Tech Brew Ride Home - (Profile) Casemark.ai

Episode Date: June 29, 2024

Find out more about Casemark at Casemark.ai. Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to another weekend bonus episode of the Tech Meme Ride Home. I'm Brian McCullough, as always. This is a portfolio profile episode.

And Chris is joining us for this one, Chris. Hello, hello. The reason being, because we're going to talk about an investment that the Ride Home AI Fund made, which obviously the Ride Home Fund made as well, but this is one of our AI companies from this year. We have with us, Scott Kvieten. the founder of Casemark AI, which you can find out more at as you're listening at casemark.aI. Scott, thanks for coming on. Thanks for having me, Brian.

Good to see you both. Good to see you, Chris. Yes. So let's right off the top. Case Mark, as maybe the name implies, is AI in the legal space. So give us just sort of like the elevator pitch for what Case Mark does, Scott. Yeah, I think the sort of shortest version of this is we've created kind of the easy button for attorneys for specific discrete tasks that they'd like to accomplish with AI. But we do it with sort of security and privacy in mind.

In other words, we deal with the implications around data leakage, training, making sure that this is safe for them to use. But then by boiling that down to really simple tasks that they try to accomplish, specifically what we focused on early on were things. like deposition summaries, case summaries, trial and hearing transcript summaries. And that's kind of the early phase of this, but what we're realizing is that over time, attorneys are realizing, wow, there's actually more you can do with AI. And once they see the simple use cases, they realize there's a bunch of ways that they can essentially take all this unstructured data that they have and then leverage AI to transform it into, you know, interesting structured analysis and reporting for what they're trying to accomplish. So you're talking about,

things like case summaries, like depositions, as you say, hearing summaries, contract review, discovery response. One of the sort of narratives for how at the very beginning, in this first inning of this era of the AI revolution, it's about getting rid of the busy work, or at least obviating the busy work of freeing folks up to not have to spend so many hours on stuff that allow you to focus on other more important things. Is that sort of the focus right now, what I just described, like, let's get rid of spending hours going over depositions and things like that? Yeah, let's get rid of the TDM work. You know, it's easy to lose the details or, you know, miss something because you're tired and you're on the fifth hour of going through a, you know,

what might be a boring deposition. And, you know, humans are, you know, fallible in that sense. And, of course, the flip side response then we get from some attorneys is, well, wait a minute, aren't you going to take away my opportunity for a bunch of billables? And our response is, well, no, one, you still should review what we're producing. But a lot of attorneys also have some, you know, artificial limits in place, whether it's on the defense side, they can only bill so much for what they're, what they get from these summaries. So it might take them six hours to do a summary, but they can only bill two. So they lose that four hours. Or on the plaintiff side, they just want to be able to do it as inexpensively as possible because they're doing it on a contingency basis.

So any costs they accrue then count against any potential future settlements. The other thing that I think a lot of folks forget about with respect to AI right now is that they sort of, you know, they talk about this doom and gloom around how it's going to replace all these paralegals and associates. And if we were at 100% AI usage today right now, I would agree with that. However, what people forget is that AI is actually going to enable, it's going to make it really, really easy for law firms to litigate. So we're about to go into a hyper growth in terms of litigation that's going to happen. So I think we're going to see a five to 10x increase in litigation, which means there's going to be a demand for not only the AI solutions, but you're going

to need humans to sort of arbitrate and traffic control on all this. So what we're doing is removing the burdensome, time-consuming tedious tasks and allowing attorneys to actually use their critical thinking. And we actually think this is going to increase their job satisfaction for sure. Insert lawyer joke here. I don't know if the phrase, an explosion in litigation is what some people want to hear. But I'm looking forward to that. Yeah. Well, and listen, if you're a lawyer, yes, I can see that as well. So sue me and says no one ever again. Yeah. Okay, let's let's imagine that I'm a law firm listening to this right now. What does it entail for me to start using your tool? Like, Is there, like, does it plug into my existing workflows easily?

Like, what are, what does it take to get running with Casemark? Yeah, we tried to keep it as simple as possible. And we also tried to solve the problem in the simplest fashion as we could. So you can essentially sign in, you know, log in with your email address, your law firm's address, whatever it is. You upload a file, you choose the workflow you want to run. There's a couple options in there if you want to choose. Otherwise, you can just do the defaults and click go.

And then a couple minutes later, you're going to have some results that you can then download and use. And I think what's interesting here, too, is sort of the work product that we generate is sort of a word document or a PDF. And while that might be boring to us as technologists for attorneys and law firms, their sort of, you know, programming language or software of choice happens to be a word document or a PDF. And so when we can actually download that, that PDF of a summary of, say, a deposition, and then we can actually, what we do is, we append the transcript. In other words, the source transcript is appended there. That's actually a really powerful, encapsulated litigation tool that they can then

drop into their case management solution, or they can forward it onto their insurance adjuster. And we're not completely replacing how they do their work. In other words, they can jump in, use our tool, and then drop it into the way that they do things. And attorneys really like that, because it's a way for them to kind of try these things. So often I see these really beautiful products

that are AI powered. but they require that the attorneys completely change how they do business. And if you think about a law firm that has, you know, 100 people, it maybe has 25 attorneys and then 75 supporting staff, if they all have to then learn some piece of software and then that software is continuously changing and that just, that will completely screw up their daily workflow such that if a law firm, you know, if all the paralegals lose an hour because of some new software, guess what, that's 75 hours of, you know, billable time that you've lost. And that's a big, big deal. And so we try to make our tools as simple and as easy as possible

