Episode Transcript
[00:00:00] Speaker A: Welcome to the Relay, the legal show for personal injury law firm owners, presented by Lexamica, the number one attorney referral network. I'm your host, Gabriel Steeritz. My guest today is Jacob Mallerby, founder and CEO of mass tort ad agency formerly known as X Social Media. Now here's why I wanted Jacob on most people on the show. When we talk about AI, we're talking about what's coming and maybe theoretically how this stuff is being used. Jacob can talk about what already happened and how he's used it in his own company.
More than anyone else that I've ever met, Jacob is using AI And I mean, look, he's not a young whippersnapper, but he is the smartest, most cutting edge AI user that I've met and really interesting stuff that he's doing. And I think Jacob is a case study in what it looks like to push AI to the absolute boundary right now in real business use case, not just theoretically. And Jacob has the credibility. He's put more than a quarter of a billion dollars through Facebook for plaintiff firms over the last decade. Hundred plus mass torts. And today he's running the OPER with a very small team. And look before that, as the AI guy, he was the Facebook guy. He invented Facebook advertising for mass torts and man, he's got interesting stories. We're going to talk about how he sued Elon Musk and in my version, he won. So we've got a David and Glad story, a category that everyone loves to hate. And in production, answer to the question that every firm owner is asking right now, which is how far does AI go? Jacob, welcome to the Relay.
[00:01:27] Speaker B: Yeah, I appreciate it, Gabriel. Always fun to talk about AI for me right now and for all the people out there that are scared about AI My philosophy is AI will create more of the work you like and less of the work you don't like doing as lawyers. That's how I see it. I think there's going to be more lawyers out there helping all the underserved people because it's not going to cost as much to do a case anymore.
And just this week, actually I released a book free on my website, a Lawyer's Guide to AI. I actually show you how you can install your own AI inside your office to keep all the proprietary data inside your office. Don't trust the cloud. The cloud is not your friend with personal information.
So you want to keep it inside. And I teach you how to build a few things and that is.
[00:02:23] Speaker A: And did you write the book or Did AI write the book?
[00:02:25] Speaker B: AI and I wrote the book in about eight hours.
[00:02:29] Speaker A: So there it is.
[00:02:31] Speaker B: Yeah, so we did it together. And that's the whole point, right? You have AI do stuff for you with human oversight.
[00:02:41] Speaker A: Walk me through your journey with AI and maybe even a couple steps prior to that. So before Chat GPT launched, what was the headcount at your agency? What kind of roles did you have going on? Were you running efficiently? Did you have 400 people working for you? You've been really one of the staples in mass tort social media advertising for a very long time. And if we have time, we'll talk about your ex social media story. But where were you before AI and at what point did you start moving more into putting your, your, your agency on. On AI?
[00:03:17] Speaker B: Yeah, so we were about 20 people and now we're down to about five people. We do most of everything with AI today and AI keeps up with all the data, keeps all up with all the numbers that come out of Facebook. There's so many metrics that a human probably couldn't keep up with it. So AI keeps up with all the metrics for us. It tells us what we should shut down, what we should scale and what we should pause.
So everything today, down to images, down to compliance with SB 37 in California, for example, is run by, by AI at my company. But we didn't really start getting into building with AI until about February of this year when the AI agents came out through a product called Open open flaw, which was when I saw that you can dash my wife. I just left the building and I just dove died deep, deep into it. Because I was like, man, if we now have somebody that can actually do the actual work for us and not just ask them a question, I'm just fully in on that. And, and I dove into it for probably a month to get really, really deep. But I will say for anybody listening, you take 30 minutes, you ask Claude what you can do and what, what he can build for you. And it's that first 30 minute barrier you just gotta get over.
And I'm not technical, I'm not, I've never coded in my life. I've never done anything like that.
[00:05:04] Speaker A: So I'll ask the question that I think everyone is wondering immediately. Like we read the headlines, OpenClaw launched, everyone's excited about it. And then people in China start losing their life savings by giving OpenClaw their credit card. You're an interesting success case of open claw. And I think a lot of people have kind of Said, hey, that was really dangerous. We're not sure that that was ever a good idea.
