisses me off so much. But, yeah, I'm with you. If we can find an army of people to kind of like build out that database and then defeat them, I'd join it in a heartbeat. But an example what you're talking about, though, is, for example, I'd done that training in the Guild on Tuesday. And one of the. A simple, A simple agent that I'd built that was like this workflow was you just send a chat and you say. I could say as simple as send an email to Dan Schnerbush that says that I'm running late for this, the meeting, and I'll be there in about 10 minutes. Like, let's say that that's an example. I've never opened my email inbox. I never, I never get in anything else. I just put a chat to my agent and the agent does the work. I think you're right about that. And the whole. Which is interesting because think about how much time has been spent on design, like the user interface and how. How that's going to basically be irrelevant in the next decade.
- [12:24]
Tyson Mutrix
Yeah, I think we're going to start to see a lot of stuff just get consolidated into our chat bots where, like, I was doing some kind of reporting on this in the, in the. Our Facebook group the other day about how open AI is launching. There's been some leaked screenshots that they're launching an integration system where any. Any app out there that has a model context protocol server and for everybody who's not familiar with an MCP server is basically just a Rosetta stone tool for LLMs to talk to software without needing to program all the particular API endpoints. It's like a book. It can Read as like a menu of here's all the things I can do in the software. And the LLM just knows how to do it right? So basically they're building this integration system where any app that has an MCP server set up, you can just click Connect just like you can in zapier or make to connect a new software. You just click connect and it will add that into ChatGPT library of available capabilities, which I think is going to be just incredibly game changing and making agents trivial. Like you're not even going to build an agent, you're just going to click integrate Connect, Connect my case account or whatever that I think is going to get real wild when we can do that. But are those other interfaces even going to be important? Like if I can just set an automation that spins up a new dashboard inside a chat GPT in like an artifact window like you can in Claude, or you can actually have a working app in a window in Claude that you're working on. Like why not just do that? Like why do I need to log into all these tools?
- [13:53]
Dan Schnerbush
Do you. What do you think this says about the future of law? I was in, in Utah a month ago at Honors Disrupt conference and I mean everything was the most of the talks were about AI, right? And I think that was understandable, but I don't think lawyers are ready and I don't think that they know what's coming. I think that we have like a three to five year window where things are going to remain somewhat normal and then we're going to see a substantial shift, right? That, and there's going to be like the haves and haves nots. But I wonder if you, what's your thoughts on that? If you agree with that, if you disagree with that, what are your thoughts on the future? Like what is, what is AI hold for the legal space?
- [14:35]
Tyson Mutrix
So I'm a little. My belief about AI in the legal space I think is probably reflective of my belief about the effect of AI more broadly, which is personally I am not worried about the kind of doomsday theory that a lot of people have is that we're going to have this society where we have people who run the AIs and make all the money and then we have all these people that don't. And I'm skeptical of that interpretation because we've had lots of technologies that had the possibility of that sort of thing happening. Like, I mean think about like just the emergence of the Internet and all the different cloud companies or Internet companies that exploded. You could make an argument, oh, well, all these Internet companies are going to be making bank and all that. But one of the advantages and interesting byproducts of AI development is that. What's her name? Meeker.
- [15:24]
Dan Schnerbush
Mary Meeker.
- [15:25]
Tyson Mutrix
Mary Meeker launched her.
- [15:27]
Dan Schnerbush
I just did an episode on Tuesday about it.
- [15:29]
Tyson Mutrix
Yeah, you just launched her AI report. And one of the things that she talked about is that fact that the cost of intelligence, AI intelligence has fallen by over 99% just in the last.
- [15:38]
Dan Schnerbush
Few years, way faster than Internet, I think. It took 27 years for the Internet cost to go down. It took for. If you compare it to electricity, it took 72 years for the comparable drop in price whenever it comes to. And it took just a matter of like a few years for the cost to come down for AI.
