Latest / AI Tools for Practicing Lawyers / Episode 021 Law Firm AI: Custom Made Or Off The Shelf?

Transcript

Episode 021 Law Firm AI: Custom Made Or Off The Shelf? · AI Tools for Practicing Lawyers

• 0:03
  SPEAKER_01: Welcome to AI Tools for Practicing Lawyers.
• 0:07
  SPEAKER_01: Practical, no nonsense guidance to help attorneys put AI power to work in their practice right now.
• 0:15
  SPEAKER_03: Lawyers are getting bombarded with AI products.
• 0:20
  SPEAKER_03: Top of the list, Harvey.
• 0:23
  SPEAKER_03: Close second, Logora.
• 0:25
  SPEAKER_03: Always there, Claude and ChatGPT, and let's not forget, a million and one verticals.
• 0:31
  SPEAKER_03: But what if the right answer isn't buying another product?
• 0:35
  SPEAKER_03: What if someone came into your firm, studied how you practice law, and built an AI operating system specifically for your office?
• 0:45
  SPEAKER_03: Well, that's what today's guest claims to do.
• 0:48
  SPEAKER_03: I'm Ron Drescher.
• 0:50
  SPEAKER_03: I'm Heather Gardner.
• 0:51
  SPEAKER_03: And welcome to another episode of AI Tools for Practicing Lawyers.
• 0:56
  SPEAKER_03: Today's guest is Jerry Dearden.
• 1:00
  SPEAKER_03: I found Jerry and his company, David AI, in my social media feed and thought it would be a good workflow options solo episode, but I changed my mind after talking with Jerry.
• 1:13
  SPEAKER_03: Jerry, welcome to our podcast.
• 1:16
  SPEAKER_00: Hey, hey Ron.
• 1:17
  SPEAKER_00: Hey Heather.
• 1:18
  SPEAKER_00: Thanks for having me on.
• 1:19
  SPEAKER_00: I'm super pumped to chat all things AI, all things AI and law firms, and anything else we want to talk about.
• 1:25
  SPEAKER_03: So see, this is why this is why I wanted to have you, and I apologize for interrupting because I I loved your energy when we talked, and uh I really wanted to have you on the show.
• 1:37
  SPEAKER_03: What is David AI?
• 1:40
  SPEAKER_00: Great question.
• 1:41
  SPEAKER_00: So David AI exists because we believe that every business really should be implementing or at least thinking about AI, but most either don't know where to start or don't have the skill set to do it.
• 1:56
  SPEAKER_00: And David, we call it David because we believe in giving the Davids.
• 2:00
  SPEAKER_00: So the small firms, the small businesses, the same tools and resources to go up against the Goliaths.
• 2:06
  SPEAKER_00: Or we do have some larger clients as well.
• 2:08
  SPEAKER_00: We want to give them the tools and resources to take on Goliath-sized goals.
• 2:12
  SPEAKER_00: So AI's out there.
• 2:15
  SPEAKER_00: For some people, it's exciting.
• 2:16
  SPEAKER_00: For some people, it's really scary.
• 2:18
  SPEAKER_00: For some people, they might even hate it.
• 2:21
  SPEAKER_00: But at the end of the day, I would say in the next six to 12 months, maybe there's a longer window.
• 2:27
  SPEAKER_00: I think it's shorter.
• 2:28
  SPEAKER_00: If you're not at least using AI in your day-to-day, you're going to be falling behind.
• 2:34
  SPEAKER_00: And that's why we're here.
• 2:35
  SPEAKER_00: We believe in helping businesses to do just that.
• 2:39
  SPEAKER_00: And a little bit of what you said at the beginning, Ron, we believe instead of having to fit a software around your firm or around your business, you should build a software that is specifically modeled for your business as opposed to fitting your business within the software.
• 2:55
  SPEAKER_00: We had a digital marketing business that we did for a really long time.
• 2:58
  SPEAKER_00: So while we help on the AI side, we also are very experienced on helping firms know how to grow and how to leverage AI on that side of things as well.
• 3:06
  SPEAKER_00: And yeah, excited to dive deeper into Ron Ron.
• 3:11
  SPEAKER_00: Ron scares me a little bit.
• 3:12
  SPEAKER_00: He's always asking the tough questions.
• 3:14
  SPEAKER_00: So I know we're going to, we're going to get some softballs, hopefully, but we're going to get some fastballs here too.
• 3:20
  SPEAKER_00: We'll we'll see what happens.
• 3:22
  SPEAKER_03: Well, I'll tell you going in what my goal is.
• 3:26
  SPEAKER_03: My goal is not to put you on the spot or not to in any way communicate that there is anything wrong with your product or your company.
• 3:35
  SPEAKER_03: My goal is to help listeners identify whether they and David would be good fits for each other.
• 3:43
  SPEAKER_03: Yeah.
• 3:44
  SPEAKER_03: So, all right, I understand once upon a time you started as consultants and you became developers.
• 3:50
  SPEAKER_03: Is that correct?
• 3:52
  SPEAKER_00: Uh yeah, yes and no.
• 3:53
  SPEAKER_00: I would say that maybe for the first three months of business, it was me and one of my business partners that were definitely a little more consultative.
• 4:02
  SPEAKER_00: And we realized, hey, we we need a technical co-founder.
• 4:05
  SPEAKER_00: So we brought in a technical co-founder because we were identifying a lot of solutions that we believed we could build, me and my business partner.
• 4:14
  SPEAKER_00: But as we started to try to build them, we were like, but we need we need someone with more technical knowledge, more technical background.
• 4:20
  SPEAKER_00: So we brought in a technical co-founder, and that way we kind of tied the knot between, hey, we'll we'll look at your business and we'll tell you where AI fits the best and where it's tied to an actual business outcome.
• 4:32
  SPEAKER_00: But we can also, if you would like, build it for you.
• 4:36
  SPEAKER_00: And that way we tie that knot between instead of just saying, hey, here's something that would be nice, we can say, hey, here's something that would be nice, and uh we can deliver it to you and uh have it actually be effective and executed in your firm.
• 4:50
  SPEAKER_03: How did you happen upon law firms as your primary customer focus?
• 4:57
  SPEAKER_00: Yeah, yeah.
• 4:58
  SPEAKER_00: I I will I'll I'll add right now we're we're pretty industry agnostic.
• 5:02
  SPEAKER_00: We do have a lot of law firm clients.
• 5:04
  SPEAKER_00: And part of the reason for that is one, we identified it as an opportunity because more than any other industry, legal is spending the most money on AI, which is kind of cool.
• 5:14
  SPEAKER_00: So for whatever reason, law firms, lawyers are a little more open to uh the idea of AI.
• 5:21
  SPEAKER_00: And I think the other end of that is that a lot of legal work is kind of repetitive.
• 5:28
  SPEAKER_00: It's uh it's work that AI is really good at doing, document intake, reviewing documents, doing research.
• 5:36
  SPEAKER_00: These are things that AI really excels in.
• 5:38
  SPEAKER_00: And because of that, we we saw law firms as an opportunity where one, we found at least a lot of law firms are really open to the idea of AI.
• 5:46
  SPEAKER_00: And two, AI does a lot of the work or the workflows or the jobs that law firms need to do.
• 5:53
  SPEAKER_00: It can do them really well.
• 5:54
  SPEAKER_00: So it's it's a good uh marriage between actually having the demand for it and uh being to being able to actually use it.