and deliver a solution that basically mimics what they do by hand today. And then in the future, we'll evolve these tools to be sort of more advanced and sort of ingrain them in more of their daily process. But again, try to be as lightweight as possible today. That seems to have the biggest impact for us. I think the other piece, too, is by being a lean and mean startup, up and because you're looking at the entire sales and marketing team, we don't have time to be able to follow up and reach out to a lot of these folks. So we do have a self-service model, which

actually really helps these folks, because maybe it's the paralegal, the law firm administrator, the overworked office manager who's in charge of this. And they can jump into our solution. They can try it. They can break it without having to talk to a salesperson because they're always worried about asking a question that they think might be considered stupid, which obviously we would never say that. And we don't think that. But, but, you know, the legal industry is full of people who, you know, sort of shout down at, you know, the people who are the paralegals and associates by design. It's a weird industry. I can say this because my wife's an attorney, so I consider myself attorney adjacent. And so just to hear some of the stories of, you know, her when she was a

younger attorney, you know, she's now been doing this for almost 15 years. But just to hear how they just get sort of browbeat all the time. It's just, it's astounding. Anyways. Yeah, the last one, and then I'll let Chris give you some questions, too. But obviously here, one of the concerns would be, you know, I can't have these sensitive documents fall into the wrong hands, leak, be trained on. So how are you thinking of and dealing with things like privacy, security, stuff like that? For, you know, I mean, everyone that their business is business critical, but you're dealing with legal stuff

and you could blow up cases if you do it wrong. So what's your process? Yeah, for sure. So we have relationships with Amazon, Microsoft, and Google to actually have sort of private cloud infrastructure for all of the data that we ingest. And so the best way to describe it is when we ingest private data from our customers, we basically drop it into a container

along with the LLM of choice that we're using. We then do our transformations on it and then output some summary, and then we tear that whole thing down and it goes away. In other words, we don't actually use models that are, like we don't connect to Open AIs, APIs, we do everything inside of Azure with respect to Open AI, but we also are using Gemini, Claude,

a little bit of Lama, some mistral in there. But again, that sort of same theme runs true that we're not gonna train with their data. And we lean really hard into that. And we've gone through quite a few, not only sort of those third-party risk assessments that you have to do for some of the larger firms, but also having the CIA offices or their security teams coming in and really kicking the tires, getting into our source code even to see and verify, trust and then verify on these things.

And so I think that's a really critical piece there. And I don't know if it's verboten or not, but if I can share my screen, I can kind of show something really quick if that's okay. Is that sure? Sure. Knock yourself out. I mean, if you're listening on the podcast, this isn't going to knock your... Oh, yeah, yeah. I will try to narrow it, yes.

Yeah, and so what I've got here is basically kind of an high-level architecture diagram of how our system works. And kind of at the top of this are a series of work. We call them workflows. And, you know, the things of that would be like a deposition summary, a trial and hearing transcript summary, a case summary. And that sits on top of what we call our sort of workflow engine. And what the workflow engine does is it securely ingests this data from our customers and then puts it in a variety of different places,

whether it's an elastic search or a rag database or a vector database, such that when we ask the questions that are done by these workflows, which are essentially these sophisticated prompts, prompting chains that we do, the workflow engine knows where to retrieve that data, how to do it, how to verify that it's the right data, all those kinds of things. And then the second piece is that sits on top of what we call our LLM routing engine. And early on, we knew that we kind of wanted to be the Switzerland of providers in this space. In other words, we want to be cloud, LLM, and then in the very near future region agnostic.

And the reason that's important is our customers don't know or care what models we're using. What they're doing is signing a deal with us that we're going to deliver what they need at a decent price performance, that it's going to be accurate, secure, private, all those things. And then what we do is our LLM routing engine allows us to essentially do that across a range of different providers. So we're never beholden to any one provider. And like we said earlier, we're in the first inning of this thing. And we don't know who the winner is going to be. And there might not be a winner. So this ability to be able to pick and choose the different models for specific things that they're really good at is absolutely critical.

The last thing I'll mention, and then I'll sort of stop here, but is that all of these LOM models have something called content filter. built into them. And that content filtering is designed as a CYA mechanism to make sure that the big players are protected against or protecting sort of the general public from doing, you know, untoward things with their models like making meth or building bombs, like that kind of stuff. But when it comes to law firms, they actually have to deal with some really sensitive topics. Things like, you know, hate speech or law-making math. Or making meth, yeah. And we actually even had one of our, we just posted this story this week, but Lisa Peck is a civil rights and employment attorney in Northern California,

and they just won a $20 million verdict. And it was an employment law that had a bunch of racial undertones in it against Stanford health. And it was actually really, really powerful the way that it worked. But in any case, it's exciting to see how these things, these solutions can solve these problems. And so we've been able to figure out how to deal with that content filtering in such a way that we deliver the right. result to folks, even if it is sensitive content. And again, those are the kinds of things that I think a lot of folks aren't really thinking about quite yet.

So just to jump in there, like one, I just want to understand a little bit more about the blog post that you previewed there and what case marks participation or involvement was. But I think to make the point, I think what you're raising is super interesting because there's a great deal of conversation about alignment and, you know, making sure that these things, you know, whether it's less so about concerns about hallucinations and more about saying, like you said, like reasonable things. and not insulting someone or saying, you know, stuff that they're not supposed to.