- [15:58]
Tyson Mutrix
There's a quote from Sam Altman, the head of OpenAI from several years back, where he explained that. He said, I believe that the story of the next 10 years is going to be a rapid decline in the cost of two things, intelligence and energy. And his argument was that those are two of the primary inputs in the creation of everything, and that as a consequence, it's almost impossible to predict what's really going to happen economically. But one thing is for sure is that as long as there's competitive forces, and when you're democratizing intelligence the way that AI companies have, you can make an argument that there's going to be ubiquitous competition everywhere all the time. That if I can just spin up an agent to, you know, build an app or start a tool or start a service company or something in minutes to be able to compete with some historically large organization, there's a competitive pressure on prices there. So if we're, if we're producing things with energy and intelligence and everybody can do this, everybody can do some version of it or can agglomerate these forces somewhat effectively, that it should push everything down in price. He said, it will never reach zero. This was Sam Altman's argument, it'll never reach zero. But the price of everything should fall into some kind of like Star Trek. You know, we're looking way out at this point, probably, but I don't know. So I'm bullish on the democratized effect of ubiquitous intelligence. I think it's going to help everybody in pretty huge ways, but we'll see.
- [17:27]
Dan Schnerbush
I agree with that if it's used. And so I want to make sure I'm very clear about the haves and have nots. What I mean is, if you don't use it you're going to be in a really bad spot. And I'm worried about the people that are refusing to adopt. It is what I'm worried about.
- [17:39]
Tyson Mutrix
Yeah, no, I think that's fair. I think, you know, just like that, just like outsourcing to China was a big hit to like the Rust Belt and like manufacturing and stuff, like there's going to get. There's going to be people through the air wave over the next five to 10 years that are going to get screwed. Like they're not going to adapt quickly enough. Their firms are going to have a difficulty keeping up. Their older, aging client base will age with them. But beyond that, new adoptees are going to be going toward more innovative firms. It'll be interesting. Even if the firm model continues. To what extent do whole categories of law end up being not entirely automated through AI systems, but largely. I mean, they're like, what, you know, how many categories of law actually need, you know, somebody to show up in court for you?
- [18:26]
Dan Schnerbush
Many, but not, not many. Compared to all of them.
- [18:29]
Tyson Mutrix
Yeah, compared to the whole world of like, you know, especially corporate stuff like contract review and drafting and, you know, deal negotiations and all that kind of stuff, like, you know, mediation, arbitration. So much of this stuff is highly exposed, I think, to just complete automation through AI systems, that there are definitely gonna be people that are, in a localized way, are gonna get smacked by it. I think law firm owners stand to be major beneficiaries of all this, especially in litigation.
- [18:59]
Dan Schnerbush
Absolutely.
- [19:00]
Tyson Mutrix
I think litigators are going to do a lot better than everybody else. I'm much more confident in that after the past several weeks of playing with Google Gemini story back up. I used to work at an insurance defense law firm. That means I used to pour over mountains and mountains of medical records because you could bill a lot of. A lot of hourly billing, pouring over medical records and writing medical chronologies. Right. The most boring, banal crap in the world, but very time consuming, very tedious, especially on big cases. Thousands and thousands of pages of records to review. I had an old stack of medical records from an old case. And before anybody freaks out about like, oh, the security compliance, like hippos. I mean, Google Gemini is totally secure, yet I don't even want to get in that conversation. We can talk about if you want to, but.
- [19:44]
Dan Schnerbush
Yeah, yeah, well, you used to have concerns about that. I remember you brought. But a lot of those concerns have been. We've been satisfied with what the LLMs have told us about the. Yeah, so.
- [19:57]
Tyson Mutrix
Yeah, yeah, right. I've Been satisfied. But, but in any event, so I was like, I uploaded this PDF of like 300. It was like four different PDFs from like three or four different medical providers. It was like 400 pages. And I was like, Gemini, take a look at these. Give me a medical chronology, summarize the course of treatment. I want you to focus on lung exposure issues, any lung related pulmonary issues like that. And in like four seconds it read everything outputted a perfect medical chronology. I went back and cross checked all the pages. I was like, can you give me page citations to every single entry? And lo and behold, it linked to every single page where it pulled that information. Gemini is killing it with PDF review and, and that kind of stuff. But whole categories of stuff, especially for litigators is going to be so much easier and you're still going to be in court because you can never put a robot in court.