• 6:04
  SPEAKER_03: You know, when I when I found you on my social media feed and I filled out the little form for a demo, you thought I was a regular law firm and you confirmed that our our audit was going to be at such and such a time.
• 6:20
  SPEAKER_03: If I had been a regular law firm, what would the audit have looked like?
• 6:27
  SPEAKER_00: Yeah, that's a great question.
• 6:28
  SPEAKER_00: So, really, when we audit a firm, what we want to understand are one, what's the size of your firm?
• 6:36
  SPEAKER_00: What's the specific niche that you focus on?
• 6:39
  SPEAKER_00: And then do you have multiple departments?
• 6:41
  SPEAKER_00: Do you have multiple employees?
• 6:43
  SPEAKER_00: Do you have multiple attorneys?
• 6:44
  SPEAKER_00: So we want to understand kind of the makeup of your law firm.
• 6:47
  SPEAKER_00: And then we want to go step by step of, hey, Ron, what does your day-to-day look like?
• 6:52
  SPEAKER_00: What does your week to week look like?
• 6:53
  SPEAKER_00: What does your month to month look like?
• 6:56
  SPEAKER_00: We're trying to identify what are the common workflows or the common tasks that you execute on day-to-day.
• 7:03
  SPEAKER_00: And then if we take it a step deeper, then we want to look at what's your tech stack?
• 7:07
  SPEAKER_00: Are you currently using any softwares to execute on that?
• 7:10
  SPEAKER_00: And basically, how are you doing it now so we can understand where AI could best fit in and either assist you or completely automate or take over an entire workflow or task.
• 7:22
  SPEAKER_03: Have you talked with any law firms that already had an solution in place and it just wasn't working for them?
• 7:33
  SPEAKER_00: That's a great question.
• 7:34
  SPEAKER_00: I would say that most of the law firms we talk to are already using Claude, for example, that's Anthropic's product.
• 7:42
  SPEAKER_00: And I think Claude did a great job of penetrating law firms.
• 7:47
  SPEAKER_00: So a lot of law firms that are lawyers we talk to are using Claude to review documents or even help draft documents.
• 7:55
  SPEAKER_00: But we we call that reactive AI, meaning that they're still having to go into their Claude or into their ChatGPT and say, hey, do this.
• 8:03
  SPEAKER_00: Um what we want to help identify is are there opportunities where you can make your AI proactive?
• 8:08
  SPEAKER_00: Can you make it an actual employee?
• 8:09
  SPEAKER_00: So I would say that's the biggest one.
• 8:11
  SPEAKER_00: Also, a lot of law firms we talk to have already had conversations with Harvey or with Lagora, and they've just found that those solutions are almost too enterprise for them.
• 8:22
  SPEAKER_00: They're too big.
• 8:22
  SPEAKER_00: They don't want to pay one to two K per seat to have that in their firm.
• 8:27
  SPEAKER_00: It's and a lot of it is more focused on enterprise level kind of situations.
• 8:33
  SPEAKER_00: So I would say a lot like have wanted it and researched it, but there's not really a tool that fits how they currently like to do things, especially on the smaller side.
• 8:44
  SPEAKER_03: I want to unpack something you said because it's I it's really interesting to me.
• 8:49
  SPEAKER_03: You talked about reactive use of AI tools versus proactive use of AI tools.
• 8:55
  SPEAKER_03: And you and you mentioned the example of typing into a chat bot, kind of one prompt at a time, as reactive use of AI.
• 9:04
  SPEAKER_03: What would proactive use of AI look like in a law firm?
• 9:08
  SPEAKER_00: Yeah, proactive use would be, for example, when a client sends you a document and you download it, you put it into Claude, that would be reactive, right?
• 9:19
  SPEAKER_00: So you're taking it, you're like, hey, review this document, blah, blah, blah.
• 9:23
  SPEAKER_00: That's a reactive use of AI.
• 9:25
  SPEAKER_00: Proactive would be that you've built an AI agent, or we can just call them an AI employee, that automatically sees someone sent you a document in your email, downloads the document, puts it into Claude, finds whatever you need to pull from that document based on your workflows and frameworks, and then pings you and says, Hey, Ron, Jerry Deirdon sent in this document we needed.
• 9:48
  SPEAKER_00: I reviewed it.
• 9:49
  SPEAKER_00: This, this, and this needs follow-up or looks good.
• 9:52
  SPEAKER_00: I put it into our practice management software, whatever.
• 9:55
  SPEAKER_00: So that that's the proactive use of AI.
• 9:57
  SPEAKER_00: Instead of you going to Claude and saying, hey, review this or draft this, it automatically just doing it for you because it knows it's supposed to take that.
• 10:08
  SPEAKER_03: Well, that sounds like a really exciting capability.
• 10:13
  SPEAKER_03: How do you engineer that?
• 10:16
  SPEAKER_00: Yeah, yeah.
• 10:17
  SPEAKER_00: So that's that's where that audit or fitting is really important because while most firms do things relatively the same, Ron might have his own proprietary approach to how you do a specific task within your law firm.
• 10:30
  SPEAKER_00: So understanding that, documenting it, and then basically we take that and we use, we can use clawed code or codecs or even traditional code, and we program an agent using that workflow to then be able to execute on the task.
• 10:47
  SPEAKER_00: And that that's a random snippet I'll lay out there, is that I would say is really important for law firms, especially if they want to implement AI in the future.
• 10:55
  SPEAKER_00: But I would say even if you're not going to implement AI, if you just want to be able to bring in new employees or anything like that, is to make sure that you're documenting your processes, because that's going to make it easier to train an employee or make it easier to train an AI agent on how to execute on a task.
• 11:13
  SPEAKER_03: So I I'm not quite ready to let go of the audit process because something that I always experienced when I was in practice, and I tried to implement something new in the firm, is the onboarding process is so different from one company to the next.
• 11:33
  SPEAKER_03: And sometimes there was just insurmountable friction in that process.
• 11:39
  SPEAKER_03: How do you how do you help law firms get past that, for lack of a better word, reluctance to open their vaults?
• 11:51
  SPEAKER_00: Yeah, yeah.
• 11:52
  SPEAKER_00: So are you saying like it was always a pain onboarding?
• 11:55
  SPEAKER_00: Like if you brought on a new software, like a vendor like us, you'd be like, oh, it's kind of a pain onboarding this.
• 12:02
  SPEAKER_00: Or okay.
• 12:03
  SPEAKER_03: That's exactly right.
• 12:04
  SPEAKER_03: Like, all right, I have this audit, great.
• 12:05
  SPEAKER_03: They sent me a 40-page uh questionnaire to fill out.
• 12:09
  SPEAKER_03: Uh is that the experience, or is it a can you somehow, you know, tighten that experience for your customers?
• 12:18
  SPEAKER_00: Yeah, yeah.
• 12:19
  SPEAKER_00: So part of the reason we do the audit is because we want to avoid the friction.
• 12:24
  SPEAKER_00: In the AI space, there's a lot of shiny objects.
• 12:28
  SPEAKER_00: There's a lot of things that people think would be really cool to build, or maybe even that they want to build, or there's probably even lawyers out there who have gone into Claude themselves or Replit or any of these no-code tools and been like, oh, I want to try to build this agent, see if it can do this, or this intake process, whatever it may be.
• 12:45
  SPEAKER_00: Part of the reason we do the audit is we're looking for one, is it possible to build what we want to build?
• 12:52
  SPEAKER_00: Is it reliable?