You're running so much on it. Have you had close calls? And if not, what is it that you think you've been able to that have put those guardrails on it?
[00:05:37] Speaker B: So I was maybe on openclaw for the first three days of my life in working with AI and found out that it wasn't really built for industries, it wasn't really built for business with the safeguards and, and the security and all that. So I moved super fast over to what I would think is the perfect stack. First off, it's free to use, which is always great. It's called cloudflare.
It has the workers, it has the databases, the workers.
You can use the name worker as an agent and interchangeable.
And it's really a beautiful stack to work with together with Claude because Claude just makes the code, you put it on a worker and boom, you have an agent and now you can, now it can do real work for you.
So I moved super fast of openclaw, but openclaw was the aha moment for me that this could be done.
[00:06:36] Speaker A: Okay, so openclaw really cast a vision for you and then you found here's something that's safer, that I can control effectively to, to make work happen. And I, I don't know how familiar our listeners will be with the idea of deterministic and non deterministic workflows, but deterministic one, essentially step A, step B, step D. And it always works in the same way. That's like your kind of traditional software coding and workflows. And non deterministic is AI like you give it a prompt and it'll give you every single time it's going to give you a slightly different output. How much of what you're doing is you figured out how to get AI plugged into step by step workflows and how much of it is you just put a prompt into an AI agent and it magically does what you ask it to.
[00:07:18] Speaker B: I do it this way. I tell Claude or the AI what my idea is which, which what you call a prompt. I just call him my, my co founder and I just talk to him like, right. So I tell him my idea and then my favorite word is expand after my idea. And then he comes up with all these extra things that we could build to this idea to make it really, really good. And then he asked me, is that what you want to do? I said, let's go. And then he starts building and Then we have a product, I put it up, I look at it on the web to see what it looks like, what it's missing, what is wrong, what is right. And then we go back and we, we make changes until it's just like I want it to be. And all these ways that developers or coders, maybe like you, Gabriel, are thinking.
I never thought about things that way. I just had the vision of the product I needed to solve a problem.
And then Claude was that middleman that just got shit done for me, so to speak. Right. And so maybe it's a constraint to AI if you have that developer mindset, maybe it's better to just be full of good ideas or you have a lot of problems that you want to solve and you just go out and solve them.
[00:08:48] Speaker A: So you had this aha moment when you saw openclaw and what it could do. You went and said, okay, let me find good infrastructure that's not going to make me lose my life savings. What was the big aha moment in your business? What was the first big thing you were able to automate that was painstaking, tedious, frustrating in the past, and you were able to put AI on it.
[00:09:09] Speaker B: So first I had what I call a kindergarten project where I just wanted to make different agents to see what they could do. And I just call that a fun project to understand what my capabilities.
So I made this website called My Kiara, which is my character in Japanese, a little nod to my son that is all into anime right now. And I just made agents that there's an agent called Salt over there that you can take a picture of what's inside your fridge and it will tell you what to do for dinner, right? Like my little son could take a picture of the snake and it would tell him if it was a dangerous snake or not. Right? Like you could talk to these agents about different subjects. There is a biohacking agent, right, that can tell you about what kind of stack you should be on to optimize your health. Like just funny stuff, right? Yeah, so. So that was my first project, but then for business it was sorry.
[00:10:14] Speaker A: And you were doing that in Cloudflare. So you, you got Cloudflare and then you said, okay, let's do some zero stakes building here just to start into it. I love it because I, I've found the same thing to be true. Like I just, there's a sweet spot of I can never just go like learn something. I need to build, I need to learn by building. And the best first thing, I built an app Called Snow Day over, over the holidays when, when we were looking openclaw came out, I was like, I have to try this. And so I was like, cool. Like, let's go. Whole website, start to finish the app. The back end, all of has no actual value, but it was an end to end. I can package it up and say I'm done with this.
And I think what you're doing to anyone who's like, hey, how do I get started? That's the thing. Pick. I don't care what it is. A snake identifier. That's perfect. Assault in your fridge. Like, I don't care what it is. Like, you know, a dog walking tracking app. Like all of that's perfect. I love that.