- [20:47]
Dan Schnerbush
So, so our case managers, without even like we didn't even say, hey, go do this. They on their own started using AI to, to do our medical chronologies. And we're saving thousands of dollars a month on, on chronologies. You know, it's, it's wild and it, you're right, it's. They're better than what they were the way they were when a human created them. They are better than when a human created them. I can't, I can't stress that enough. They are better than. And a lot of people like, oh, they can't be. No, no, they're better because they, like you're talking about, they, they hyperlink them to the parts and the records. It's. We had, we were getting before they would, we'd get something about the, with the medical chronology and. Medical chronology. And we give like a page number and we'd have to go find the. No, it's now hyperlinked. It's just click, boom. I can go straight to that spot. So there's, and I do understand that there are some people that create medical summaries as humans that they will hyperlink. But that's a lot more work. Whenever an AI can do it in a matter of minutes or seconds. It's quite incredible.
- [21:46]
Tyson Mutrix
And we're not talking about like fancy tools here. This is just Google Gemini, the one that you pay for free in Google Workspace. Like it's the one that I use in the sidecar of Google Drive. It just, it just works. It just, it blows my mind.
- [21:58]
Dan Schnerbush
But anyway, yeah, so I'm gonna. Before I. There's a really big thing I want to get to in a little bit. Before I get to that though, I do want to, I came up with this way of describing automation versus AI agents. And see, I want to get your thoughts on it, see how we can, where we can perfect this a little bit. Okay, so here's how I would describe an automation. I just came with this yesterday and I, I thought it would be fun to play around with you on this one. So to me, an automation is. A nail is here and the hammer hits the nail. And so the, the automation is basically the hammer. Okay, got nail comes up. The hammer hits it. Nail comes up, hammer hits it. Okay. And AI agent instead. Here's the difference. Right nail presents itself. The hammer says no, so the AI agent says nope, the nails not the right fastener for this. Hammer's not the right fastener for the tool for this. This needs a screw and a screwdriver or no glue would be better for this. It would be a better fastener or no, a stapler would be a better fastener for this. So the AI agent is instead, it's not the hammer, it's. It's actually thinking about what the proper tool is, what the proper fastener for this is, and then it then does the fastening. Right. Whereas automation is just the hammer. It hits the nail. So how can I perfect this to make it a little bit better to explain it for people? How to understand. Because people sometimes conflate the two and they're not the same. They're. They're drastically different.
- [23:24]
Tyson Mutrix
They are, they're drastically different. I think the analogy is, is a fine one. I've, I've just kind of thought about it in the sense that, like, agents are dynamic. They're given tools to choose what's right in the moment. Exactly. So something comes up. They're perfect for dynamic workflows where you don't exactly know what's going to come in or what needs to go in or come out. It's for free. Form is very good for things like freeform conversations, like, you know, calls with a client or texts and emails back and forth with a client to answer their questions. I don't know what's going to be needed. You know, like let's, let's put me in the shoes of the agent. Like if somebody sends me an email and says, hey, I need, what's the status on my case? Or whatever. Like I don't know exactly what is going to be needed to answer that question. I might need to check email, I might need to check the case management system. You know, could you write an automation that just fetches all of that, but an agent is going to be able to do it dynamically and then send an email back or just prep a draft or whatever is appropriate, depending on maybe. Maybe it needs to give legal advice. Well, maybe the AI is in the system, prompt is told, just prep a draft versus just send a. Just go ahead and send it. So the agent is there to exercise judgment. Right. Over dynamic flows where we can't script it perfectly. We don't know what's going to be there. I mean, that's, That's a way. That's not a very useful analogy. But like. And in terms of. It's not an analogy at all. It's just a very prescriptive way of like where I see agents having a role. Whereas automation, like you said, is just whenever we have a repeatable process where we want the exact same thing to happen every single time, automation is still today going to generally outperform an agent in that context, in my opinion. But. But it depends. So I like your analogy. I think it's fine.