• 12:53
  SPEAKER_00: Is it scalable?
• 12:54
  SPEAKER_00: And is it tied to an actual business outcome?
• 12:57
  SPEAKER_00: And then we're also looking through the firm to understand which employees are most ready or most open to AI.
• 13:04
  SPEAKER_00: And um, and then we need to make sure that we're training them on it.
• 13:07
  SPEAKER_00: So we really focus on those grand slams first so that we're building something that is actually useful, that's easy to use, and that's actually going to help with the work.
• 13:18
  SPEAKER_00: And I think focusing on those grand slams instead of focusing on a bunch of different shiny objects ensures that we avoid the issue of building something that never gets used or building something that can't be used because there's not technical know-how.
• 13:36
  SPEAKER_00: This should be an easy thing that is implemented in the firm in a simple way, as opposed to having to learn a whole new software or train a whole new employee.
• 13:47
  SPEAKER_00: I don't know if that answers your question, but it's it's the reason we do the audit is because we don't want to waste our time on things that aren't going to actually be used.
• 13:57
  SPEAKER_00: So we dive deep, identify those solutions, present them to Ron, for example, if you did have a law firm, and then we say, these are the grand slams we identified.
• 14:07
  SPEAKER_00: Um, how do you feel about these?
• 14:09
  SPEAKER_00: Good.
• 14:10
  SPEAKER_00: Okay, we can even show you a mock-up of how they work, and then um the goal is for them to be as simple to use as possible.
• 14:17
  SPEAKER_03: See, what I'm trying to understand is where the line is between having a good experience with just buying your solution as opposed to having a good experience with building your solution, especially now that so many of these frontier models have become builder friendly.
• 14:40
  SPEAKER_03: I mean, you know, Heather is always amazing me with the things that she's building that read all of her emails and summarize them overnight.
• 14:52
  SPEAKER_03: She wakes up, she's got a digest of everything.
• 14:55
  SPEAKER_03: Are you seeing that more and more lawyers want to be a little build friendly?
• 15:00
  SPEAKER_03: Or are you seeing, you know, lawyers are just, yeah, just just put the product on my desktop and and leave me alone?
• 15:09
  SPEAKER_00: Yeah, that's awesome, Heather.
• 15:10
  SPEAKER_00: I I would love to see some of the stuff you built.
• 15:12
  SPEAKER_00: So yeah, I would say I've found most are becoming builder-friendly, and it's because we want to build these agents where they like to live and where they like to interact.
• 15:24
  SPEAKER_00: For example, we have some lawyers that love using iMessage.
• 15:28
  SPEAKER_00: And if we can just have them send an iMessage to their agent and get information or execute on tasks, that's great.
• 15:35
  SPEAKER_00: For others, they like having a web app that just is on their locally hosted device, or in other words, their laptop.
• 15:40
  SPEAKER_00: There's a fancy way of saying laptop.
• 15:43
  SPEAKER_00: And they just want to be able to log into a web app and interact with their agent there.
• 15:48
  SPEAKER_00: So we just try to meet them where they most like to interact or where they most like to converse and execute on tasks.
• 15:56
  SPEAKER_00: And for others, it could be in Microsoft Teams.
• 15:58
  SPEAKER_00: For others, it could be in Slack.
• 16:00
  SPEAKER_00: So there's a variety of ways where we can architect it so that it fits into how they like to work and their workflow.
• 16:09
  SPEAKER_00: And that's where we've had the most success.
• 16:11
  SPEAKER_00: Instead of building something where they have to log into a software or a platform they never use, they can just message it.
• 16:18
  SPEAKER_00: They're more likely to actually use.
• 16:21
  SPEAKER_03: So are there are there firms that should never build custom software?
• 16:26
  SPEAKER_00: Yeah.
• 16:26
  SPEAKER_00: Yeah.
• 16:27
  SPEAKER_00: Part of the audit is always kind of a build versus buy uh mentality.
• 16:30
  SPEAKER_00: I mean, I think most firms probably have or need software at this point.
• 16:35
  SPEAKER_00: I'd be, I haven't talked to any law firms yet who don't have some sort of tech within the firm.
• 16:41
  SPEAKER_00: And we're always looking at a build versus buy debate because sometimes we might be more expensive than something, or sometimes we might have a tool, might have a capability that they've just built really well.
• 16:53
  SPEAKER_00: And it makes more sense to tap into that.
• 16:56
  SPEAKER_00: So that's part of our audit process as well.
• 16:58
  SPEAKER_00: There, there is a chance that we come back to a firm and we say, hey, you should keep this, this, and this tool, but we could build this, or hey, we should keep all the tools, but we can build you an agent that helps you to centralize it all in one place instead of you having to log into Cleo and log into another tool, and uh it just kind of consolidates the process.
• 17:20
  SPEAKER_03: You know, Heather, you're talking to lawyers all the time, and they all have these different systems, whatever, right?
• 17:29
  SPEAKER_03: Every law firm builds things differently.
• 17:31
  SPEAKER_03: When you're looking at a law firm, do you ever think, wow, they really should be doing X, Y, and Z?
• 17:38
  SPEAKER_02: Oh, yeah, every day, all the time.
• 17:40
  SPEAKER_03: Like, what's what's the thing you're seeing that that these law firms should be doing that most law firms are missing?
• 17:48
  SPEAKER_02: I I think in our industry, we're still seeing so much resistance to AI.
• 17:53
  SPEAKER_02: I just feel like there is a lot of missed automation, time-saving efficiency tools that could be easily implemented, but we also can't make any attorney or law firm do what we wish they would, of course.
• 18:08
  SPEAKER_02: And when you're in my position in the service that I provide, all I can do is provide information, not of course make them change their procedures or software.
• 18:20
  SPEAKER_03: So, Jerry, do you do you get a lot of resistance from your customers as to the things you want to recommend?
• 18:28
  SPEAKER_00: Just just you, Ron.
• 18:30
  SPEAKER_00: You were the only Yeah, I think uh we definitely do.
• 18:34
  SPEAKER_00: I think one of the the biggest concerns is actually something you brought up, Ron, when you first started using Chat GPT was like hallucinations.
• 18:41
  SPEAKER_00: They they don't want the AI to make them look bad or make a mistake or offend a client or just do something that it's not supposed to do.
• 18:51
  SPEAKER_00: And that's why it's really important to architect these solutions in a secure environment that is built around exactly how they like to work and stays within those frameworks so that you avoid like those hallucinations, so that you avoid any major mistakes or errors.
• 19:09
  SPEAKER_03: Can you tell us about any success stories you've had where a client has been resistant or intimidated to bring AI into their firm, but after you provided them a solution, the light bulb went on over their heads and they said, yes, yes, thank you.
• 19:28
  SPEAKER_03: This is what we need.
• 19:30
  SPEAKER_03: Could you tell us anything like that?
• 19:32
  SPEAKER_00: Yeah, let me let me think.
• 19:33
  SPEAKER_00: I mean, there's a few.
• 19:34
  SPEAKER_00: I would say that oftentimes it happens actually at lower levels of companies.
• 19:40
  SPEAKER_00: Uh usually those that we're selling to, they they want it, they know they need it.
• 19:46
  SPEAKER_00: They they might be surprised at how how good it is.
• 19:49
  SPEAKER_00: We might start with just one solution and then they open their mind to, wow, okay, this kind of like Heather built, like, hey, this helped me manage my email and my calendar.
• 19:57
  SPEAKER_00: Can it actually do this as well?
