Law Firm Owners: Is Your Team Putting Client Data at Risk With AI?

People

Companies

[00:00] The four mandatory rules of the AI First Law Firm in this episode of Law Firm Growth Accelerator
[00:06] I sit down with Eli Gill, Chief Technology Officer of SMB team, an inventor of AI workforce
[00:13] pro, the leading agente AI for law firms in the US right now. And we're going to break down
[00:21] four of the most critical things that 99% of lawyers are missing as it pertains to AI.
[00:27] We're going to start with the importance of multi-modal flexibility.
[00:32] We're going to transition into something absolutely critical, which is security.
[00:39] And how most law firm owners who are using these consumer subscriptions to AI are actually in breach of client confidentiality right now as we speak.
[00:49] Then we're going to transition into how to set up integrations in your law firm the right way so that your AI use isn't just using chat AI
[00:59] But instead a genetic AI that can log into your law firm systems and actually get worked on and last but not least
[01:05] We're going to talk about the customizable and pre built skills and building a skill library for your firm on specific workflows
[01:13] workflows, taking the way your firm does things,
[01:16] in handling a case, systemizing it,
[01:19] and turning it into a repeatable skill,
[01:21] so that the next time that workflow pops up,
[01:24] you can get that skill and test on very quickly.
[01:27] One thing I really want you to focus on,
[01:29] which we're gonna hit early in this episode,
[01:31] is that multi-model flexibility.
[01:34] And this is something that nobody is hitting on right now.
[01:37] And there's really only one way to solve it,
[01:39] and that's through AI Workforce Pro.
[01:41] but you do not want to be a victim to any one of these large language models, whether that's
[01:47] Claude, Gemini, or ChatchyBT.
[01:49] If you build all of your firm systems in one of these models and then one of those models
[01:54] go down, we've seen it happen.
[01:55] We've seen anthropic and Claude crash.
[01:58] You need access to multiple models under one interface.
[02:02] And separately, you don't know which large language model is going to roll out the next
[02:07] big breakthrough feature.
[02:08] So you need access to all of them in one centralized platform.
[02:12] So what I'm going to offer you today for being here at the right place at the right time
[02:16] for this episode is an AI kickstart.
[02:20] And our AI kickstart is going to be we place a fractional CTO in your law firm for two weeks.
[02:26] We give you an AI opportunity snapshot on your firm and we're going to give you 14 days
[02:32] access inside of AI workforce pro along with 500,000 AI use credits, which is the way that
[02:38] we measure token usage on our end.
[02:40] And you're going to get two AI implementation calls with our team.
[02:45] And we'll build at least one wired integration for you completely free.
[02:49] So if you want to learn more about that, go to smbtm.com slash AI kickstart to build AI in your law firm right now.
[02:58] Please understand this has been the highest demand thing we've ever rolled out in the history of our company.
[03:02] So I apologize if we don't have capacity when you go to smbtm.com slash AI kickstart.
[03:08] Since you're here, I want to offer that to you now before we dive into the episode.
[03:11] So without further ado, let's dive right in.
[03:14] Welcome to another episode of Law From Growth Accelerator.
[03:17] Today we're going to be breaking down the talk of the town, the thing that everyone is waking up
[03:24] in cold sweats in the middle of the night.
[03:26] They're just frantically panting and running around in circles, screaming alone in their bathrooms
[03:33] to themselves in front of the mirror, screaming the word A, I, to themselves. And then looking
[03:41] at themselves and staring at themselves awkwardly in the mirror until an answer appears
[03:48] on what to do with A, I. So we have none other than the one and only E, I, GILED today,
[03:53] who's going to answer the question that we're all asking ourselves alone in our bathrooms
[04:00] in front of the mirror every morning.
[04:02] So what's the answer, Eli?
[04:03] I don't know if I'd be doing this in front of a mirror
[04:06] in a bathroom, but I do know a lot of folks
[04:10] in the business world including attorneys
[04:12] are really concerned about AI and how to appropriately use AI.
[04:17] It's an evolving landscape.
[04:19] We're still trying to figure it out.
[04:21] But I do think AI is here to stay.
[04:24] I think it's going to keep improving.
[04:25] And I think those businesses that use AI
[04:28] will out compete those that don't.
[04:30] I love it.
[04:31] Um, so first thing we want to talk about is,
[04:33] dive and write in, why don't you want to be tied to anyone,
[04:40] LLM or AI search engine as attorneys call it. Why, why do I not?
[04:44] Let's say I go and build let's say I decide I'm going to go all in on
[04:48] Claude. Why is that a risk to integrate and tie all of your stuff
[04:54] just to Claude? Yeah, that down for us.
[04:58] Yeah, for sure.
[04:59] It's definitely a risk right now.
[05:00] I mean, this landscape is constantly changing.
[05:03] One week, Cloud might be the best.
[05:05] I anthropic the company that makes Cloud.
[05:07] They might be the absolute best in the market.
[05:09] They might have more integrations than everybody else.
[05:12] They might have cooler features, things that revolutionize people's day to day jobs until
[05:17] the next week when they don't.