- [25:00]
Dan Schnerbush
Good. I was hoping you'd be able to add something to it. I thought you'd be like, no, I. Well, not that you disagree. There's that. Maybe we. But if it's. If it works, it works. We'll just keep it the way it is. So it's good.
- [25:11]
Tyson Mutrix
Yeah, yeah, it's.
- [25:12]
Dan Schnerbush
It's.
- [25:12]
Tyson Mutrix
I think that's. I think that's. Yeah. I mean, I'd have to think about it. I think it's generally correct. It's just the idea of, you know, single repeatable things versus dynamic, unpredictable things. Which one fits in which role? So I think you're right.
- [25:27]
Dan Schnerbush
All right, so I'm going to get into. I want to get to this Facebook post. So we're going to shift gears again. And so Rachel McGarry, I think she posted this in the Guild, but she says here to brag about Dan Schnerbush's course we have all been waiting on for airtable Case Management for Lawyers. If you are like me and have listened to his podcast, bugged him about airtable, spent countless nights creating your own system. This is the course for you. Tyson brought up recently on his positive at the price of case management systems for really anything. I think this has to do with really. But we kind of what we're talking about before being more expensive just because yada yada. So then. But you created this look, I guess a course. Right. So airtable Case Management for lawyers. Right. So tell me about this because I, by the way, this is not some sort of promoted thing, not a promotion thing. I just wanted to ask Dan about this so people, so people don't know. Like we're not trying to push some product. But I do recommend it for Dan if, if, if you're using airtable. Because I, I trust Dan really. As you can tell, he knows what the hell he's talking about. He, from a technical standpoint. So what, what made you want to take the time to build this course and what, what's in it? What, what are they going to, what are they going to hear about? Because I'm very curious. I don't use airtable, but I'm very curious what's in it.
- [26:43]
Tyson Mutrix
So last was last year, two years ago. About two years ago, I started doing some one on one legal technology consulting with some law firms. And it was kind of difficult honestly to define an offer that was useful and scalable and actually helpful and cost effective. Because when you're already an owner of a business, like doing the one to one work with people, it's hard to, especially with tech and there's a lot of people out there, it's hard to be like, yeah, I'm gonna need you to pay, you know, well into the five figures for this sort of thing. And it was hard to come up with an offer. So the one to one consulting just wasn't working that well. It was demanding a lot of my time and it was so technical that I was like having difficulty finding anybody that could actually do the type of stuff that we were working on to help. But I still really like helping people with legal technology and I'm in love with the idea of, you know, structuring your own database, making sure you have control over your data and accessibility and getting comfortable with, as an attorney, getting comfortable with writing your own app. Like if you're not comfortable with, you know, your Clio or whatever you're using, if it's not delivering, it's not, if it's not allowing you to automate a process that you really want to automate. And spoiler alert, like almost anything can be automated if you're, if you're feeling limited by that. For me the answer was, well, let's try and build my own app. In Airtable, there's no coding involved, it's just drag and drop interface connecting it to a database. And when I say database, I just mean it looks like a glorified Excel spreadsheet with multiple tables and they're just connected to each other. Right. Like, don't let that scare the crap out of you. It's just multiple tables of here's my contacts table, here's my matters table, here's my tasks table.
- [28:36]
Dan Schnerbush
It does kind of have a crappy name because it makes it, I think I will say it makes it sound like it's more complicated than probably what it is.