Any number of times you might have asked chatypt for instructions to do something, you know, that is, you know, borderline or could result in borderline content, and then it shuts you down. So given that, you know, part and parcel to lawsuits and legal issues are, of course, as you say, sensitive topics, I guess if you just dig in there a little bit more, when you talk about these content filters, how does that actually work? Because one of the questions that I think people might have going forward is if we we see more LLM technologies built into operating systems and places like that, but they're also designed in such a way to prohibit the discussion of sensitive topics,

then that is actually not giving us agency and use of those tools in a way that's effective. And in your specific use case, you have to be able to talk about sensitive topics. Yeah, I mean, it's definitely very tricky. So in the way that the content filters work and it's implemented differently across all the different LLMs, but really by and large, it's sort of they have a set of, filters and you can back them off. In other words, you can say, you know, I want the highest filtering, middle, low. And then in some cases, you can turn it off, although you tend to have to have an actual relationship with the provider of that LLM. And then the flip side of that

is you can also use what's called an uncensored model, which is a model that's been built effectively as an open source model. And they've basically taken the governors off. In other words, they've said, okay, anything goes, you can do something here. And so we use a mix of those to be able to solve these problems for us. So as we're going, through and summarizing these documents, we might hit a content filter. We have internal to our LLM routing model, what we call content filter failover. So then we could take that same chunk and try it against another model or we can just go straight over to an uncensored model.

Now, we wouldn't default to that uncensored model because we're potentially exposing ourselves to a bunch of liability there. And so that can be kind of difficult. And in the case of Lisa Peck, she came into that case. She had three weeks when she got added as co-counsel. So her and her team had to get up to speed really quickly. They had 30 depositions they had to get through. So they started feeding their depositions into our solution, and they kept failing.

And so when we dug into it, we're like, oh, it's failing on the content filter. And so Stephen, who's my CTO and Chris, you know Stephen, he's this kind of hacker guy. He's like, oh, yeah, we can figure this out. We just route around it with this uncensored mistral. And so we were able to solve that problem for them really quickly. So she was able to get up and running really, really fast. And not only that, she said this completely changed how she will prepare. for trial now because she can get up to speed much, much quicker. And more importantly, they also

were taking what are called the dailies. So at the end of a day of trial or the day's proceedings, you can go to the court reporter and ask for the uncertified transcript. And so it's basically when they hack out the shorthand. And she would take that and feed it into our system and generate a summary for the day that she could then stand on to her constituents. So whether it's people who had a bearing in the case or the folks who were actually involved directly in And she could just email that off and then not have to update them via phone or email. They could just read for themselves what happened. And then she could go and her and her team could focus on the next day's proceedings.

Because when you're in trial, it is intense. You have to manage the, you have to see what's happening with the jury. You've got to deal with opposing counsel. You've got to make sure the judge is happy. There's interruptions. There's a witness might get tired. All these things that you have to do is just absolutely brutal. And this was a four-week trial.

So she said that the dailies were such a critical tool for. her to be able to keep people a prize and then like get that off her plate. And that's what's really exciting. And I think in the future, what we're going to start to see are real-time tools that are listening in the courtroom. And if somebody gets up on the stand and says something that doesn't corroborate with what they said in the deposition, six months ago, that's going to have any huge impact. Or if an attorney mentions law, attorneys love to mention case or cite some law and say, oh, ORS164.5 means that, you know, this case is moot. And it's like, well, actually, that's a lot.

not what that means. So if in real time you can pull up not only the actual law itself, but also the sort of layperson summary so that you can tell the judge, actually, Your Honor, this is what it really means. I'll let opposing counsel read for themselves. But, you know, and so those are the kinds of things that I think are going to be really interesting and how this is going to change litigation, especially in real-time trials are going to change, I think, significantly here. So just to double click on that, like I think, which is a phrase I've never used before, I don't think. Nonetheless, it makes me wonder about the, one, the depositions. And two, you know, just as, you know, Brian and I are constantly either using, you know, Fathom or Granola or other

tools for recording our calls and then having notes come afterwards, it sort of is mind-blowing to imagine that there is still, not that there doesn't need to be, but that there is still sort of a human stenographer in the courtroom who's writing everything down sort of like super fast. And so you imagine these courtrooms, you know, and, you know, typically like, you know, the male judges with like the big wigs or whatever. And like, the thing hasn't changed. I don't think we have wigs anymore, but I know if you're talking about it. But like, but like, when you're describing this to me, like, and the fact that it's not, I mean, maybe there are some courtrooms that allow recording instruments and allow some 70s era technology. Like, what you're describing in this futuristic courtroom where everything is being fact checked kind of like in the moment or at least reference back to things that it said before.