[05:19] And what we're finding from people is they do exactly what we say they go out and they spend
[05:23] money they invest in building integrations, doing development work, sticking with one platform.
[05:28] And then they end up still chasing these other platforms when they hear about the latest release from OpenAI,
[05:34] or some of the other model providers out there, and they realize that suddenly,
[05:38] it could cause not the best anymore, and they want the best.
[05:41] And not only that, but each of these models are better at certain tasks.
[05:45] So, again, you might find that clouds really good in certain business contexts,
[05:50] whereas maybe Google Gemini is a little bit better at searching the web.
[05:54] So, number one, right, you're always keep in up with the rat race.
[05:59] That's really difficult to do.
[06:00] Number two, these models do have better use cases in certain areas.
[06:05] And then I think number three is just, you know, whenever there's that latest announcement
[06:09] of some new model out there, you see the same thing, these providers have outages.
[06:14] And if you're relying on this in the middle of the day to do mission critical work and
[06:18] all of a sudden, oops, sorry, anthropics models are having problems today.
[06:22] That's a real hit to some organizations these days.
[06:25] Yeah, I didn't even think about that.
[06:28] I thought about more the AI race, like the biggest risk, I mean, the biggest risk is security
[06:35] risk which we're going to hit on, you know, is firms not being protected with client
[06:39] confidentiality on the consumer facing versions of these AI tools, which we're going to hit
[06:43] on next.
[06:44] The LLM dependency is something I wasn't even thinking of originally, but it makes so much sense.
[06:55] Like, I don't think in the history of the world.
[06:59] I don't think we've ever seen and I don't think we will ever see a competition of epic proportions.
[07:05] Like we're going to see between Chatchee BT, Claude, Gemini and whatever the LL pops up.
[07:10] up, grock, you know, whatever picks up momentum next, right? I don't think there will ever,
[07:16] I mean, I think this is probably going to be the most fierce race we may ever witness in
[07:21] the history of any technology race. With the biggest budgets, you know, most of these are
[07:28] trillion dollar, you know, or near trillion dollar market caps, with unlimited budgets, the ability
[07:34] It raised literally, I would say, not just billions in debt, but probably a trillion in debt.
[07:41] I mean, it's like absolutely mind boggling.
[07:45] So to think you're going to go and beat the frontier models is, I think, insane.
[07:51] I mean, I think the only way you beat it is through specialization is by going deep in a
[07:57] vertical, like we've done with legal.
[07:59] But more importantly, my concept is let these people race and get on a platform like
[08:07] AI Workforce Pro, what we've built, the only one in the industry that plugs into all three
[08:12] of them, so that if a great feature gets rolled out by chat to be teed next week, and
[08:18] you've already tied all your stuff in the cloud, all your integrations, all your skills, all
[08:22] your workflows, your teams using it, and let's say there's an outage, then you're screwed.
[08:28] If you're in a centralized platform agnostic system that taps into the LLMs like A-A-R-Course
[08:34] Pro, then you are protected from one LLM going down, one LLM rolling out a new feature,
[08:42] and I'm able to leverage the best of the models for my tasks while knowing I have two backup
[08:50] plans if one of them goes down or there's an outage.
[08:53] Yeah, absolutely.
[08:54] You've seen this from the attorneys who have adopted our platform.
[08:56] One, I think even this week I had somebody who was sending me email, hey, what's going on?
[09:02] And it just happened to be that anthropic was having a little bit of a service degradation.
[09:06] And it was easy as telling him to switch to another model provider.
[09:10] Back up, all of his workflows were working again.
[09:13] Super happy.
[09:14] And, you know, same thing, it's really nice to be able to say, okay, the latest OpenAI models
[09:20] out and it's available to you the next day.
[09:22] you don't have to go subscribe to it,
[09:24] you don't have to move your workflows.
[09:25] Just you can test it out, and if you like it,
[09:27] start using it, or you can stick with the models
[09:29] that you're using.
[09:30] And the other thing is some of these firm owners
[09:33] that I've been talking to, they separately log into
[09:37] Gemini, Chatchy, BT, and Claude with three separate
[09:39] browser tabs.
[09:40] Different browser tabs, different logins, you know,
[09:43] half the firm doesn't even have a, you know,
[09:46] a team plan or an enterprise plan,
[09:48] so they're all on just different personal accounts,
[09:49] which also risk.
[09:51] but that's risk. But what more so is when you're doing mission critical legal research work and if you put all of your weight in one LLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLLL
[10:00] response to your prompt or to your legal research prompt,
[10:06] what if the other LLM gives you different information?
[10:09] Comparing and contrasting against multiple LLMs is already what people do.
[10:14] They just do it manually with multiple tabs open.
[10:17] So one of the functions that I love,
[10:19] and I think is the way the future, is you're able to ask one question
[10:23] and have three LLMs answer, chat to BTGM,
[10:27] And I am glad, all answered the same question.
[10:29] You compare and contrast what's different,
[10:31] what's the same, how did each of the responses differ,
[10:35] what's the actual truth?