[00:11:07] Speaker B: So build something fun, right? Build a website for your church group, right? Build. Build something for your kids scout class, right? Like just build a website. Just do something right? Like try it out. So I, that. So that was my first project then in my business, which is mass torts, we track 46, 45, 46 different mass torts. So I built a site called Tort intel that basically tracks it all for me. Updates with Chrome Jobs. A Chrome job is like an automatic update where you just tell the agent or the worker, go Fetch all these APIs and see if anything updates. So Tort intel now just. I actually haven't been on the website in a few months, but it just keeps updating all the mass torts. So that was my first real product where people were like, wow, okay, that's pretty cool because we spend a lot of time going to the courts to figure out what's going on. And here's a website that just tracks it. All right, so, so that was. That felt like a win. And then I was like, well that's great for a lot of lawyers that they can track all the mass charts, but let me build something for my company so I don't have to spend so much time in Facebook anymore with all the ads we are running. And so I built something called Watchdog that basically keeps up with all my ads and can make images for my ads. Can see if the image is good or bad and it will tell me if the image is good or bad after I'd spend some money on it. So that probably automated away about 80% of the manual labor in, in, in running ads.
[00:12:53] Speaker A: There's something to call out here and I think because a lot of people think AI is like this magic thing where you just, you go chat with Claude and they try that and then they're disappointed because the one shot, like, hey, I did, I tried one thing and it was pretty good, but then it doesn't have real actual value. What you're doing, Jacob, is using AI to solve for real problems.
I'm an, I'm a tech CEO now. My degree is in, in art in photography and design. And you have a design mindset which is, let's go define the problem. What, what is it that I want to solve for and then find the best way to do that. And that is a very different approach that I think a lot of people can learn from, which is like you said, you've got you and you keep raising the stakes. You went from personal project to zero downside public facing project with this tort intel AI. And by the way, I'm on the website now, it's working perfectly. Tort intel AI, it's still running. Like you're not even maintaining it. It's there, like, that's so cool. And then now you're going into real business process that has stakes, right? Like you shut off the campaign too early, that's actually bad. You can change budget, you can like, and maybe you're just in the monitoring phase, but that like stair stepping approach is like, that is a framework that is really important for people to understand as they listen to your journey on this.
[00:14:11] Speaker B: Yeah, I would think 80% or something is automated. Of course. We spend a month looking at what AI gave us as this should stay on, this should shut down and all that. How are the images working that AI makes? If the disclaimers right, is the spelling right?
Like AI can check the spelling on the image, right? It has vision, right? So you can see if there's a spelling mistake or hallucination in any kind. You just make all these checks and balances and then of course you monitor it as a human and with human judgment to see is it good, is it bad. And once you have rejected some images, AI learns from that, right? And then everything in Facebook right now is you gotta come to the problem from a different angle. So we made eight or nine different AI agents that make different style of images. You know, one is the passionate one, one is the journalist one. One is a luxury image we say, where it's like, it's just a really good looking image with all the numbers and all that in it. Like, and then you see what works the best. And it's kind of incredible, right? We had two people doing that job before keeping up with Facebook ads and now AI do it better than probably I could do it because it Looks at more data, it looks at more metrics, it even gives us a head up when our images start getting stale. And it looks at the frequency and the click through rate over time to see is the image getting stale, is it about time to change it out? I actually think I have a much better ad agency right now than I've ever had because of AI Sell.
[00:16:03] Speaker A: But again, I'll call out, it's. You say AI, it is AI inside of highly structured workflows that you already understood. And if you. And, and I think this does apply to personal injury mass tort law firms because you do have these highly structured workflows. Like you actually have the thing that Jacob's talking about.
But just going and throwing that into a Claude chat or chat GPT is not what we're talking about. We're saying, okay, create the workflow, get in cloudflare, whatever your, your AI harness is, and you can go do that research, you're smart. And then start to test your way toward it because you've, I'm sure you put Jacob, many guardrails on this early warning signals this is a deviation. The spin goes too high, goes too low, it goes too soon. Like, I'm sure you have all of those notifications set up such that nothing could just spin out of control without you paying attention.
[00:16:51] Speaker B: No, it's not adding money and it's, it's telling me what's wrong. And I'm actually clicking the button to say shut it off. Right.