• 19:59
  SPEAKER_00: And I think once they open their mind a little bit to the capabilities that AI has and they see it actually working, then they can get excited about what more it can do.
• 20:10
  SPEAKER_00: Lower levels, I think a lot of employees, and maybe Heather even sees this in like the paralegal space, they are worried AI is going to replace their job.
• 20:19
  SPEAKER_00: So we've met resistance there sometimes.
• 20:22
  SPEAKER_00: And we always position it as hey, this is like an Ironman suit for you.
• 20:27
  SPEAKER_00: This is an Ironman suit for your employees.
• 20:29
  SPEAKER_00: We want it to amplify your work.
• 20:31
  SPEAKER_00: We want it to be something that helps you to enjoy your job more and um do more with less.
• 20:39
  SPEAKER_00: And I think once people experience it, once they have it in their hands and they see what it can actually do for them, that's when the resistance goes down.
• 20:50
  SPEAKER_00: So I I can't think of a hyper-specific story, but generally we found, hey, we want to show you at least one thing that this can do really well.
• 20:59
  SPEAKER_00: And typically, once people experience that, they're like, okay, so if it can do my email and my calendar, could it also review documents?
• 21:07
  SPEAKER_00: Could it also update my practice management software and do some of these things that people don't always necessarily take a lot of time, but maybe take, I call it brain calories.
• 21:19
  SPEAKER_00: You still have to log into your practice management, you have to update the client form, all that, blah, blah, blah.
• 21:24
  SPEAKER_00: This can streamline a lot of that stuff.
• 21:27
  SPEAKER_03: So that's really very interesting because I had the sense when we spoke that David was kind of a one and done company.
• 21:37
  SPEAKER_03: They cut you, they come in, they do the audit, you, you, you, you make your build, and then you move on to the next customer.
• 21:44
  SPEAKER_03: Am I wrong about that?
• 21:45
  SPEAKER_00: Uh yeah.
• 21:46
  SPEAKER_00: I mean, I would say that that might be the case for some firms that we work with where they have developers in-house or software engineers that maybe they just didn't have the bandwidth, and we helped them do a project, and then they can manage, maintain, and update it.
• 22:00
  SPEAKER_00: But we are building a piece of tech.
• 22:03
  SPEAKER_00: And with a piece of technology, we hope everything always goes the way it's supposed to be planned.
• 22:09
  SPEAKER_00: But there are issues that can come up.
• 22:11
  SPEAKER_00: So we always stay on board to make sure that it's always running as efficiently as possible.
• 22:15
  SPEAKER_00: And then on top of that, you all have probably noticed that Claude, ChatGPT, they're always coming out with new models.
• 22:22
  SPEAKER_00: Sometimes Claude's the best at one thing, and then the next month ChatGPT is the best at it.
• 22:27
  SPEAKER_00: And then randomly, perplexity might be really good at another thing.
• 22:31
  SPEAKER_00: So we also make sure that whatever tasks an agent is doing, that they're using the best model for that task.
• 22:38
  SPEAKER_00: And that's constantly changing.
• 22:40
  SPEAKER_03: Big in news maybe a month ago was that Kirkland and Ellis was spending a half a billion dollars to build their own AI system.
• 22:51
  SPEAKER_03: And I I never really understood what that was.
• 22:54
  SPEAKER_03: Were they building their own LLM or were they doing something else?
• 22:58
  SPEAKER_03: So not talking necessarily about KE, but talking about the solutions that you provide.
• 23:05
  SPEAKER_03: Would you ever just build an LLM out of whatever open source technology or whatever for a company?
• 23:14
  SPEAKER_03: Or do you think that's kind of like frying an egg on a volcano?
• 23:17
  SPEAKER_00: I think it's definitely like frying an egg on a volcano.
• 23:20
  SPEAKER_00: I would love to build, I would love if every company could have its own LLM, but it's way smarter to just tap into what Claude, what chat, like what Anthropic, OpenAI, et cetera, have already built.
• 23:33
  SPEAKER_00: But I will say we do encourage all the companies we work with to have what we call their own company brain, which you could think as its own company LLM, which is basically a model within your company that is trained on all of your processes, it's trained on all your workflows, it's connected to all your tools.
• 23:53
  SPEAKER_00: And this, it's not an LLM in the traditional sense when you think of like ChatGPT and Claude, but it is your own mini-nternal LLM that is connected to everything within your company so that whenever you bring on a new employee, they can tap into that.
• 24:09
  SPEAKER_00: Whenever you want to train a new AI agent, it can tap into that.
• 24:13
  SPEAKER_00: And it's basically just uh it's like Google Drive on steroids, you could think of it that way, where it can pull any and all information about your company.
• 24:22
  SPEAKER_00: So, in a way, you can still build your own LLM and you should have that company brain.
• 24:27
  SPEAKER_00: And if nothing else, I would say if if uh everyone on this call or on this podcast did anything today, and you're already using ChatGBT or Cloud, go in and make sure that you're connected to the different connectors that those tools have.
• 24:42
  SPEAKER_00: So your Google Drive or your uh Gmail or your Microsoft Teams, whatever ecosystem you're using, connect all those to your cloud, to your ChatGBT, and that can start acting as your company brain over time.
• 24:54
  SPEAKER_02: If you have the enterprise level, just to clarify for everyone, not if you have the free open source tool, please do not do that.
• 25:01
  SPEAKER_02: Yes, yes.
• 25:02
  SPEAKER_02: Thank you.
• 25:03
  SPEAKER_03: All right.
• 25:04
  SPEAKER_03: Wow, Heather, what a perfect segue to uh the harder questions that you have to ask.
• 25:12
  SPEAKER_03: So anyone who's going to be building an AI operating system in your firm, there's there's questions that have to be answered.
• 25:20
  SPEAKER_03: For example, how do you protect confidential information?
• 25:23
  SPEAKER_00: Yeah, yeah.
• 25:24
  SPEAKER_00: And uh something, Ron, you actually brought up on our last call that brought me to to repentance was uh adding a dpa to to our agreement.
• 25:34
  SPEAKER_00: So so we do have that now, by the way.
• 25:36
  SPEAKER_03: Well, that's great.
• 25:36
  SPEAKER_03: How did you how did you go about creating that?
• 25:40
  SPEAKER_00: Because of you.
• 25:41
  SPEAKER_00: You asked, you asked me, you're like, do you have a DPA?
• 25:43
  SPEAKER_00: And I was like, I don't, I don't think so.
• 25:45
  SPEAKER_00: And you were like, you should get one.
• 25:46
  SPEAKER_00: And I was like, okay.
• 25:47
  SPEAKER_00: So we went to our lawyer and he was like, Yeah, we actually should.
• 25:51
  SPEAKER_00: So it's a part, it's a part of our uh agreements now.
• 25:55
  SPEAKER_00: But I will say it's not just security, that that data protection, it's just not just a document.
• 26:01
  SPEAKER_00: It it starts with how you architect the solution.
• 26:03
  SPEAKER_00: And that's again why we start with the fitting, and part of the fitting is even preferences, but we want to look at is this something that we should build on a locally hosted device?
• 26:13
  SPEAKER_00: Meaning that we build it on a computer or Mac Mini or whatever it may be that lives in your firm.
• 26:20
  SPEAKER_00: So I like to think of it like the movies, like the only way someone could access that is if they go in there and they like plug a USB into your computer and pull hack the information.
• 26:29
  SPEAKER_00: There, it could be privately on the cloud.