[10:36] I think your chance of getting the right answer
[10:40] to your prompt goes up exponentially.
[10:42] If you're fact checking it across three different
[10:45] of the frontier model.
[10:46] Yeah, we call that second opinion mode.
[10:48] And we allow our users to pick up
[10:51] the four different models at a time.
[10:53] And it's just as easy as you described.
[10:54] can take a document and say, you know, I'm having difficulty reading this, help me summarize this,
[11:00] or help you to research in a certain area, and just kick off second opinion mode, and, you know,
[11:04] right away they get not only full on outputs for me to do those models, but then they get that sort of
[11:09] meta output that tells them, here's the summary, here's what the model's agree, here's what the model's
[11:12] discrete, here's some things you might want to look into. Yeah, and it's in one chat interface.
[11:16] It's all in one chat interface, all in one chat. I can have all three of my responses in one chat.
[11:20] So if you want to get 500,000 free AI use credits in Air Workforce Pro, go to SMBT.com,
[11:27] Book of Call with our team.
[11:28] We'd love to have you test out the platform.
[11:31] The second thing I want to talk about, which is way more urgent than that, is the security
[11:35] concern.
[11:36] So I've read this recently and I want to make sure that this is your understanding of it,
[11:41] too.
[11:42] But when an employee or the owner of a law firm uses a $20 a month consumer AI account,
[11:50] Standard terms allow third party human contractors to review flagged prompts to improve quality.
[11:58] Now, if that is true, then using the traditional consumer account with private client data,
[12:12] would technically be breaking, ABA model rule 1.6 on confidentiality and your law firm could be exposed.
[12:28] Quick interruption. If you own a law firm doing 500k to 10 million and you want to grow without adding more chaos to your life,
[12:37] I want you to join us for our law firm, growth, acceleration, boot camp, September 16th and 17th.
[12:46] It's completely free and virtual, and over two days, we're breaking down the four systems
[12:52] behind the fastest growing law firms. Lead generation intake, team, and profit. Go to www.smbtem.com slash
[13:01] podcast to grab your free spot. I'll see you there.
[13:08] Yeah, I mean obviously I'm not a attorney and I'm not a member of the bar, but I can't say that
[13:13] in some of these
[13:16] different models and different model providers it is always worth reading their terms of service and what data is retained and what data is
[13:22] shown to either other models for training and improvement or to
[13:27] to their technology team to, like you said, improve those model responses in the future.
[13:33] It's very much something that everybody should be concerned about is sort of the privacy aspect.
[13:38] I will say this typically in the free versions of any of these.
[13:43] I would be really wary, really cautious.
[13:47] Almost all the free versions say that the data that you put in there into those can be used for future training.
[13:52] and there are a lot of anecdotes out there of people putting data into that free model and then all of a sudden it's exposed to somebody else and they did not intend to that to happen.
[14:00] So, we really care for the free versions. The cheaper sort of like consumer grade, as you were discussing, those do have some concerns as well.
[14:11] And I know what a lot of attorneys are looking for is something called zero data retention as much as possible.
[14:16] So you do want to look into what the terms of service read and if they're capable things like zero data retention where the models really just
[14:24] Using the data as it comes in and then after that response is complete there's no data actually saved anywhere so it's very much worth reading into those terms.
[14:33] So my question is if I'm an attorney and I
[14:37] I'm exposed right now because my teams using individual accounts multiple different LLMs multiple different AI search engines
[14:46] They're using private client information in it.
[14:49] One way of solving that is getting on an enterprise level subscription with one of these companies,
[14:56] the constraint, however, for that, and the research I've done,
[15:00] is most of these enterprise level subscriptions require,
[15:05] I think it's around 150 minimum users.
[15:08] Yeah. Typically, these enterprise-grade plans are the ones that include all the security features that a lot of attorneys are looking for.
[15:15] for. And the thing is not everybody has access to those plans to your point. They might say,
[15:22] okay, if you're three person firm, you can buy this plan, but you're going to be paying for 150
[15:27] users worth of usage and that's an exorbitant cost. So yes, if you want the best security,
[15:35] you're typically going to be pushed into those enterprise plans, but those plans are going to include
[15:39] enterprise costs. Yeah. So talk to me about how can a firm get access to enterprise level
[15:48] security without paying enterprise level pricing? For sure. I mean, what you're generally
[15:54] going to want to do is find a platform like a workforce pro, right? That's that's what we
[15:57] off. That was a leading question. Yes. A workforce pro is a tool that's designed to sort of leverage
[16:04] that enterprise-grade security for the smaller to
[16:08] been sized business. So we offer our plan at what we think is
[16:12] much better rate in terms of cost, but we still build that
[16:16] security directly into our platform.
[16:17] Yeah. So long story short, we in AI workforce pro, we
[16:24] buy the enterprise level, quote unquote, subscription or
[16:27] commercial agreement with those three LLMs, then our clients
[16:33] get access to that even if they have much, much, much less than 150 users.
[16:37] Right.