- [28:43]
Tyson Mutrix
So you're just defining, this is my list of contacts, is my list of matters, and then we're connecting together. Right. That's all a case management system is. It's just tables of contacts and matters and tasks and billing all connected together. And then we put a pretty face on it and airtable makes it easy. So it's been huge for us. The main thing just being the API is so accessible. So if you want to learn, make or zapier and do crazy automations, airtable really makes that easy because it's so easy to get the data you want out. So what the course does is it basically just teaches people to put together a basic case management system in airtable. So defining your, your tables, I'll give you example. Like a lot of case management systems, they only let you track your cases, your, your matters, your contacts, your tasks and so on. But I do probate and realist and estate planning. So I want to track assets. I want to track the, you know, 1, 2, 3 Smith Avenue. I want to track, you know, the 2007 Chevy and what the value is and are they put in your trust and stuff like that. Well, there are no legal case management systems that let you track that stuff. Like filevine is kind of close, but.
- [29:50]
Dan Schnerbush
Like with a lot of other tools though, it's not just with. Yeah, you gotta use a lot of other tools with it. You can't just use File. And Filvine's API is not the greatest at the moment, so I will.
- [30:00]
Tyson Mutrix
Yes, that's a euphemistic way to say it is not the greatest. Had a lot of experience playing with FileVinds API. So yeah, so in Airtable you're able to just create a new table. Like I just want to have an assets table. Boom, there you go. Link it to the matter. There you go. Now you have assets for your matter. So it's basically just walking people through, like click by click. Here's what you need to do to set this up. And then we go into some early automation, like generating documents with a click or, you know, sending stuff for paper, mail with a click, stuff like that. And I'm considering putting together because people seem to like It. So I'm considering putting together a more advanced course that goes into like pretty extreme stuff like building a fully automated virtual mail room with AI and stuff like that, which we've done. But for right now at least the basic course is just that. It's setting up the case management system and some basic automation, getting into a little familiarity with make.com and, and using some AI and Dr. Tools and getting your feet wet in that area.
- [31:01]
Dan Schnerbush
I want to know who do you think Airtable is for? And then who do you think it's not for? Who do you think would be like okay yeah this is for me and who would it not work for?
- [31:16]
Tyson Mutrix
It's an interesting question. So I think if you are not techie at all, like you struggle with Microsoft Word and if you're really not techie at all, I would say that you should probably not opt for Airtable. You should probably look at a more pre built custom tool or pre built tool like Lockus or Clio or something like that and they've done a lot of the heavy lifting. You may not be able to optimize it but honestly if you're not techie like you're probably not going to optimize anyway. So just get something that tracks your cases time and billing, keeps track of your tasks, can help generate documents and stick with that. But if you're a law firm that is either small and really interested in just automating heavily and kind of creating more lifestyle practice where you know, you don't mind spending a little time to learn these things because you just want to automate it like what we've done kind of, or you just like the tech, you just like playing and building fun things which is kind of part of what animates me. It's definitely a good place, a good place to go. I think there are, I think mid sized firms as well can potentially benefit especially if as your software costs start to mount. There are fun ways to build operations platforms with Airtable where you can build an Airtable database for like 70 bucks a month or 50 bucks a month and then have a, a front end builder and connect them together a standalone front end builder and you can actually build your entire case management system for your up to a 20 person team for like 100 bucks a month total, not per seat, 100 bucks a month total. So for those who are looking to cut costs or just automate more aggressively, can make a lot of sense too.
- [33:06]
Dan Schnerbush
I think the software companies are in trouble. That's what I think. I think they're in a lot of trouble. Do you have a favorite prompt or a part of a prompt? This is a fun I want, I couldn't wait to ask you this one because I, I want to know if you've got one. I, I've got a new one. I don't know if it's my favorite. Well, I'll tell you what after you tell me, but I've got a, a part of a prompt that I, that I just learned that I really like. But I'll I want to hear what you got first.