• 26:32
  SPEAKER_00: So there's a lot of different ways to host the data that can be smart.
• 26:36
  SPEAKER_00: And then uh the next level to it would be governance, meaning like your partners at the firm should have different levels of access than the associates or paralegals.
• 26:48
  SPEAKER_00: Like you, they they can know what people get paid and all that stuff, but maybe you don't want the rest of the employees to be able to pull any of that information.
• 26:56
  SPEAKER_00: So we architect all of that to make sure that internal data is secure and safe, as well as any external data that you're pulling into the firm.
• 27:04
  SPEAKER_00: And and again, there's there's a variety of ways to do that, but the DPA, a big one, like you mentioned.
• 27:09
  SPEAKER_00: And then uh from there, just uh making sure Heather brought up a good point with uh the models that you use.
• 27:16
  SPEAKER_00: Everything we build, we're not building on the free models or anything like that.
• 27:20
  SPEAKER_00: So how we have our agreements with OpenAI and Anthropic is they're not gonna train their models using the data of the solutions we build for clients.
• 27:30
  SPEAKER_03: So then there's some cost issues.
• 27:34
  SPEAKER_03: And I I'm gonna get to how much you charge for a build, but I'm not quite there yet.
• 27:39
  SPEAKER_03: I'm more interested in, for example, who's who who's paying the token bill for for the usage of the AI underpinning uh for the anthropic or for the open AI juice, for lack of a better word.
• 27:55
  SPEAKER_00: Yeah, yeah.
• 27:56
  SPEAKER_00: In the build phase, we we cover that cost as we build it.
• 28:00
  SPEAKER_00: And then once we've passed on the solution, we're still there to manage, maintain, and update it.
• 28:05
  SPEAKER_00: But that cost, we we build it within their own uh accounts and everything.
• 28:10
  SPEAKER_00: So that that token cost is passed on.
• 28:14
  SPEAKER_00: But we are always focused on minimizing token cost.
• 28:18
  SPEAKER_00: For example, you don't need God to write your emails.
• 28:21
  SPEAKER_00: So you don't need Fable Five to be writing your emails.
• 28:25
  SPEAKER_00: A model like uh haiku or Sonnet or whatever might be a better fit to write emails.
• 28:31
  SPEAKER_00: So that's just an example of every agent we build.
• 28:35
  SPEAKER_00: We'll make sure that we're minimizing token costs as much as possible.
• 28:38
  SPEAKER_03: Who owns the code after you implement a system?
• 28:44
  SPEAKER_00: Our client always owns the code.
• 28:45
  SPEAKER_00: So everything we build is proprietary to them and they they own it.
• 28:51
  SPEAKER_00: Again, if they don't have technical knowledge, we're still there to manage, maintain, and update it.
• 28:56
  SPEAKER_00: But anything we build is is owned by the client.
• 28:59
  SPEAKER_00: They paid us for it.
• 29:00
  SPEAKER_03: All right.
• 29:01
  SPEAKER_03: That all right, so now we get to the really good question.
• 29:04
  SPEAKER_03: How much does it cost to hire David to build a system in a law firm?
• 29:09
  SPEAKER_00: I mean, you you mentioned the the law firm earlier, uh Curtin, I forget the Kirkland Ellis.
• 29:15
  SPEAKER_00: Yeah, yeah.
• 29:16
  SPEAKER_00: So so half a billion is it sounds good enough to me.
• 29:19
  SPEAKER_00: I don't I'm no, I mean, our our ethos is uh going back to David, we we really believe in in helping small business.
• 29:29
  SPEAKER_00: And so we try to be as affordable as possible.
• 29:32
  SPEAKER_00: Um, starting with the fitting, our fittings, depending on the size of a firm, are anywhere from$500 up to$5,000 one time.
• 29:40
  SPEAKER_00: And then as far as builds go, it really just depends on what we identify to build in that phase.
• 29:47
  SPEAKER_00: But most of our builds are anywhere from two, maybe up to$10K per month over a three to six month period.
• 29:54
  SPEAKER_00: And then it goes way down from there.
• 29:56
  SPEAKER_00: The managed maintenance and updates are anywhere from 500 to 1,000 or even less than that sometimes per month.
• 30:02
  SPEAKER_00: Really just depending on what's identified to be built, the complexity of it, how many agents, all that stuff.
• 30:09
  SPEAKER_03: Could you without, I mean, you could name them if you want, but could you describe your last really satisfied law firm customer?
• 30:19
  SPEAKER_03: Who were they?
• 30:21
  SPEAKER_03: How many lawyers?
• 30:22
  SPEAKER_03: What was their practice area?
• 30:24
  SPEAKER_03: Where were they located?
• 30:25
  SPEAKER_03: Were they brick and mortar?
• 30:27
  SPEAKER_03: Were they AI native?
• 30:29
  SPEAKER_00: Yeah, yeah.
• 30:30
  SPEAKER_00: Um so we built uh an AI paralegal, which Heather, we should connect after because it'd be interesting to talk about all the your paralegal stuff.
• 30:40
  SPEAKER_00: But uh, we built an AI paralegal for a law firm here in Utah that was a relatively small firm, I would say 10 people, and basically their paralegal was retiring.
• 30:52
  SPEAKER_00: So she was leaving, they needed to fill that gap.
• 30:55
  SPEAKER_00: They were the ones that thought of AI as like, hey, maybe we could have AI do this.
• 31:00
  SPEAKER_00: So they came to us, we built this AI paralegal, and it's been really successful in the sense that going back to interacting with the client where they want to, where they're most comfortable or used to interacting.
• 31:13
  SPEAKER_00: This guy loves iMessage.
• 31:15
  SPEAKER_00: He messages all of his employees on iMessage.
• 31:17
  SPEAKER_00: He's like, I'm not going to be in Microsoft Teams.
• 31:20
  SPEAKER_00: If you need to get a hold of me, text me.
• 31:22
  SPEAKER_00: So he just texts this AI paralegal.
• 31:25
  SPEAKER_00: It executes on a number of tasks for him, including managing his email and his schedule, reviewing documents, helping with client intake.
• 31:33
  SPEAKER_00: So that's been really successful.
• 31:35
  SPEAKER_00: We started with that.
• 31:36
  SPEAKER_00: Now we've built them a voice agent as well.
• 31:39
  SPEAKER_00: The next thing we're working on for them, which is kind of interesting, is um, which I would say is a little more outside the box, but this guy kind of is thinking that way now that he's been experiencing AI, is building an avatar of himself.
• 31:52
  SPEAKER_00: Um, you've probably seen these before where it looks like Jerry and it's recorded my movements and even hand movements and mouth movement.
• 32:01
  SPEAKER_00: It's recorded my voice, and then it can take uh a video of himself so that he can have more interactions with his clients um without almost like cloning himself, right?
• 32:12
  SPEAKER_00: So he can be uh doing multiple things at once.
• 32:14
  SPEAKER_00: So we're still we're still building that right now.
• 32:17
  SPEAKER_00: It hasn't deployed, but I'll I'll update you on how it goes once we uh try it out.
• 32:22
  SPEAKER_03: Every company these days has like a little box on their website.
• 32:26
  SPEAKER_03: Talk to AI, and uh you have the chance to ask questions, and AI gives you half complete answers, and uh it it proceeds like this.
• 32:36
  SPEAKER_03: But this way, instead of having a chatbot give half complete answers, the avatar of the lawyer could actually do the same thing.