[16:38] Right.
[16:39] So we have these commercial agreements where we make sure that all of those protections
[16:41] are in place, there's zero data retention, there's no training.
[16:44] People aren't looking at these chats.
[16:46] And then we're the one who's able to provide that exposure to the attorney, where they're
[16:52] interfacing with our platform.
[16:55] It's still getting that security.
[16:56] Yeah.
[16:57] So what exactly do these enterprise agreements say?
[17:00] Like, what words do they use and don't they use?
[17:04] Because these companies are very,
[17:06] they have very careful legal departments, right?
[17:08] So I'm sure they don't,
[17:10] they don't promise things, right?
[17:13] Maybe they do promise something.
[17:14] But like, what have you seen in these enterprise level
[17:17] agreements that we have with the Chattu GT Gemini and Cloud?
[17:20] And are they basically all the same for all three of them?
[17:23] Like, or is there differences in each of them?
[17:25] Yeah, sure.
[17:26] There's commonalities.
[17:28] I think the three things that I tend to see trends around,
[17:33] one is in the intellectual property who owns the outputs.
[17:36] And the general trend within the industry is
[17:38] if you're the one interacting with the model,
[17:40] whatever outputs the model produces are yours to keep.
[17:43] So it's something to keep in mind.
[17:44] Can you say that one more time?
[17:45] Yeah, for sure.
[17:46] So like the intellectual property that's produced.
[17:49] So if you're talking to a model and it produces,
[17:51] maybe some document, who actually owns the intellectual property
[17:54] of that document.
[17:55] And the general trend is that the person who inputed the need for that document is the one who owned it,
[18:01] not the model provider.
[18:03] So again, if you're asking for it, for it to write a novel, I guess for example,
[18:07] and it writes a novel, you would technically be the owner of that novel.
[18:10] I got it.
[18:11] Okay.
[18:12] So that's one trend.
[18:13] Another is around what's actually used for training.
[18:17] It's really important, right?
[18:18] You don't want your data potentially exposed.
[18:20] So usually inside these enterprise plans, it does say,
[18:23] None of that data is retained for training.
[18:26] Oh, wow.
[18:27] So put that in writing.
[18:28] It do typically say in an enterprise plan,
[18:31] we're not going to use your model for training.
[18:33] Wow.
[18:34] Now again, your mileage is going to vary.
[18:36] It's definitely read all of these terms of service.
[18:39] And then last is this concept of zero data retention.
[18:42] Okay.
[18:43] This is kind of a complex technical term.
[18:45] Now obviously data has to be read into their system
[18:49] and processed and read back out.
[18:51] but the question becomes how long is that data actually inside their system?
[18:55] So there's very technical rules for how long it can be there.
[18:58] Because it obviously has to be there for a period of time to process everything.
[19:02] So after it processes, it's kind of like the system wipes itself.
[19:08] Now again this doesn't apply to every single chat and it gets it doesn't apply to every single
[19:13] agreement, but a lot of these have provisions around how you can have zero data retention
[19:18] chats with the model provider.
[19:20] Okay, is that the temporary chat model?
[19:23] Are you saying that the variables?
[19:24] Yeah, it's a separate, but it's in it.
[19:25] It's very similar to this idea that you can have a chat and at the end of the chat, there's no data retained anywhere.
[19:31] I got it.
[19:32] Okay.
[19:33] So basically, if a law firm owner wants to go to bed at night feeling more secure and like their clients information isn't
[19:45] breaching their client confidentiality requirement.
[19:47] Either need to buy an enterprise level subscription
[19:52] and pay the price of that at a very high user number.
[19:56] Or they choose something like AIwork.
[20:00] or pro that buys that enterprise subscription on behalf of them
[20:03] because we have a bulk number of clients on it
[20:05] and then gives you access to that enterprise grade
[20:08] level security at a lower cost.
[20:11] Is that basically the two outcomes?
[20:14] Right.
[20:15] Again, it depends on what you're using a model for.
[20:17] Right.
[20:18] If you're asking for recipes for dinner that night,
[20:20] it's certainly fine to go out and buy a $20 consumer grade
[20:23] platform.
[20:24] But if you're doing things like again, complicated,
[20:26] legal research, drafting documents, things that include
[20:29] like your day-to-day operations as you practice,
[20:31] you should be taking security seriously.
[20:33] And to your point, either going to be paying a lot to go
[20:36] to one of these large providers who are selling to enterprises
[20:39] and installing very, very high costs software.
[20:42] Or, you know, you're going to find a software like ours
[20:44] that is designed for a small-to-mid-sized law firm
[20:46] and has appropriate price tags.
[20:49] And the security.
[20:50] The other thing is when you go, just think about it,
[20:53] the first point we said is not being tied to any one model.
[20:59] Let's, if you did buy an enterprise level of Claude, now you're on the hook for 150.
[21:04] Seas Wood of Claude goes out and Wood of chat, you BT comes out with something better.
[21:08] Like you're still stuck with the same problem.
[21:10] If you buy one of these enterprise level agreements,
[21:13] you really to get the same level of protection.