and it just kind of puts into stark contrast and relief what it means to bring a type of accurate memory and recording into the courtroom situation and how the whole artistry and to some degree artifice of like legal challenges can change through this technology but you're dealing with the court system so i guess my question is sort of maybe multifold one is how do you imagine like the legal profession kind of changing and adapting with these new technologies and how much is the courtroom itself a headwind against some of those things that you just described and sort of like real-time fact-checking or whatever being inhibited by the way that the court might actually not allow these certain technologies to come into the courtroom yeah i mean that the interesting

thing around court reporters um you know the whole concept of court reporters obviously is you know you do take shorthand, but if you want to use, like if you do deposition and you have a court reporter transcribe it, the court reporter then has to certify the results of that deposition. In other words, they have to literally swear to the accuracy of it. And if you don't get that, you can't actually use that testimony in court. So that's one of the reasons why court reporters and that whole concept still exists. What's fascinating is there's a huge shortage of court reporters nationally here in the U.S. I would think so, yeah. Because it's, it's, it's, you know, everybody's, you know, going into what would be considered sort of greener pastures,

but it's a very, very interesting space. And then to your point, you know, the legal system itself, just the adoption of technology takes a long time. And I think for good reason, right? I mean, if, you know, you think about the folks who are staffing these offices. Well, it's a system that's based on precedence by definition. Correct. Yeah, yeah, yeah, for sure. And so it just, the folks who are staffing these offices may not be up to the latest and greatest in terms of technology. So the idea of them bringing in some real-time reporting system that, by the way, if it breaks, what do they do? Like, they can't call an 800 number and find somebody to come in. And then, by the way, the wheels of justice just grind to a halt,

which is bad. But to your point, like, how is this going to impact things? I think what's really exciting is when you think about the way a deposition happens. I mean, here in the U.S., there's 11.5 million depositions that happen annually, and it's a huge, huge business. It's a huge market. But these attorneys, to take a deposition as an attorney is actually a process. It takes you several years to get good at it, especially a contentious deposition. And so one of the things that we've seen is folks who can take real-time depositions and then transcribe those. And let's say before you go into the deposition, you know there's 15 questions you want to have asked and answered. And the way the depositions work is you get one bite at the apple. You get to ask those questions and then

that's it. You don't get to go ask later. So if you're an attorney that's a junior attorney and it's really contentious and you sort of lose your sense, and you didn't get those 15 questions to ask and answer, guess what? You're kind of host. So imagine if you start at 9 a.m. It goes till noon and you have a lunch break and you could feed in that transcript and then you could compare it to the 15 questions and the AI could come back and say, oh, great, you're nine of the 15. So you got six more that you need to get answered after lunch. And so now all of a sudden we can up-level attorneys that wouldn't normally have the skills to be able to do these things. And to me, that's very exciting. Not only is a job satisfaction thing, but it's also

helps them kind of up level to a certain extent. And more importantly, yeah, sorry, go ahead. Well, I was going to say just one more on that, I guess, that topic, which is around, I guess, like talent and how the field itself is changing. One thing that I think Brian and I have been witnessing and observing as we've been talking to other founders going into other spaces, of course, where AI is starting to be deployed, is just hiring is like sort of in, I mean, I suppose this is like the meme and the reality going around, but that people can't hire enough staff talents. They can't keep them. They can't retain them.

Whether it's because now there's a whole new world of influencers or whatever it is, it seems as though, like you said about the court reporters, that there's going to be deficits in certain roles. And so the need for AI to actually perform those duties is becoming more and more significant. And I guess the other thing that I would add to that is that you also have in some cases, is older or more experienced staff retiring or moving on or whatever it happens to be. So maybe the legal profession is alive and well and it's super vibrant and it's got a ton of new people coming in. But maybe you can speak a little bit to that, right?

Because if there is sort of a huge number of junior folks entering into the force but they don't have that level of experience, then AI does become kind of like a tool of assistance so they can execute at a higher level and maybe learn the ropes faster. than if they were to do this on their own. Yeah, I mean, I think there's a bit of that. I mean, I think the reality is over the last, you know, 30 plus years, the legal profession has really fought sort of tooth and nail and utilize sort of pointy elbows to box out what technology could do to undermine,

which is effectively a giant pyramid scheme, right? The idea is that you work your ass off and if you get to the top of the pyramid, you get to reap the benefits of all the work of all those people below you. And oh, yeah, you're supposed to, you know, talk down to them and all those kinds of things. And so what I actually think AI is finally going to be the thing that is going to disrupt this model, whether law firms like it or not, it's going to happen. And so I think there's a huge opportunity for these early career attorneys to either spin up their own firms that can handle a lot more bandwidth or for firms to be able to focus on specific things that they're really, really good at, but that they could do at scale around, you know, specific subsections of litigation. That you couldn't do that before because you had to have maybe a relationship with the client and build that over time so they would drive business to you. I think a lot of that is going to change here over time.

And, you know, we'll see where it goes. But, I mean, I'm very, very excited just based on what we see and how we see attorneys using this stuff, especially the early career attorneys who are like, holy Toledo, I am no longer, you know, going to be at the, you know, the whim of these senior partners. But the other thing, you know, to your point, Chris, is we see a lot of. attorneys who are using it sort of on the quiet. In other words, they use their Gmail account to sign in to use our solution. And if I'm a junior attorney and I'm having a hard time with some motion or some filing, I can't ask the attorney above me because you're like, are you an idiot?

How do you not know this? Everybody knows this, you know, and then they feel like a moron. Whereas if they can ask an empathetic, you know, infinitely patient chat bot that will say, well, actually, it looks like there is a problem here. You might want to ask a question. And then when they ask an informed question, then, you know, the partner will still likely, you know, make fun of them. They'll go, oh, wow, this person actually had a really thoughtful question. You can see my, you know, my concern about the legal industry and their, you know, hassling of people.