• 32:43
  SPEAKER_03: Do you think clients be more or less satisfied with that?
• 32:47
  SPEAKER_00: We'll we'll find out.
• 32:48
  SPEAKER_00: Um, I do think something I always tell clients is AI is the worst it's ever going to be, and it gets better every day.
• 32:55
  SPEAKER_00: So what voice agents could do six months ago is drastically different than what they can do now or or how conversational they are.
• 33:05
  SPEAKER_00: And that's the same with AI avatars looking more and more and more real.
• 33:09
  SPEAKER_00: And I would say, unless you're like deep in the weeds, like us three are, you would look at a lot of these AI videos and you'd be like, I don't even, I would I wouldn't even know.
• 33:17
  SPEAKER_00: We notice the little indicators, but the the average day person isn't noticing a lot of that stuff.
• 33:24
  SPEAKER_00: I do think it depends on the the type of law you're in.
• 33:27
  SPEAKER_00: Uh I wouldn't recommend it for more sensitive type cases, but this guy's doing a lot of transactional stuff, and I think because of that, he's open to testing this out to see if he can uh scale uh more efficiently.
• 33:42
  SPEAKER_02: That uh that's what I was gonna ask.
• 33:44
  SPEAKER_02: What area of law did that Utah firm do?
• 33:46
  SPEAKER_02: That would be really interesting to know.
• 33:49
  SPEAKER_00: Yeah, yeah.
• 33:50
  SPEAKER_00: So transactional, like estate planning, um, lease agreements, some of that kind of stuff.
• 33:56
  SPEAKER_03: I'm getting the feeling that most of the uh AI tools that we're seeing are trying to fill that space.
• 34:07
  SPEAKER_03: And, you know, spell book it was probably the first to market for doing that.
• 34:13
  SPEAKER_03: But so many of them say, well, we take your company or your firm's playbook and we use that to create these agents to make an NDA or a construction agreement or an asset purchase agreement or uh, you know, a merger and acquisition type suite of documents.
• 34:34
  SPEAKER_03: And that seems to be the highest level with litigation maybe a little bit further, a little bit behind.
• 34:43
  SPEAKER_03: Is that what you're seeing?
• 34:44
  SPEAKER_00: Yeah, yeah.
• 34:45
  SPEAKER_00: I think uh the that space is a space hungry for solutions right now because Harvey and Lagora don't really fit the uh the their firm size or their firm operations.
• 34:59
  SPEAKER_00: So it's definitely a space that that has a lot of hunger for it.
• 35:03
  SPEAKER_00: And I think it's a space too where there are a lot of firms we talk to, especially being here in Utah, where you actually don't have to be a licensed attorney to own a law firm.
• 35:13
  SPEAKER_00: Um there's a lot of people that are just looking to maximize their margins and AI helps to do to do that effectively.
• 35:20
  SPEAKER_00: And it helps with valuations.
• 35:22
  SPEAKER_00: If people are ever looking to exit, they if they have a proprietary piece of software that is trained on their specific firm, that that can help with with an evaluate with evaluation.
• 35:34
  SPEAKER_03: Jerry, before we move into the next segment of our show, do you have a sound bite that you want to give to our listeners so that they can really have a good sense of David AI?
• 35:48
  SPEAKER_03: I think I think you've covered the waterfront, frankly, but I would love it if you have a good soundbite for us.
• 35:54
  SPEAKER_00: I would say no matter your size, the David should have the same tools and resources to go up against the Goliaths.
• 36:01
  SPEAKER_00: And what's nice is while it might cost a little more up front, it won't cost that that much forever.
• 36:06
  SPEAKER_00: You own it, it's proprietary to you, so it's going to perform as great as possible all the time instead of you being dependent on the tool you purchased off the shelf actually doing the task that you're hoping it's going to do.
• 36:21
  SPEAKER_00: You can make sure that it does the task that you want it to do.
• 36:26
  SPEAKER_03: I do think that that really articulated your your mission.
• 36:30
  SPEAKER_03: So now we're gonna transition to our practice signal segment.
• 36:36
  SPEAKER_03: And this is where we hear from lawyers in the wild who have real problems and how maybe AI can help with those problems.
• 36:45
  SPEAKER_03: I set up a system.
• 36:46
  SPEAKER_03: Um this is a tease for season three.
• 36:51
  SPEAKER_03: Heather doesn't even know this yet.
• 36:54
  SPEAKER_03: Tease for season three is users and builders, which is how I see the AI world evolving.
• 37:03
  SPEAKER_03: And I'll leave it at that.
• 37:05
  SPEAKER_03: But as a I'm a builder all the way, and I've set up a system that's not yet complete to have AI scrape certain subreddits to help me find and summarize good practice signals.
• 37:17
  SPEAKER_03: So this is the first one that I've found using this system.
• 37:21
  SPEAKER_03: And the title is the insurance defense building trap, and it comes from Reddit.
• 37:26
  SPEAKER_03: From now on, they're all going to come from Reddit.
• 37:28
  SPEAKER_03: Unless I see unless I see something amazing.
• 37:30
  SPEAKER_03: Because you know, the anonymity of Reddit tends to make people a little bit more open with their problems.
• 37:37
  SPEAKER_00: Yeah.
• 37:38
  SPEAKER_03: So an insurance defense attorney describes routinely losing billable time because insurance carrier billing guidelines reject entries for activities such as conferring with senior attorneys, delegating work to assistants, reviewing large volumes of medical records, internal collaboration.
• 38:01
  SPEAKER_03: Now, the work, they gotta do that work.
• 38:03
  SPEAKER_03: And it's ethically appropriate.
• 38:05
  SPEAKER_03: But, you know, these insurance companies, they they don't like to pay.
• 38:10
  SPEAKER_03: So how would you use AI to address this problem?
• 38:18
  SPEAKER_00: So to make sure I understood the problem, because I I feel like there was a few moving parts there.
• 38:24
  SPEAKER_00: So the attorney is spending a lot of billable hours because they're having to basically coordinate between multiple different employees and multiple different tasks to then go to the insurance and try to get insurance to actually pay.
• 38:39
  SPEAKER_03: My it's very much like a problem that medical offices have.
• 38:45
  SPEAKER_03: You may go into the doctor's office and have the doctor look at two or three things, and the doctor or the nurse practitioner or the medical assistant or whomever you're dealing with has to apply the correct codes so that Medicare or the insurance companies will recognize them as legitimate codes for service and reimburse.
• 39:08
  SPEAKER_03: Now we're seeing that in a law office where apparently the top levels of the lawyers they haven't really developed a protocol within the firm for translating these kind of services into an insurance company acceptable language.
• 39:32
  SPEAKER_03: Okay.
• 39:32
  SPEAKER_03: I mean, I could very easily see this something coming that that you might you might start seeing this on insurance defense companies.
• 39:41
  SPEAKER_02: Seems like a perfect thing for David to be hired for.
• 39:44
  SPEAKER_00: Right.
• 39:45
  SPEAKER_00: Yeah, thanks.
• 39:45
  SPEAKER_00: Thanks for my idea.
• 39:46
  SPEAKER_00: I'm gonna I'm gonna run some ads now at uh these types of lawyers.
• 39:49
  SPEAKER_00: I mean, this this is cut and dry like a workflow or task that AI excels at.
• 39:57
  SPEAKER_00: It doesn't sleep, right?
• 39:59
  SPEAKER_00: It's really great at recon following a process, gathering information, documenting that information, and then passing it to the relevant parties.