When, when I make an image, I'm actually clicking the button to upload the image. Right. Like there's still human involved. It just does all the boring stuff. Right. That just takes time. But there's still the judgment call made by me or one of my people. Right. So, so you still have that in place.
[00:17:22] Speaker A: But like you said, you reduced headcount like specifically because of these workflows that you built. You no longer need someone to go do certain parts of the workflow.
[00:17:31] Speaker B: Yeah, yeah. And if we just take it one step further, where we went after that was there is so much public data that is available via APIs, wildfires, hailstorms, medical malpractice, employee law, asbestos. So much public data that you can all of a sudden pipe into a map and you can visually display that data and then you can overlay it with Facebook and say, I want to target these people that went through a hailstorm 12 months ago because there might be a first party insurance claim there. So all of a sudden the way that lawyers find clients changes because of AI because now we can put clients in front of them where we have a much higher probability that they have a case versus I'm just blanking in the whole state. Right. I know where that hailstorm happened. Right. So now I can just hit the hailstorm. So there's an ally in the AI for lawyers already using those things.
[00:18:38] Speaker A: And so you found real practical value in that piece as well, and running
[00:18:44] Speaker B: campaigns today with it for law firms. Right, because they see the value of not doing the shotgun approach. Right, like do the sniper. And you know, we know where these storms went through. NOAA has the data. It's free for everybody. All we have to do is put it on a map and make the connection between the map and Facebook to get the data over to Facebook so we can geotarget the hailstorm.
I mean, it's really nothing but a thing once you, like, get into it. I mean, it really isn't.
[00:19:18] Speaker A: So let me ask you this because I, this is, I think, going to be really interesting.
You, Jacob, you're the business owner and the CEO and you have people working for you. You personally went through this transformation from pre AI and maybe less hands on as a builder personally, although I don't know, we haven't asked that yet, to 100% builder in the weeds. Understanding the process, clearly that's worked for you. What's your advice to someone who's running a law firm with twenty, fifty, a hundred people? If I'm that owner, do I walk out of this saying, well, now I need to go through a personal journey of transformation, or is there a path, are there other paths that lead to the same end result, which is that you've massively increased the productivity and the efficiency of your business. And look, I know you. You're not just here spouting a bunch of BS about AI stuff like you actually walk it out, otherwise I wouldn't have had you on the podcast. Like, you actually are doing it. And you, all the stuff you're talking about is real. Like, you don't just, like, make things up. So for someone who's looking at you and saying, oh, wow, okay, so Jacob personally spent the last six months of his life, and I would say it's a personal transformation even more than a business one. Is that the necessary gate through which every owner must pass, or is there. Do you think that someone could delegate that transformation to someone else on their team?
[00:20:41] Speaker B: Yeah. So first off, I have a developer on my team, right. That develops. If you call my 800 number, you will talk to Miles, who is our AI assistant that will talk to you about everything that we do and set appointments in my calendar and all that. My developer builds that. Right. He used my brain that I had built up in Cloudflare, so it knows everything about the business, but my developer built that. So the question is, for a business owner running a law firm, the good thing about AI is he knows all the issues, he knows all the problems, and that's the first thing you have to identify. Where are people spending their time and what do they hate to do? That's the first thing to automate. Right? And once you understand, once you take that 30 minutes and just say, hey, I have a problem in my business, it is this. How can I fix it with a stack that is Cloudflare, for example, what do I need to build? And it's 30 minutes. And you will know so much more. You will have taken that first step.
And it's intimidating to begin with, but after the first 30 minutes, you go, like, I can do that, right? Like. Because it's really nothing but a thing once you get into it. And it's very exciting, obviously. I mean, I'm excited about.
[00:22:13] Speaker A: That's the title of the episode, It's Nothing but a Thing. AI with Jacob Mallerby. It's Nothing but a Thing.
[00:22:19] Speaker B: No, but it really isn't. Once you get into it, you find out that all these barriers and the coders and the developers and how they held the key to the kingdom that's gone right now. It's about who understands the problems and possibly the solutions will be AI to those problems.