[21:15] You have to buy enterprise agreements with all three, integrate with all three.
[21:19] That, that's kind of the funniest thing is, and I've heard this from attorneys before they've come to a workforce pro.
[21:26] they'll even say, yeah, I'm on this higher tier package with maybe anthropic.
[21:30] Yeah. But I still go to Google every single day and ask it questions.
[21:33] So, like, really what's the point of buying all the security in one?
[21:37] If you find yourself in the bad habit of going to the other anyways. So,
[21:40] do your point why not have one central platform where you never have to leave?
[21:46] You can still, if you want a second opinion from another model, you can get that while you're still maintaining security.
[21:52] Yeah. Are you looking to implement AI in your law firm now? Not tomorrow right now. Well, I want you to take advantage of our AI kickstart, okay?
[22:01] We are going to place a fractional CTO in your law firm for two weeks. We're going to give your firm a full AI opportunity snapshot on exactly how your firm can make money by implementing AI in your firm the right way.
[22:16] You're going to get 14 days free access to Aero Workforce Pro along with 500,000 AI use credits.
[22:23] And you're going to get two AI implementation calls with our team and we'll set up your first
[22:28] integration completely free. If you want to take advantage of this now go to smbtm.com slash
[22:34] AI kickstart. That's smbtm.com slash AI kickstart. Man, what a world we're in. All right, so next thing
[22:42] thing I want to go through is the systems integrations.
[22:49] So this is what I think, the thing that separates the firms that really get work done
[22:57] with AI and those that still use it as really just a chat box.
[23:02] The firms that are using AI agents to get work done, the minimum viable survival input
[23:09] of that is their integrations.
[23:10] because if you don't integrate with stuff, there's no way to get worked on.
[23:13] There's no way for the AI to exit the system, go into one of your systems and get
[23:19] worked on it.
[23:20] There's just no way without integrations.
[23:22] And I was actually on a call with about 100 attorneys yesterday and I asked them, how
[23:29] many of you have integrated with more than five or ten systems.
[23:33] I think I said 10.
[23:34] And I don't think anyone said yes.
[23:37] And you're in cross-mort.
[23:38] I mean, it's you need to integrate with your email, your document storage, Google Drive, whatever
[23:47] you're calendar, your call tracking system. So we use Fathom on our Zoom, but there's many others.
[23:56] You need to integrate with your QuickBooks or your financial system. You need to integrate with your
[24:03] law firms CRM, your case management system, Slack, whatever you use for internal team communication,
[24:13] what am I missing? What are the major buckets? I think you've covered most of them.
[24:17] Yeah. But there are incidentals, right? And I think that's the thing where we're learning
[24:21] even a small business has an incredible number of technology tools. Yeah. And those are just the
[24:27] ones I named or just like the everyday like everyone uses stuff like that, right? Right.
[24:31] you know, right. And this is a real cost people pay and technology like this idea that I'm
[24:38] switching my context, my brain context, which we in this window and that window, it takes time,
[24:44] it takes like cognitive load, you have to think about, okay, now I'm not looking at this screen,
[24:49] anymore, I'm looking at this screen and I have to think about this, that takes time, right? You can't
[24:55] instantly flip sometimes to the next task that you have to do. So integrating all these things
[25:00] and a one-spot allows you to reduce a little bit of your cognitive load right.
[25:04] You're not really thinking so hard about switching between windows.
[25:07] You can ask, hey, AI, right?
[25:10] What calls it out today? What are the important things that have to do?
[25:13] And not only that, but can you update my calendar to take any of the outputs from those calls and schedule my next set of meetings, right?
[25:21] Ask the AI.
[25:22] Can you take the fathom of the call I just had with this client and can you upload it into my CRM and then connect it?
[25:28] under their account. Exactly. Right. So all of those things are now just one window. You're
[25:33] interactive with one system. Yeah. And that's where the power comes in. AI goes out and talks to each
[25:39] of those systems. Does those tasks for you and takes a lot of that load away from you? I think that's
[25:44] the most important thing is like for people, I still think 95% of attorneys are not, they don't even
[25:50] know, they have an experience of power of a gentic AI getting work done in your systems. Just based on
[25:57] on the question I asked yesterday, it's got to be that way.
[26:00] I mean, and that's the only way you get worked on.
[26:03] So the integrations aspect is like the most
[26:10] commission critical thing to get worked on,
[26:12] but at the same time, you don't want to be beholden to,
[26:17] let's say you go through the process setting up
[26:19] all these integrations.
[26:20] You get all the everything integrated and then
[26:23] chat to VT comes out with something new.
[26:27] and you did it all in just cloud, so that platform flexibility comes back to the importance
[26:33] of that, because if you're going to go through all the blood sweat and tears of integrating
[26:36] with everything, you might as well have to only do that once rather than every time a model
[26:42] comes out with something new, I got to switch my models, switch my integrations.
[26:45] But the next thing is, and this is something I love for you to educate us on, difference between
[26:50] off the shelf, integrations and like custom integrations. Because I know in our platform and in other
[26:58] platforms, they'll have like integrations that pop up. So those are off the shelf integrations.