• 40:09
  SPEAKER_00: And then again, since it doesn't sleep and it doesn't have emotion, it's just going to keep following up until it gets what it needs.
• 40:16
  SPEAKER_00: And that way, this attorney, for example, can continue to focus on other high-leverage activities that are going to help grow the firm while the AI handles that entire process that you mentioned without him having to sit there and and do a lot of that uh manual and tedious labor.
• 40:35
  SPEAKER_03: So, Heather, what do you think?
• 40:37
  SPEAKER_03: Do you think that there's a solution for this?
• 40:39
  SPEAKER_02: Well, I do think that this sounds like a perfect situation for Jerry's company to come in and build the solution for the firm because I do believe that there is an AI solution for this.
• 40:51
  SPEAKER_02: And I think that it would be a custom build or a very AI competent attorney or staff member in that office building it.
• 41:03
  SPEAKER_02: We know that we just interviewed a firm owner who did something very similar, not in that space, but in the bankruptcy space that built multiple skills to address, you know, needs and pain points in their office.
• 41:16
  SPEAKER_02: And insurance defense, I know exactly what they're talking about because I briefly worked in that area, not uh a really fun area of law.
• 41:24
  SPEAKER_02: But even as a pure legal, they want you to bill six to seven hours per day.
• 41:30
  SPEAKER_02: And the massive caveat to that is you have to use the exact right language when you're entering your time entries, or the insurance company will not pay.
• 41:41
  SPEAKER_03: You see, I I think this is tailor-made for AI.
• 41:46
  SPEAKER_03: But you know, and and this is the problem, and this is why I I brought this up at the beginning, Jerry, talking about resistance to onboarding, because I mean, lawyers are so busy uh day in and day out to have to add that extra labor burden on top of it to prepare for maybe buying some software.
• 42:08
  SPEAKER_03: I get it.
• 42:09
  SPEAKER_03: But what what the law firms have to do is they've got to create a protocol of acceptable I was gonna say coding acceptable time records.
• 42:20
  SPEAKER_03: And they have to somehow bake that into their AI.
• 42:23
  SPEAKER_03: And frankly, every day a lawyer should upload their time records to this AI skill or module or whatever we want to call it, and then have AI output a uh insurance company-friendly translation of the timesheet.
• 42:41
  SPEAKER_03: I mean, uh AI is built for that kind of process.
• 42:45
  SPEAKER_03: I mean, could you what what would what would be involved in implementing something like that for a law firm?
• 42:51
  SPEAKER_00: Yeah, yeah.
• 42:52
  SPEAKER_00: It it it could be as simple, like Heather said, as building skills um that can help accomplish that.
• 42:58
  SPEAKER_00: That's still on that reactive side of AI where you have to call those skills.
• 43:04
  SPEAKER_00: Programming an agent to just follow this exact workflow, you can set it to work on a loop where it just consistently does that.
• 43:11
  SPEAKER_00: Or even when you start doing those tasks, it starts tracking it.
• 43:16
  SPEAKER_00: So then it can take it, like you said, put it in acceptable language and then send it off.
• 43:21
  SPEAKER_00: If you give it the right parameters for the language, it'll do that language the way it's supposed to be done each time, as opposed to human who maybe gets tired or maybe doesn't stay up to date on the trends.
• 43:33
  SPEAKER_00: If it does change, AI can help avoid those pitfalls.
• 43:37
  SPEAKER_02: Yeah, this is something that could be fed into your AI.
• 43:43
  SPEAKER_02: This is this is a language in your time entry.
• 43:46
  SPEAKER_02: This is a very specific pain point.
• 43:47
  SPEAKER_02: And it is you have to have specific language in these time entries.
• 43:51
  SPEAKER_02: So I feel like with an enterprise level tool, you could go and upload a year's worth of time entries from every Staff member in that firm that is doing billable hours that's being sent to insurance companies so that you have the examples of correct language.
• 44:08
  SPEAKER_02: And then you could also put in a restricted list of phrases and tasks, et cetera, that cannot be billed, you know, in any of these codes.
• 44:19
  SPEAKER_02: So that's going to be things such as admin anything administrative, for instance, cannot be billed.
• 44:24
  SPEAKER_02: There, there's just so many things.
• 44:26
  SPEAKER_02: But it seems like that would be a very easy to fix, but an extraordinarily time-consuming process to get set up.
• 44:34
  SPEAKER_00: Well, a a fun, a fun fact for the listeners is what Heather just described is called machine learning.
• 44:42
  SPEAKER_00: It's basically where it's taking years and years of data and it's training itself on what's acceptable and what's not.
• 44:49
  SPEAKER_00: And then another part of machine learning is reinforcement learning.
• 44:53
  SPEAKER_00: So while AI out of the box, we would hope it just does it right every single time.
• 44:57
  SPEAKER_00: And this is a little bit speaking to the pain point, Ron, of there is some time investment when you build a new AI tool or an AI agent that you have to treat it almost like an employee when it starts out.
• 45:09
  SPEAKER_00: Like, hey, this is like a junior level employee.
• 45:11
  SPEAKER_00: We got to train it, but we we have it in place.
• 45:14
  SPEAKER_00: We're going to train it.
• 45:15
  SPEAKER_00: And that's called reinforcement learning, which is a subset of machine learning, which is basically you tell it almost like a dog, like, okay, hey, you did this, it this was correct or acceptable.
• 45:26
  SPEAKER_00: Here's a treat.
• 45:27
  SPEAKER_00: Oh, this was incorrect, never do that again.
• 45:29
  SPEAKER_00: Kind of like Heather said, you can't bill for administrative hours.
• 45:32
  SPEAKER_00: So it's like, hey, you you included the administrative hours, never do that again.
• 45:37
  SPEAKER_00: Okay.
• 45:37
  SPEAKER_00: That's that machine learning.
• 45:39
  SPEAKER_00: And so yeah, it does take a little time, but then once you've trained the model, it uh it's going to save all that time and all that headache for the months and years to come.
• 45:47
  SPEAKER_02: And then build, and then build an AI review agent to review every time entry before the billing is sent to the insurance agents, insurance agency.
• 45:58
  SPEAKER_02: That way you don't get as many sendbacks because that is the pain point for insurance defense firms is that every month, once that billing goes out, they're getting a whole lot back that needs to be reworded, redone, or just is not going to get paid because it was not logged correctly.
• 46:15
  SPEAKER_03: You know, what I'm seeing is the development of a billing module within an AI environment so that the attorneys, you know, when I started practicing, I handwrote my billing records.
• 46:30
  SPEAKER_03: By the time I was done, I used a product called Time Solve, which is not a super popular product, but it worked really well for me.
• 46:39
  SPEAKER_03: Uh now I see billing really being captured in an AI framework where the lawyer or whoever the timekeeper is types the billing information into an AI chatbot.
• 46:54
  SPEAKER_03: Doesn't have to be using Fable, could be using really the lowest level of uh model that you have, because yeah, it's not a high processing function.
• 47:04
  SPEAKER_03: And then it's spit it, I like to use the word spit out for output, but output's the more professional.
• 47:11
  SPEAKER_03: Yeah, I like industry friendly.
• 47:14
  SPEAKER_03: It spits out your time records in the appropriate way.
• 47:17
  SPEAKER_03: I I see that as something that David AI will be offering to its customers very soon.