But by and large, I think is the answer. It is ready for disruption, for sure. And to me, that's exciting. And we're, you know, we're at the crux of this. And we're seeing it, right, like all day long. Now, the flip side of that is the news media and the story. will say that the stuff is getting adopted at lightning pace and that everybody's using it and Harvey's a, you know, $2.2 billion company.

I got news for you. This stuff takes a while to adopt. And if you're a startup in the legal space and you think that you're going to go from, you know, one to 10 million in revenue overnight, your boncos. Because to really get these things utilized and folks using it on a regular basis and paying for it, you have to make sure that it's working all the way to. down to the office manager, the paralegal, they have to understand it, they have to trust it, they have to know how to integrate that into their process. And there's a reason why private

equity is so dominant in the legal tech space is because churn is so low for people, but you don't switch case management products because that's like a nightmare for a law firm. And so anyway, so that's the other thing that I think a lot of folks forget about. This stuff isn't going to happen overnight. I mean, it's going to happen. It's just it's not going to be overnight. So slow and steady is going to kind of win the race here, I think. Yeah, so I guess the other question, you know, for listeners who don't know, Scott and I go way back. I mean, we probably have one of the original. Spread Firefox. Yes, yes, exactly.

Back when you were at the Oregon State University running the computer lab, basically. The open source lab, yeah. The open source lab, thank you. Back in 2004. So, you know, we go back 20 years. And, you know, we've seen each other through a number of different eras and waves of technological development from open source into. I was an advisor to Urban Airship, which is now just airship, I believe, and essentially is focused on the productization of push notifications once the iPhone came out and provided a

solution to replace SMS. The reason why that I think is relevant is because Scott and I have been in these moments of transformation of basic set of behaviors and norms where, you know, to sort of invoke McLuhan, the idea is to take an existing set of media content and move it into a new one and it's just more efficient. And so in a similar way, you've got these depositions and it's going to be more efficient to process them. And yet the transformation doesn't actually happen until you're a few years in. And because you've built on the existing substrate or set of behaviors that exist, you can then start to make subtle tweaks and changes to process.

So my question to you is, as you're seeing this, you have this interesting challenge to on the one hand deliver a product and a service and a tool that meets the legal profession where it is currently. Like you said, you're not going to replace the whole case management system, you know, one and done, but over time, you get to redefine the way that that work actually is executed. So when you think about when you play out case mark over the next several years, you know, one, how do you see this moment, you know, with AI being different from past technological revolutions in terms of you building product? And then, And then two, how do you see the industry, the legal profession, sort of changing,

presuming you're successful? Yeah. I mean, you know, history doesn't repeat itself, but boy, does it rhyme. I'll tell you what. I just feel like we're going through, you know, what we saw with, you know, the original dot-com bubble, Web2O that you and I live through Chris, you know, mobile 2.0. To me, what's happening is a lot like what happened with the iPhone. with respect to AI.

And what I mean there is AI is having what I would call an iPhone moment. We all had mobile phones in our pockets when the iPhone came out. But when we saw the iPhone for the first time, we said, oh, this is what a mobile device is supposed to be. And while there were people who had sort of 20 years of experience with mobile when the iPhone came out, that was kind of all out the window. And AI has been around for 40 years. But when we all saw ChatGBT, we said, oh, this is what AI is supposed to be. And so what that does, just like with the iPhone, is it creates a moment in time where people are saying, oh my gosh, I have to have that.

I have to have that. So that creates an opening for a company like ours to create that as a wedge. We have an answer for you for your AI solutions because that's what we've said we can do. And do we have a specific product or feature or point solution right now? Yes, we do. But really, the way to win in this to me is to build a partnership with customers over time because you're navigating what is a disruptive cycle that will take a. decade to accomplish. And this is exactly what we saw with urban airship. And so it's been very,

very interesting to sort of to see that, you know, relate as it relates to, you know, AI and this moment in time with sort of legal and anyways, to watch how this is playing out. But again, you know, the other thing that's also very interesting is there's a lot of companies raising a lot of money at valuations that are ridiculous and they don't have product market fit. And they probably won't get product market fit. And it's like me having gone through this, a gazillion times, I'm shouting at the top of my lungs, like, oh, my God, we're really doing this again? But, like, you know, again, that's how these cycles go, and that's okay. But when we think about where we're kind of where we want to be or where we think this is going to go,

is I see a lot of people saying, like, we're going to be this AI assistant for, you know, law firms, and you can ask it questions and it's just going to do things. And what you end up usually with is this sort of mile, wide, an inch deep solution that doesn't really solve specific problems. And so what we said is, wait a minute, what if we can solve from the bottom up? In other words, we can solve these specific discrete tasks really, really well. And we'll get them bulletproof because we're going to throw thousands and thousands of tries at it. And we're going to use it, all these people are going to use it, such that when we do layer on that assistant down the road,

we can then look at a case folder and say, okay, well, we see some pleadings in there, some transcripts, some medical records. Here's our suggestion. We think you should do some deposition summaries and medical chronology. And then, by the way, we'll sum the whole thing up with a case summary report that you could forward on to somebody. And oh, by the way, we have a couple next best actions you should take because there's a couple filings that are doing here and you should check those out. Right. So that's where I think this is going to evolve. I mean, that's what I'm talking about there is five years of work, right? And it's not going to happen overnight and people aren't going to adopt or trust it overnight. But that's where we're going to kind of get, I think.