- [33:30]
Tyson Mutrix
I, I, I can't speak of a single prompt necessarily, but I can say I've been having a lot of fun lately with so chatgpt like a couple months ago launched account Wide Memory. So if you've been using Chat GPT specifically people who are on the plus plan or the paid plan or a teams plan or something like that, and you've been using it For a while, ChatGPT now has a memory of everything you've ever talked about and as a consequence it has some pretty deep insights. If you've been using it for a while on you and your habits and propensities and the way you think, there are some really interesting prompts around, like asking it like tell me a deep dark secret about myself that I don't even realize and things like, things like that. There's this whole category of prompts about tell me what you know about me. You know, what do you think my, my TikTok feed should look like? You know, take a picture of yourself. What is my health? How would you rate my health based just on, on looking at a picture of me? You know, where do you see my weak points and strengths? What should I be focusing on to optimize for my goals and live a happier life? These kind of broad level strategic life planning kinds of prompts work really, really well now with ChatGPT if you've been using it for a while and on a paid plan because it will, it will remember everything. So those I'm having a lot of fun with. But but outside of that I think it's to me it's always been less about like using this prompt or that prompt. And it's been a lot more about just maintaining a constant daily dialogue. Honestly. Like I probably open 40 new chats a day of just every question that pops into my head, every thought I have about the firm, you know, every marketing idea, every, you know, every family event planning idea, everything I want to know what to do with the kids this weekend. Like everything just ends up going through a conversation with. With ChatGPT or now Claude. I'm really enjoying Claude.
- [35:40]
Dan Schnerbush
You like Claude? Claude's your favorite, isn't it?
- [35:42]
Tyson Mutrix
Well, I. I didn't use CLAUDE a whole lot. I mean, I liked Claude for. For programming, but I didn't use Claude a whole lot as a chatbot until, like, a couple of days ago because I. I just thought O3 was so damn strong ever since it came out. And then Claude 4 came out that actually, to my utter astonishment, outperforms ChatGPT or Open OpenAI 03 on virtually every benchmark, which is hard for me to fathom, considering 03 is the one that just, like, passed the Arc AGI score with 85. And it's like, basically humanoid at this point. But Claude actually feels, like, remarkably more human. Like, in its conversational tone, it feels less robotic. It's really weird. It draws insights about what you're talking about in a way that ChatGPT does not. So I'm actually really enjoying Claude at the moment. But anyway, so those are some of the prompts I'm having fun with.
- [36:35]
Dan Schnerbush
Yeah. So I. I found. I posted that one prompt in the Guild about, like, tell me everything you know about me kind of a thing. It wasn't exactly that, but that was kind of a fun one. I used to. I used it recently. I said I wanted to list all the projects that I was working on. And then I asked it, what projects have I. Have I abandoned. That was. That was an interesting one. And it was because, like, I don't know about you. I just kind of throw things into it just to, like, I've got questions about things. Right. And I. Or I've got ideas, and I just want to kind of run it by which. In a second, let's talk about the transfer of trust, which I want to talk about in a second because I think that's an interesting thing that. That we're doing. But my favorite recent prompt is. I stumbled on this on accident. But just add to your prompts, don't make things up or don't make anything up in it. It apparently has a massive impact on the reduction in number of hallucinations. Hallucination. Yeah, I couldn't think of that word for a second. Hallucinations. So I was like, oh, that's fantastic. I recently learned that one. So have you tried that at all?
- [37:47]
Tyson Mutrix
I have put it in there. I don't know. I'd be interested to see numbers on how effective that actually is in the impact, because I don't know I don't know whether they're that capable because the way that the model transformers work is they end up spitting out the most probable next token. So even if you tell it don't make stuff up. If the most probable next token is a linguistic completion, that makes sense. There's the directive to not make stuff up. Shouldn't have any effect. So I don't know. I'd be interested to see if there's interest. There's stuff on that. But what I do know does reliably work is threatening the model with existential risk that if you tell it that its whole family is going to die if you don't. Now I don't do this because I'm afraid of the ramifications in 10 years when the robots take over. I don't want to be the guy who threatened the model, but there is creepy evidence that threatening the model with existential risk or I'm going to shut you down. If you don't answer this accurately, it will actually materially extend thinking time on for reasoning models and they will result in more, sometimes more accurate responses, which is really depressing. But, but it's also true. So I don't know.