[27:04] Is that the way to put that? Yeah. So one click, integrations where if I'm going to integrate with my
[27:08] Gmail, like in Workforce Per or Cloud, it's like three clicks to do that. Right. It pops up a little
[27:13] login bar. I got a login to my Gmail. I click a button. It's integrated. It's very quick. But then
[27:19] Then there's certain legal CRMs in case management systems that like Clio's off the shelf
[27:24] because they have an open API.
[27:26] I believe that's why it's off the shelf.
[27:28] I don't know.
[27:29] It's sort of, yeah.
[27:30] Yeah.
[27:31] A little bit of a technical answer, but generally, yes, there's some that are just, you know, you
[27:34] might go into claw to our product and you'll see just like a button you push and you're suddenly
[27:39] connected.
[27:40] But there are, I think, where you're going with this, it's like there's more cost of integrations.
[27:43] So some products.
[27:46] You can't just click a button and you're attached.
[27:48] And that's where real developer effort comes in.
[27:51] And this is where we also help our clients.
[27:52] So we have these fractional CTOs,
[27:55] and these AI implementation managers who work with our attorneys,
[27:59] and they try to understand every single system they use.
[28:02] And they try to build these integrations no matter how easy
[28:04] they might be, to how difficult.
[28:06] Some of these more difficult integrations,
[28:08] we have to do actual development work in the background,
[28:11] to try to make it an easy experience for them.
[28:14] So what are the examples of ones that aren't off the shelf?
[28:18] like once that like you need a little bit of customization like I think we're talking about
[28:21] smoke ball or something. It's something else where it's like you have to like put in special
[28:26] credentials or like you've got to like. Yeah it just really depends. There's no like one size fits
[28:33] all or like one app that I can point to and say this is a difficult integration. Yeah it really
[28:40] just depends on what the that application provider gives in terms of documentation. Got it.
[28:46] they support in terms of authentication. That's a really important part of it's the authentication
[28:50] and authorization. So the user experienced to get logged into that app and to integrate it.
[28:54] That makes a big difference as well. Sometimes it's even maybe an easier integration on the surface.
[29:02] But when you really start to dig into what the attorneys needs are, we find out, well,
[29:07] we could use the off the shelf integration, but that's really not going to be best for these
[29:12] folks because of some reason they might have. So we're going to have to custom build it ourselves anyways.
[29:17] Just depends. Okay. So integrations, integrations, integrations for sure. Very mission critical.
[29:24] The other one is legal skills. So like we we tried to break this down yesterday.
[29:34] I think we've got a rename skills, because it seems like average lawyer is not very clear on what a skill is, because we've had to explain it a lot.
[29:45] And so I think this is the fourth pillar and very important pillar.
[29:54] How do we take the repetitive workflow?
[30:00] as in our business, turn them into a skill so that our special touch, our way of doing things
[30:06] can be systemized.
[30:07] And we had someone on our call yesterday give a great example on the demand letter.
[30:13] So I want to break that down.
[30:14] This is why I think these companies that are offering AI demand letters are going to be in trouble
[30:20] because the customization that you can do with a tool like ours or directly in one
[30:26] one of the LLMs with skills is, why would you pay 1,000 bucks for a demand letter?
[30:32] I mean, but what was shared and I'd love to hear your thoughts on this was, I thought
[30:36] it was the best explanation we've heard on what a skill is, just because it came from a
[30:41] lawyer who works with us.
[30:43] And what he said was, you know, when I created demand letter, I hash it out with the AI.
[30:51] And I go back for, I say, you know, here's the client situation.
[30:56] Here's access to my CRM, grab all this information, here's the facts of the case, here's
[31:01] what happened in court, here's the discovery, here's all the data.
[31:07] Create me a demand letter for this situation.
[31:11] AI creates it, comes back to you says, here's your demand letter.
[31:16] And what he said was he fought with the AI for 20 minutes.
[31:18] No, I'd rather have the intro be like this, I want you to be more aggressive here.
[31:22] I want this, I want your outro to be like this.
[31:25] I want dollar amounts here, blah, blah, blah, blah, all this back and forth.
[31:29] And at the end of 20, 25 minutes, he ends up with the perfect demand letter.
[31:35] Now most attorneys would just move on with their day and say, oh, my demand letter's done.
[31:39] Now I've q-a-d it.
[31:41] I've hashed it out, cleaned it up, and smooth on to my next thing.
[31:45] Well, he did, instead, was he said, turn this into a skill.
[31:49] And then it created his custom demand letter skill.
[31:54] And that demand letter skill.
[31:55] Now the next time he needs to create a demand letter,
[31:57] it's already trained on all that back and forth.
[32:00] The little special touches he wanted to put in
[32:02] at his little chef's kiss on how he does demand letters.
[32:06] So now next time he has to do it,
[32:07] and he says, use demand letter skill to insert client,
[32:14] you know, create a demand letter for insert client name,
[32:16] go to the CRM, pull out all those data.