And just like when the first push notifications got sent, I immediately saw, I was like, oh, wow. And it was companies like Starbucks that saw it that said, you know, okay, yeah, cool. We have a mobile app. And in our mobile app, we have our menu. But what they saw was, oh, wow, this is going to change how we do ordering, how we do tipping, how we do, you know, stored value. Like all those things that have a fundamental change to your business that, you know, manifest themselves in an app, but really mean you have to change how your store works and how, like, all those things. So that's the transformation that legal is about to go through right now. And it's going to be painful and it's going to be hard.

But those who navigate it and find the right partners to help drive that are going to come out the other side, you know, know, way, way stronger and way, way more profitable, in my opinion. So you mentioned that your attorney adjacent and you've gone into a little bit of your background, but can you give me sort of the inception of the idea of this company, maybe touching on where you were when you started working on this company and like where the lightball moment came from? Yeah, I mean, I think, so just on me, I mean, I'm a serial entrepreneur, I, you know, cut my teeth at Amazon, turn of the century. I have the dubious honor

being on their Y2K team for what was basically the biggest nothing burger ever, was really active in a bunch of open source, open technology stuff. Just like Chris said, you know, he and I were there when Mozilla spun out of AOL, and we both helped kind of get Firefox 1.0 out the door. Then again, we partnered up on making sure that open ID and OAuths and pulled, you know, Facebook, Microsoft, and Google together to say, this is it. And then, you know, we created the Open ID Foundation, which we, I think I was a chairman for a little while, and then we let that off into the world. And that's created a really awesome thing there. And then I started a company called Urban Airship to do push notifications.

And I really moved over to the business side to sort of build, scale and sell BDB enterprise SaaS companies. After that, my co-founder and CTO, who I still work with this to this day, Stephen Osborne, we started a point of sale system for the cannabis industry that we sold in 2017. Most recently, we sold a transparent ledger for physical assets company. So it was a blockchain company, except that instead of saying, if you build it, they will come. We actually had a customer who had sports member abelia that we were doing. Anyways, we sold that last year, ended up being an exit out to Fanatics because our customer got acquired. So we got kind of swept up into that. And so I had a team.

And this is literally June 1st of 2023, so just over a year ago. And we were kind of looking at what we're going to do. And my wife, obviously, being an attorney, she runs a firm here in Portland. They do Oregon, Washington, and Idaho, insurance defense. So they have big retailers and, you know, let's just say, you know, driving or vehicle-related businesses. You know, they support all those. And so she kind of jokingly said, hey, why don't you do something that helps me for once? And like a lot of attorneys, she had the first experience, which was she went into chat,

UPT. She asked her the question. The response was lackluster. And she said, well, this is dumb. And so I said, well, you know, I actually think there's something here. And so, you know, like I do when I go into a new, when I'm launching a new company, the first thing I want to do is get product into markets so we can increase our pace of learning, so we can figure out what the heck's here.

So we immediately said, let's launch a Word and Chrome extension, a Word add-in and a Chrome extension. So we found the open-source solutions out there. We hired the devs. And then we launched something within a month, basically, about four weeks. And we learned really quickly that attorneys don't want those. They forget about the add-in. They hate paying for software that they might not use that month. The idea of SaaS to them is just they just don't understand it because they bill for the time that they work.

So they just don't understand it. In the fall, we tried some fine-tuning of models for firms. But then we were sort of left with this whole concept of, okay, cool, we fine-tuned your model. Here's a prompt window. Go ahead and start, you know, use AI. And they're like, what the hell do I do with this? And so that let us down launching what we call legal prompt guide.com, which is a really simple solution. It's a freebie site there.

It helps attorneys figure out how they're going to do prompting and really understand, just like I took a generation for folks to figure out how to get those Google searches right to extract what you want out of Google. The same thing is true, like 10X for chat, GPT or just chatting with any kind of generative model. So that's when we realized, oh, we have to take what we learn from all these prompting and turn those into easy buttons, leveraging these methodologies and the things that we've learned to be able to get the most out of the LLMs. And that's when we sort of launched, really launched in earnest in kind of January.

And then we had a bunch of legal tech players approach us saying like, hey, we want to license your stuff, which then led to, oh, we need an API. And now we've got folks connecting to the API. And it's literally just kind of taken on a life of its own now. And now we're stamping out all kinds of workflows for folks to be able to really increase our scope. And then anybody who's plumbed up to our API, guess what, they get to have access to any of the workflows that we have. So it just really has this multiplicative effect right now. We're kind of have landed on this, you know, AI as infrastructure play. And then, you know, from a pricing standpoint, we've tried to be really aggressive instead of saying,

here's how much it costs for an attorney to do it. We're going to charge just a little bit less. We actually are going the other way, which is we know what our cogs are and we're going to tack on a margin that leaves a lot of room for people to resell our stuff. And that's working really, really well right now. And so, yeah, that's kind of the, the host. for why we landed on it. And then my wife's been really instrumental in helping us kind of craft some of the initial workflows that we did. Because the key with the LLMs is they have the answers. You just have to know how to ask the right question. And the answer can't, the prompts can't be,