[00:22:41] Speaker A: You are a walking case study on this, which is what I think is so interesting about it. Like, you are the guy who wasn't a coder, wasn't a programmer, dove into this, and then has risen out the other side with a more efficient, more productive business than you had before. And that's. The case study is not about. It's really not about AI. Right. It's about can. With today's technology and workflows, can you build a better, more efficient, more productive business that isn't just hype? So, look, I've. We've got a wrap in a couple minutes. Is there anything that you want to show or demo for us while. While we're here? Because I'd love to take a peek behind the curtain.
[00:23:16] Speaker B: I think I should show a little bit. Here is a good example.
A law firm reached out to me yesterday and said, hey, we do employment law.
I would like to have a call with you today.
So last night I sat on my phone and I built this, right?
This is public data of companies laying off people in California in the next 90 days because they have to report it. And that API is available to anybody.
So overnight on my phone I built this for the guy where we geo target where the location is. I can take it over to Facebook. I can go in here, I have everything that I need to say on Facebook on here. I can render an image here, get the images. It takes 60 seconds, so I won't bore you with that. But up here in the gallery you can see images that are generated, right?
And now I just feel like I built business in a box for this guy, right? Like he can take this, I can do the advertising for him, but he knows what to target.
[00:24:32] Speaker A: Yeah.
[00:24:33] Speaker B: So I love that here's my watchdog where we keep up with all the campaigns and basically we could click on this one.
So it shows us what our price is right now. It shows us the different ads what should fix a hook on this new ads running.
It shows us a gallery that is down right now. We can make images here with the different agents for different images that we want to use on, on the, on the ad and we can upload it straight to Facebook. We even have different image AI's here to, to make that image.
So that's another one, another one.
[00:25:27] Speaker A: And this is the workflow app that you were talking about earlier, Watchdog. It tells you this is what campaigns are performing.
[00:25:33] Speaker B: This is what does all our workflow for us is this one.
[00:25:37] Speaker A: And for people who haven't been in Facebook ads, like this is a much better Facebook ads interface is completely insane. And this is far, far better than that.
[00:25:46] Speaker B: Okay, so then another guy came to me, I'm going to switch to another tab, want to do asbestos. So we made him an intelligent kind of thing for asbestos so he can find out where is the incidents around United States asbestos meso cases come out of. This is again public data that you just have to pipe into a map again just to show them, hey, we know your business. We can help you target these people with your business. Right.
The last thing I want to show you is how we track our own website mass taught ad agency. Specifically I want to highlight where you can check where you are in the overview if you're being cited for different keywords people use.
And we track how many times the AI searches cite us inside for what keywords. And then of course we try and make it for all the keywords. But here's where we at right now.
And then the last thing I want to highlight just to get into the whole AI stuff.
If you go to my website and click Guide to AI Free book, start reading a lawyer's guide to AI. It will get you a good idea of what AI is. And I will, in the second half of the book, teach you how to put your own AI on your computer and build a few things.
So that's maybe a good way to get started, you know, I love it,
[00:27:36] Speaker A: Jacob, it's always fun talking to you.
You've built five new products since the last time we spoke and you're in a very. But like you said, it allows this level of customization. I'm sure you used to build like one pagers in Canva or whatever for a new client and now you're building live websites. And I think that we've seen so much of that as well, which is that the ability to customize to create bespoke useful tools is just completely different than it was before. And look, you don't have to become a Jacob Mallerby. I think a lot of people should become Jacobs. But what you shouldn't do is not get into it. Go get the book, download it, start building stuff. I think the coolest thing, the most exciting thing about all this, like we all get to be builders now. Like, that's awesome. Like, and I get that not everyone wants to be a builder, but you should. You should want to be hands on and creating things. It helps you understand your business better, the problems better. And heck, if you're going to go buy AI from a vendor, which isn't a bad thing, a lot of you, that's the best solution.
Building it yourself will let you understand what you're being sold. Right? Like, you just have no excuse for not putting your hands on the wheel and understanding what's actually going on. So Jacob, thank you so much for being on the show. I've really enjoyed this conversation.
[00:28:54] Speaker B: Enjoyed it too, Dave. Well, thank you.