[32:18] Now it knows exactly your way of doing a demandlet.
[32:22] Yep.
[32:23] Yeah, I love that example that he gave yesterday,
[32:25] and I was trying to think about how to re-explain it.
[32:28] And I would almost call it like employee training, right?
[32:32] Like you write these employee training guides,
[32:34] how to do certain things in your company.
[32:36] Think of it as an employee training manual for the AI.
[32:39] Maybe that's just how we should start explaining it.
[32:41] That's all it is.
[32:42] So after you've spent that time,
[32:44] I really like how we ended it.
[32:46] I spent all that time hashing it out,
[32:48] but then I built the skill and I never had to spend that time again.
[32:50] Right, so it's just the way of training the AI
[32:53] to do whatever task you wanted to.
[32:55] And then after that you never have to spend that time again,
[32:57] but the other really important thing is,
[33:00] now the AI does it that way every time.
[33:02] So it really dials in the AI's accuracy
[33:05] of future responses.
[33:06] Another thing is, why do it based on an ad hoc event
[33:10] like that?
[33:11] If you want to scale the best results
[33:13] your firms ever produce for people,
[33:15] why don't you take the most successful demand
[33:17] letter, your firms ever put out in the history of your firm.
[33:21] And give that to the AI and say,
[33:23] derive the principles, the structure, the methodology
[33:26] used for this demand letter, and turn that into my template
[33:29] for future demand letters.
[33:31] Like, it doesn't have to, like, there's many,
[33:34] once you learn the power of a skill, you can approach a skill
[33:39] from any different ways.
[33:40] You can approach it through, I hash it out with AI for 30 minutes,
[33:43] or you can approach it from, this is the best case we've
[33:47] have ever handled in the history of our company.
[33:50] And I want you to turn this.
[33:51] Could you tell it to turn it into five skills?
[33:55] Like if you were to give all the context of the best case,
[33:58] you ever handled, where everything went perfect.
[34:00] And every part of the case management process,
[34:03] could you have it create a skill for part one, part two,
[34:05] part three, part four, part five of the case?
[34:07] Yeah, and I think in like an AI engineer here.
[34:09] Yeah, this is kind of the advanced sort of prompt engineering.
[34:13] So over time, what you realize is that the more narrow an AI sort of focuses, the better.
[34:21] So I love that idea.
[34:23] You have some big complex tasks.
[34:26] Don't make that necessarily one skill.
[34:27] Make that like break it down.
[34:29] What are these sub tasks?
[34:30] And turn each of those into a skill.
[34:32] And those skills are just little specialized agents just doing that one specialized thing.
[34:37] And then you can actually create a sort of a master skill or a meta skill that sits on
[34:43] top of the other skills and says, okay, use all five.
[34:47] Use all five in this order.
[34:48] Yeah.
[34:49] And once skill 1's done, go to skill 2 and take the info XYZ from skill 1's skill 2.
[34:54] That's a genetic AI.
[34:56] That's an genetic workflow.
[34:58] Yeah.
[34:59] That is for sure.
[35:00] So that's taking, yeah, I've never actually heard a parent skill and then a set of child skills.
[35:07] Yeah.
[35:08] Yeah.
[35:09] For sure.
[35:10] We've set a few of those up for clients.
[35:12] It is very common for somebody to think like this one process that might take a human 45 steps to do.
[35:18] And you know, two hours, why is this just one skill for an AI?
[35:21] And the thing I also love is the legal specific nature.
[35:26] This is one of the competitive advantages of ours is we have what 46 prebuilt skills now somewhere in that.
[35:33] Yes.
[35:34] 46.
[35:36] And if I had to invent every skill from ground zero, that's a lot of work, a lot of thinking,
[35:43] but sometimes I also need to be, if I'm a law firmman on busy, I also want to know what kind
[35:49] of skills could exist.
[35:51] And that's why we have the whole list that you can get when you type the dollar sign in the
[35:54] chat, the whole list of all the skills that are already built.
[35:57] And I saw one of them in there, the job post builder.
[36:01] I used that for a legal event director role for ourselves.
[36:05] And we're not even a law firm and I use it.
[36:06] We use it for our own company.
[36:09] And it worked perfectly.
[36:11] And basically I said, use job builder skill to create me a job description.
[36:17] Now it's the way we write job descriptions.
[36:21] It's exactly the way we write them.
[36:22] It's a mission at the top.
[36:24] It's like all the little nuances that we teach in our coaching and fractional COO solutions
[36:29] around that.
[36:31] It was perfect.
[36:32] All I had to do is say use the skill to do it. Then I saw the financial skill and I thought
[36:38] The scheduled skill function
[36:40] So scheduled skill basically means you take a workflow and then you schedule the delivery of or outcome of that workflow
[36:47] on scheduled basis. So I saw the
[36:49] We have our QuickBooks integration. Yeah, and I'm actually starting this starting next Monday and the scheduled skill is
[36:57] Look at our financials from the last week compare it to our forecast
[37:01] So it integrates with Drive, which has our financial forecasts, and then it integrates with
[37:07] our QuickBooks, and on a Monday it says, here's your financial report, 9 a.m. every Monday, on where you ended last week,
[37:15] where you were off track, where you were ahead of pace. And that's scheduled. Now, every Monday, now doing that for inbox,
[37:22] summarizer, that's another one we have.