pretend you're a lawyer. What you have to do is ask a question like a lawyer does. And then it will respond with a response that a lawyer would expect to see. And so those are the tricky things that we don't have the experience in. And that's why we lean on like my wife and a couple other sort of trusted advisors in that front. So yeah. It's funny. I'll just comment and then a question. You know, what are the one of the reasons why we invested was along the lines of the AI Varietals thesis that Brian and I work from. And that essentially is this concept of bringing together some subject matter expert, you know, with AI engineering or the use of generative AI. And, you know, the fact that obviously your wife is the subject matter expert in this case sort of allows you to,

you know, ground in real truth, the way in which you're bringing generative AI into a specific context where there's specific requirements. And those requirements come from language. They come from a set of expectations that people in the field, you know, you lose so much credibility if you answer the question in any other way that is not sort of legalese. And so it's not enough to, I think, like you said, like pretend that you're a lawyer and then create a prompt, there's a whole lot more that goes into that, which is around culture and norms and why communication happens a certain way. And I think it's valuable just to keep that in mind from a product design perspective is that when you're designing something in the generative AI space, you have to match the language of the

person that you're actually interacting with. So the question that I have is about maybe the, if not the metaphor, the conceptualization of bringing the output of an LLM into the legal context. And what I mean by that or what I'm getting at is the increasing interest in agents or having kind of like AI employees that work alongside someone else versus, let's say like a co-pilot, which you treat as like a chat bot, which you know is an AI and sort of sits alongside, let's say, a document versus a set of workflows that are almost like macros but lit up, you know, for like the 21st century with generative AI. And we seem to be at this moment, you know, you mentioned AI as infrastructure. structure, but there's still a question of how someone chooses to invoke these services. And we're, I think, in some ways, struggling with the right interface for this. So my thought and question is kind of about that, whether agents are a relevant metaphor for you, whether it's, you'll have multiple different sort of paralegals, but they're all powered by

Gen. I, and that eventually, you know, the junior, like, lawyer goes to one that is a specific subject matter expert, or where you imagine these workflows are the right concept and framework for delivering casemark, and you're going to stick with that because that's the one that people seem to understand, and you're going to go forward with that. Yeah, I think in the near term, what we're finding is what's the easiest way that people can kind of rock this, such that they feel confidence, one, that they can test it, try it, get a result, and then verify that result quickly and easily.

And I think that sort of manifests itself in the easy button today. that idea of a changing paradigm where you'd actually trust an assistant to do those things or a co-pilot even. I think that's it's going to be how it happens. It's just those things never happen overnight. You don't adopt those paradigms overnight. I mean, if you think about any new paradigm, I mean, the only one that I can think of that took hold overnight was sort of like the pinch to zoom or like any of the sort of touch interfaces. But you weren't doing anything new there. You were just doing something that was obvious that you would do, you know.

Well, it was like a digital version of something that, you know, kind of already existed to some degree. Yeah, exactly. And so I think that, I think that, you know, for us, we just tried to do the lowest, easiest way for people to get in and use it. I think that will evolve over time. But we always have to continuously, you know, check in with our customers on this. The other thing is that we've learned, too, is we sort of track the sessions and watch users and their behavior of our solution. I'm always astounded at how easy we, how simple we have to make the interface so that they understand it.

Because these attorneys, you know, a lot of the times can be sort of neophytes. And that's a problem. That's our problem. It's not their problem. And so often I just keep seeing so many companies that are building software that I know attorneys will never use, or at least, you know, even this current generation, the sort of youngest generation, they'll probably be able to make it work, but those aren't the folks who are making the buying decisions

or have the budget to purchase the line items that would require for these things. So, yeah, it's very, very interesting to watch how this whole thing's shaking out for sure. I've got one more before we wrap, because I always like to, you know, a lot of the people listening to this episode right now are listening for, if I was in his shoes,

what would I do? Would I make decisions this way, that way? As I'm looking at the tech meme back end right now. There literally seven hours ago was another legal AI company that announced a raise. This is a hugely competitive space. In fact, by the way, I got out of that that legal tech startups have pulled in $356 million so far this year. Now that's down slightly compared to. to last year. That's from Geekwire, by the way. But so, okay, number one. Oh, and it's twice that now

with Harvey closing that round, right? Okay. Well, so number one, hugely competitive space. Number two, you know, a lot of times VCs will encourage startups to raise big rounds because it sort of freezes the market. Like if somebody has hit a billion dollar valuation, then that stops other people from being able to raise. So, In a broad way or as narrow away as I can ask this, how are you thinking about being in a super competitive space where, A, you've got a competitor that people would look at the market and say further ahead, bigger valuation. You've made the argument also that this is early innings. But number two, again, I'm looking at the tech mean back in the amount of legal AI startups. How many lawyers are there in the United States?

That's a good question. I want to say it's somewhere in the neighborhood of like 800,000 attorneys, and that's going to grow by like 18% in the next 10 years or something like that. So a big market, but also that's why everyone's going after it. So as a founder of a company in a super competitive space, how are you thinking about that? Is it just we got to focus on the product. and everything will work out?

Or is there strategically, as you're building this company, how big is the team right now? We're 10 people right now, yeah. As you're building the company, do you make decisions based on that that you wouldn't make if you, like, had a space to yourself? You know what I mean?

Well, I mean, I think if you have a space to yourself, you don't have a business, right? Or you're doing something so crazy that no one knows it's a business yet, yeah. That's true. That's true too, but that's so rare because I mean, even the idea of sort of, you know, monkeys, you know, an infinite number of monkeys in front of an infinite number of typewriters, you know, the odds are there's going to be a lot that are really, really close to each other.