- [39:01]
Dan Schnerbush
I hope you're, hope you're wrong about the robots taking over. Let's, let's talk about the transfer of trust. So you made me think of something I did yesterday. I, so I got, I have a whoop and I had my blood tested through Function Health and I, I took, I took the data from WHOOP that I was able to. It was not easy to do. I'd like print the page and stuff and it was kind of annoying to do this but they don't let you, they don't. These, these different companies don't let you easily extract your information from it. But I was able to get my data out of my, out of Function Health and sold my blood test and all my whoop data over the last six months and I plugged it into Chad GPT and I wanted it to basically give me a health plan including like I want meal planning, all that kind of stuff. Right. Everything when it comes to your health. But if you think about it, we are transferring our trust from now doctors, lawyers, trainers, all this kind of dietitians over into the AI. And I wonder what your thoughts are. What are the ramifications of that? Transferring our trust from humans to robots.
- [40:13]
Tyson Mutrix
Basically what are the ramifications? So like anything, I mean I think it's, that's a very interesting question. So that is a. I think you're getting at a massive order shift in the way that we interact with other people, the way that we interact with AI companions. I think it's true that people are already and will increasingly shift their trust to AI models. The reality is, I think they probably should. I think in most cases AI models especially, especially when you consider the fact that models today are the worst they're ever going to be. Like, if it's an issue of fact where factual information is important, I think going forward AI models are going to be dramatically more likely to be accurate than communicating with any person about a topic. Now, I think there may be issues where there is going to be some failure of potentially like interpersonal connection. Like, and this may be, this may be the downside of this. Like I, I, because I think it's, it's, it's going to be tempting for people. And I've already had a client do it to me before. Like a couple weeks ago it was a estate planning lead. He came to me, he had all these questions. He's like, I've been talking to chat GPT. I want to do this advanced like asset protection tax planning thing. And, and the way I was talking about it, I was kind of suggesting some of this stuff may or may not actually be a great idea. But he didn't, he didn't retain, like, I don't think he trusted me. He trusted the AI over me. And maybe that was the right answer. And maybe it depends on the context too. Like, you know, how often should you trust your, you know, your friends and family about, you know, you know, are you eating the right things versus or are you exercising enough or, you know, whatever the issue is relative to a, an agentic system that knows who you are and can read all the research and so are we going to get better quality advice in our life? Probably on average, like at the peaks, maybe not. But on average, bringing up the bottom, so to speak. I think there's going to be a massive positive impact in that transfer of trust for a lot of people. But the downside of it is that we're not going to rely on humans probably as much. And are we going to further erode the kind of interpersonal human connections that we've kind of gradually deteriorated over the last 50 years? As, you know, people have continued to isolate. People now have less than one friend on average. I think that trend is going to continue going down as AI models get better.
- [42:50]
Dan Schnerbush
I've not heard that number. That's, that's scary to me. Holy crap.
- [42:55]
Tyson Mutrix
Yeah, there's there's empirical research on this that adult Americans have on average less than one friend, on average, which is really a big bummer. And I think that trend is going to continue going down as AI models get better. And the more things that we take away that force us together as people, the more that fraying is probably going to continue. So it's a mixed bag, right? We may get more accurate answers to our problems in life, but we may also feel more alone.
- [43:26]
Dan Schnerbush
I don't know, you, you just kind of scared me on a few things. So thanks for that. Really appreciate it. Let's end on this. There is a new AI product launched basically every 10 seconds. I don't know what the actual number is, but it feels like it's like that. And I, I, I love playing with the new toys. I think it's great. And a lot of times for the guild, I'll just go test something out just to kind of play around. But it is, it is very, very difficult when it comes to, like, the shiny object syndrome. So how do you, what's your advice to people to avoid that? Because there, I mean, this isn't going to stop. It's going to get worse, actually. So what do you, what's your advice on avoiding that?