• 47:24
  SPEAKER_02: Could David AI go in and, for instance, in your case management software for a firm that uses Clio, for instance, go in and build that right inside the system, an AI agent that does all of those things for an insurance defense firm?
• 47:37
  SPEAKER_00: Yeah, yeah.
• 47:38
  SPEAKER_00: With Clio, it's it's possible.
• 47:40
  SPEAKER_00: Uh it does depend on the softwares they're using.
• 47:43
  SPEAKER_00: Going back to the fitting, we look at the whole tech stack.
• 47:46
  SPEAKER_00: Every software has different API documentation.
• 47:50
  SPEAKER_00: An API, you can think of it as like a key into a software.
• 47:54
  SPEAKER_00: So for us to build an agent that can go into a software, we need the key to the door, which would be the API.
• 48:01
  SPEAKER_00: There are workarounds.
• 48:03
  SPEAKER_00: Um, there's things called like a browser automation.
• 48:05
  SPEAKER_00: Um, they just they don't work as well as an API key.
• 48:09
  SPEAKER_00: So Clio, for example, is um one that has good API documentation.
• 48:13
  SPEAKER_00: So, but it just depends on what a firm's tech stack is.
• 48:17
  SPEAKER_00: And but generally most modern softwares have API documentation so that you can have an agent that goes into it and executes on tasks.
• 48:27
  SPEAKER_03: You know, as a comment, once upon a time, I would have heard the phrase API documentation and thought runs, we're on Jetson's level automatically.
• 48:37
  SPEAKER_03: But I don't know that that's true anymore.
• 48:40
  SPEAKER_03: I think that the interconnectivity of different tools for different purposes is really going to be mainstream in the legal field.
• 48:50
  SPEAKER_03: Uh credits to Claude for coming out with its 22 connectors uh earlier in the spring.
• 48:58
  SPEAKER_03: Uh, I do see that as something that we're going to be seeing more and more of, which brings me to the Flintstones Simpsons Jetsons segment.
• 49:10
  SPEAKER_03: I see Heather have a little smirk, admiring aptitude in uh transitioning to this section.
• 49:18
  SPEAKER_03: So this is our segment where in season one, we would try to give tips for different attorneys at different levels of technology to help them advance into higher levels of AI fluency.
• 49:33
  SPEAKER_03: Uh for season two, we're looking at builds, at ways of for each level of attorney AI proficiency to build something within their practices that would be useful.
• 49:48
  SPEAKER_03: So over the past three weeks, we've been building something that may not have seemed terribly glamorous.
• 49:54
  SPEAKER_03: First, we identified our firm's knowledge assets.
• 49:58
  SPEAKER_03: Then, second week, we organized them.
• 50:01
  SPEAKER_03: Last week, we cleaned them, keeping only the versions that truly represent how we practice today.
• 50:08
  SPEAKER_03: Now we reach the point where, like, please, please tell us what the AI is gonna do.
• 50:14
  SPEAKER_03: We're gonna make those assets AI ready.
• 50:16
  SPEAKER_03: I didn't say upload them to AI.
• 50:18
  SPEAKER_03: We're not there yet.
• 50:20
  SPEAKER_03: This week's build step is to standardize them.
• 50:23
  SPEAKER_03: That means we're gonna make it easy for both humans and AI to understand what we're looking at.
• 50:29
  SPEAKER_03: If you're a Flintstones lawyer, open one of your best forms.
• 50:33
  SPEAKER_03: At the top, write a short description of what it is, when you use it, and why it's your preferred version.
• 50:40
  SPEAKER_03: If there are blanks that require judgment, make a note explaining what belongs there.
• 50:45
  SPEAKER_03: You're giving context, not just a document.
• 50:49
  SPEAKER_03: If you're a Simpsons lawyer, take your client questions and turn them into a consistent format.
• 50:55
  SPEAKER_03: Each question should have a clear title, followed by a plain English answer.
• 51:00
  SPEAKER_03: If there are exceptions or common misunderstandings, add those as notes.
• 51:05
  SPEAKER_03: Think about how you'd explain it to a brand new associate on their first day.
• 51:10
  SPEAKER_03: And finally, if you're a Jetsons lawyer, look at your systems.
• 51:14
  SPEAKER_03: Instead of writing them as one long paragraph, break them into clear sequential steps.
• 51:21
  SPEAKER_03: Start with the objective, then list the normal workflow, and finally, note any common decision points.
• 51:28
  SPEAKER_03: You're documenting the process the way you actually want it followed.
• 51:33
  SPEAKER_03: Without even trying, you're beginning to transform your firm's experience into structured knowledge.
• 51:40
  SPEAKER_03: Next week, we're going to introduce a tool that makes it even easier, which is markdown.
• 51:46
  SPEAKER_03: Jerry, what do you think?
• 51:47
  SPEAKER_03: Do you think that these are good approaches for building?
• 51:51
  SPEAKER_00: That was fantastic.
• 51:53
  SPEAKER_00: I couldn't have laid it out better.
• 51:54
  SPEAKER_00: And I think that that's a great way for lawyers, especially ones who are resistant or new to AI, to think about it.
• 52:01
  SPEAKER_00: Start with the Flintstone step, then get to the Simpsons step, and then uh get to the Jetsons.
• 52:06
  SPEAKER_00: I think how you laid that out is fantastic.
• 52:09
  SPEAKER_00: One thing I would add is that I think sometimes you can get analysis paralysis.
• 52:14
  SPEAKER_00: AI is built really well to understand natural language prompting, um, which is basically just talk to it as if it's a coworker or a friend.
• 52:25
  SPEAKER_00: And instead of having to feel like, oh, I have to document this perfectly, you can tell it all of your thoughts, you can tell it the process, and it actually can understand that really well to either train an agent or build the document to train the agent, kind of like you laid out in that Flintstones to Jetson situation.
• 52:43
  SPEAKER_00: So I I love how you laid that out.
• 52:45
  SPEAKER_00: I I might have to s steal uh the analogy.
• 52:48
  SPEAKER_03: Uh listen, you cannot steal that which is freely given.
• 52:52
  SPEAKER_00: All right.
• 52:53
  SPEAKER_03: Heather, what do you think about this week's Flintstone Simpsons Jetson segment?
• 52:56
  SPEAKER_02: I think it's exciting.
• 52:58
  SPEAKER_02: It's exciting that we get to find out what you're gonna do with it soon.
• 53:02
  SPEAKER_03: I I'm excited myself.
• 53:04
  SPEAKER_03: All right.
• 53:04
  SPEAKER_03: Well, listen, this was really a great episode.
• 53:06
  SPEAKER_03: Jerry, you are our first vendor guest, and and I and I think we'll be able to do more of these because I I I'm hoping we added value for our listeners.
• 53:16
  SPEAKER_00: Yeah.
• 53:17
  SPEAKER_00: Thank you for having me on.
• 53:18
  SPEAKER_00: The this was awesome.
• 53:19
  SPEAKER_03: All right, well, this has been AI Tools for Practicing Lawyers.
• 53:25
  SPEAKER_03: I hope you all have a great week.
• 53:27
  SPEAKER_02: Thank you.
• 53:28
  SPEAKER_02: Thank you, Jerry.
• 53:29
  SPEAKER_03: Bye.
• 53:30
  SPEAKER_03: Thank you, Jerry.
• 53:33
  SPEAKER_01: That's it for today's episode of AI Tools for Practicing Lawyers.
• 53:37
  SPEAKER_01: Thanks for listening.
• 53:39
  SPEAKER_01: We'll see you next time.