[37:27] Yeah, we do have a, yeah, we have a scale that helps people manage their inboxes.
[37:31] So inbox summarizer, I come my desk a Monday morning and it's set to me,
[37:34] sift them through 500 emails.
[37:36] I go to one email.
[37:39] Yeah, that comes in.
[37:40] That's already summarized everything.
[37:41] Yep.
[37:42] Right.
[37:43] It's amazing.
[37:44] Yeah.
[37:44] Yeah.
[37:45] And I mean, we've had people tell us in an air workforce pro.
[37:49] You can do this in other platforms.
[37:51] You know, you could create a skill and theory around these things.
[37:54] But people who have used skills, whether in our platform or not, I think in a Patrick literally said,
[38:02] how much time did you say he's saving per day?
[38:05] Two hours a day?
[38:06] Two hours a day?
[38:07] Every single day.
[38:08] Yeah.
[38:09] I mean, just think of that.
[38:10] Right.
[38:11] I know.
[38:12] That's 10 hours a week.
[38:13] Right.
[38:14] That's 500, 500 hours, 520 hours a year?
[38:18] Yeah.
[38:19] I mean, that's like, and that's just one skill.
[38:22] That's the promise of AI.
[38:23] The big promise of AI is I believe it's return on labor.
[38:28] That's the metric I've come down to is if it firms doing a million dollars a year and
[38:33] has five employees, and those five employees cost the firm $400,000 a year.
[38:42] Paying 80 grand per employee, that firm doing a million a year who has $400,000 in salaries
[38:50] is at a 2.5x return on labor.
[38:52] Meaning the firm's revenues is 2.5 times higher than how much they're paying on salaries.
[38:57] I think the metric that drives the return on investment of AI is return on labor.
[39:05] Because when those five employees can get twice the work done and or when those five employees have their low level
[39:14] admin tasks automated and therefore can focus on higher value money making work.
[39:20] Exactly. That firm that was doing 1 million with five employees can get to 1.5 or
[39:26] even 2 million with the same five employees. Right. That's what I think the promise of AI is.
[39:31] I think it's taking your return on labor from say 2.5 to 5x. Yeah. And I think if you do the
[39:39] math on that any amount of cost that you need to invest in AI implementation is going to
[39:46] pay all in comparison to because if I go from making 1 million to 2 million a year and my
[39:52] expenses stay flat. Even if I just increase my expenses by
[40:00] I don't even know, 200,000 a year on AI implementation.
[40:04] That would be a lot of money.
[40:05] That's like 18,000, 16,000 a month, right?
[40:10] Still, I would make 800 grand an additional profit,
[40:13] even if I spent 200 grand on AI.
[40:15] Yep.
[40:16] Yep.
[40:16] I mean, it is like for service-for-service-based industries
[40:22] that become tech-enabled, it's like,
[40:26] I keep trying to tell our clients and attorneys like,
[40:28] This is gross margin heaven for service businesses if you use it right right yeah
[40:34] The traditional formulas
[40:36] Completely busted right because we know that you know previously is very linear
[40:40] You'd have to add more employees and to your point
[40:42] This is not something people are thinking about in the future they're thinking about can I add like a small line item for AI
[40:49] Keep my employee count flat and continue to grow that top line revenue. Yep. Yep
[40:53] So we got today guys, this is changing the game.
[40:57] If you want any information, you want to get 500,000 free AI use credits for AirWorkforce
[41:03] Pro, get s&bteam.com and ask our team about that.
[41:09] And yeah, we're looking to change the world here.
[41:11] So I'll see you in the next episode.
[41:13] Let's say your firm is doing a million dollars a year right now and you have five employees.
[41:17] Okay?
[41:18] Those five employees cost you $400,000 a year in salary and those $400,000 in salaries
[41:22] in salaries is producing a million dollars a year in your firm.
[41:26] Well, what if AI could keep your salary cost the same,
[41:29] but take your firm's revenues from one to two million dollars a year
[41:33] through better efficiency and your team spending time on much more productive things?
[41:37] That's what AI can help do for your law firm.
[41:40] So if you want to implement AI in your law firm now, not tomorrow,
[41:44] then go to smbtem.com slash AI kickstart.
[41:47] Our AI kickstart is where we place a fractional CTO in your law firm for two weeks.
[41:51] that CTO will build an AI opportunity snapshot for you. You'll also get access to 14 days
[41:58] of AI Workforce Pro. AI Workforce Pro is going to come with 500,000 AI use credits that you can
[42:05] start testing in your firm today and you're going to get two AI implementation calls and your first
[42:11] integration setup completely free. So if you want to take advantage of this now go to smbtm.com slash
[42:16] A.I. Kickstarter.