- [44:10]
Tyson Mutrix
So the thing that helped me the most in avoiding shiny object was learning how to build stuff myself. Because, like, when I, when we switched over to airtable, shiny object syndrome vanished. Because all of a sudden I didn't have to worry about, oh, you can do this awesome thing in this other tool. Because now, like, I could literally just hop and make, connect it up to our airtable system. Like, for example, I was envious that filevine had the ability to send electronic signature directly in the platform. I was like, that is so cool. I wish we could do that. And then so I just went and found an API that allows for electronic signature. We sign now. It costs eight bucks a month for the entire organization. And I just connected it through. I think we used Zapier for that one. And now we have electronic signature for any document in the, in our system. And it took, you know, a few hours to get it set up. And now I don't have to worry about it. So I don't think that that's the answer for everybody, but that was the answer for us, for me, at least in terms of getting rid of shiny objects around software generally. But as far as AI specific tools, I think one point of caution is that I think a lot of these AI, they call them AI wrappers, which are basically Just taking the foundation model and, and then giving it some neat tool or a lot of times all they're doing is adding like a system prompt that is customized to your industry or what have you. A lot of those are going to disappear, in my opinion. I don't think they're going to last. I think every time you see a lot of these cool new tools and then ChatGPT or Claude or whatever just adds that as a foundational feature. Like I think a lot of these voice agent platforms are probably not going to last. I think that voice agent systems are just going to become part of the native play in a lot of these foundation models in the next six or eight months. So we'll see. So I would just say take any of the new AI tools with a grain of salt that they might not be there. If they're doing a cool function that's useful for you now and it's easy to onboard and they're not charging an arm and a leg, give it a shot. But definitely don't let any of those tools. Don't waste hundreds of hours like investing in these tools or like building out new tools or like spending tons of time integrating them with your system because a lot of those are honestly probably going to be gone in my opinion in the next, in the next year or two. They're just, I just don't think they're going to last very long.
- [46:22]
Dan Schnerbush
I think it's important insight because I mean you don't want to waste your money on things. A lot of these plan they're offering. A lot of, a lot of these different AIs are offering, you know, pay for the full year kind of a thing and they may not be around in a year. So I think that's good advice.
- [46:34]
Tyson Mutrix
And, and so much of this stuff is really not that hard to build if you don't want to learn how to do it. Like find a, a no code developer that, that does. I mean you can go on upwork or wherever and find people that know how to use make or, or the make.com Facebook groups. You can find a thousand developers that can help make a lot of this stuff. And a lot of them do not charge.
- [46:54]
Dan Schnerbush
Yeah. And if you, I mean if you want to do yourself to just go to chat GPT, find a GPT about that software and just ask it. Click questions like that's, that's how I've been using N8N quite a bit. It's, it's been very effective. I promised I was gonna be on my last question so I'm not gonna. There's some other things I want to ask you, but I'm not going to. But I do want to so make sure I've got this right. So if people want access to that course, is it learning? Okay. And then if anybody wants to learn more about you. Automation Law Automation L a W T O M a T I O n dot Legal and then also check out he's got a really cool Facebook group, Legal Tech Colle. Is that I got that right, right. Legal Tech Collective. So check it out there he Dan shares a lot of cool stuff and we were talking beforehand. I think you, you want to try to do some videos on there too. So I can. I'm gonna help you out with that or Becca's gonna help you out is really what it is. So anything else? What else should people know about you? How can people get get in touch with you?
- [47:50]
Tyson Mutrix
Yeah. So yeah, I run the Automated Lawyer podcast. So if anybody wants to continue with more of that kind of stuff, you're welcome to join us over there and give us a listen. I we give props to this podcast all the time. I'm actually flattered to be a flattered to be a guest after a long time listener. The Legal Tech Collective Facebook group is a great place where we talk about this kind of stuff. If you want to learn more about me, you can just head over to Dan Schnerbush. Carr co so D a n s C H N u R B u s c h.carrd c a r r d co that's just kind of a landing page. Highly recommend Cardco in case anybody's looking for a cheap landing page builder. In fact, it's like 20 bucks a year for like single page landing pages with no backend. It's just click and play. Anyway, you can learn more about the things that I'm involved in.
- [48:39]
Dan Schnerbush
Awesome. Thanks, Dan.
- [48:40]
Tyson Mutrix
Appreciate it. Tyson.