And so I feel like that's kind of where we're at right now. And the answer to this is, you know, you have to kind of play the market a little bit, but also make sure that you're focusing on the fundamentals. I mean, we did this at Urban Airship, which was let's make sure we get a sales motion in place such that we know how and who we're selling to, and we have a strong relationship with those folks, and let's lock that up. And, you know, we did raise, you know,

I don't know, when all I was there, we raised about 50 million, which is like chump change now. Our series A was $1.1 million, which is like laughable. That's not even a pre-seed now, you know? And so, but, you know, for us at Urban Airship, what we did is we said,

let's get that sales motion right. Let's make sure we have a really solid product market fit. And then we're going to watch this market. And what we're going to do is we're going to identify those players that either overraised or couldn't find product market fit or ran out of gas. And then what we did is we acquired them for pennies on the dollar. And we got great teams and tech. And we were able to fold that into our sales motion. And, you know, the uplift was anywhere from 15 to 30 percent in sort of net new growth from a business perspective.

And I think the same thing is going to happen here. And that's kind of how we're architecting our business. It's one of the reasons we only raise the $1.7 million is we're in this. this, we wanted to do a small amount to prove out some things that we think we're proving out right now. And then, you know, we'll likely raise again. And, but we don't have to because we're now throwing off a bunch of cash and building an interesting, compelling business. But as we hit 18, 24 months, my gut says that this is going to go faster than the original sort of SaaS cycle because AI is moving so fast that we'll be able to look out at the landscape and say, well, where do we have gaps?

What could we fold into our sales motion such that it'll, you know, know, we'll have some uplift there that allows us to continue to grow and scale. And it actually turn us into not just an interesting company, but a brand that people will depend on for legal in general. And that, to me, is where it gets really interesting. Because AI is going to become a feature. Right. Without putting words in your mouth, what you're, it sounds like you're saying to me is let other people get headlines with big raises. Let, you know, 20 other people raise rounds. We'll make the headline. down the road if we've executed on the sales because then we'll make the headlines because

our raise will be based on the revenue that because we've executed on the plan. Yeah, I think that's I think that's about right. But I think there's also some element of, you know, you have to play the market a little bit too. You know, obviously, you know, after we close the round and now it's, there's so much interest in the space that, you know, we get investors constantly pinging us. And I take those calls. I have those conversations with those folks.

and there's always interest, which is great. That's great. But again, we have to focus on execution. And I know it sounds boring, but that's the critical pieces right now that I think is really important. Now, I think the unfair advantage that we have is as a team, we've all worked together. It's a bunch of airship folks that have come together, urban airship folks that are putting this together. And we've all scaled companies like this before. So a lot of these startups have the challenge of market headwinds and implement,

and all those things, as well as learning as they go on how to scale a company. Team dynamics too. The team dynamics are really helpful. We can shout at each other or get angry and have a disagreement. And guess what? The next day we wake up and go, okay, cool, we solve that problem, right? And there's no, there's, because we have that relationship, I think that sets us apart as well. And so I think it's one of the reasons we've been able to accomplish so much in just a year.

And, you know, especially when I look at some of our, you know, the sort of folks that are in our same space or, you even in the same portfolio companies, like a gradient and others, you know, the amount we've accomplished and the challenges they have around just what I consider simple stuff around scaling, they're struggling with. But I always, you know, give feedback on those things and say, like, hey, maybe you want to think about this. And that to me is also kind of really fun to kind of help those other companies. So, yeah. So you mentioned the pre-seed round, which was led by a gradient, which is Google's seed fund. We are honored to be a part of it as well with the Wright Home AI Fund and the ride home fund as well.

We're excited. So happy you all are involved, without a doubt. Very exciting. And again, if you want to learn more as you're listening right now, it's casemark. . But also, if people are listening and want to learn more or want to get involved, are you hiring? Do you have an ask for this audience that you never know who's listening that might have something that they can deliver for you? I mean, we're definitely hiring.

We're looking for folks on the engineering side of things without a doubt, especially DevOps, site reliability. We're having some what I call wonderful problems around scaling in the sense that this is going really, really fast. And so we have to figure out how we're going to scale those things. So always looking for folks who are interested in a serious going concern that has the fun problems of a startup, which are scaling and those kinds of things. And, you know, if you have folks that are running, whether it's an insurance, you know, friends that are running an insurance defense firm or doing, you know, transactional work like personal injury stuff, then, hey, they should point them in our direction and have them check out casemark.

You know, we offer up a little free plan where you can, you can test it out with a couple different free summaries. And we find that when people see the free version and they sort of try it against a, especially a deposition they've taken themselves and they see the summary, they're actually really blown away. So that would be my ask was just check it out and spread the word. By the way, according to Claude, there's 1.3 million lawyers in the U.S. Oh, so I was wrong. Okay. I was going off of what I'd seen in my last Gartner report, which I probably misquoted. So great.

Okay. And Claude must have it right. I was going to say, AI to the rescue, right? You know, they have references, so I believe it. Yeah. That's true. That's fair. Okay.

Fair now. So again, casemark, casemark. Scott, thanks for coming on and telling us all about that. Chris, thank you for joining me as well. Thank you, Brian. Thank you, Chris. Thanks so much for having us. Really appreciate it.