FiredUp! - The Startup Marketing Podcast
FiredUp! is the show for marketers working in early and late-stage startups. Each week, we walk through fresh strategies and tactics to build brand and drive demand for your startup. Featuring interviews with marketing leaders, our take on the latest trends, and practical tips about PR, content marketing and growth marketing, we promise plenty of signal with some noisy fun along the way.
FiredUp! is hosted by the team at startup marketing agency, Firebrand. Learn more at firebrand.marketing today.
FiredUp! - The Startup Marketing Podcast
Inside Marketing Decisions at a LegalTech AI Firm with Jared White
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Navigating early-stage startup marketing in a crowded, highly regulated market is one of the toughest trials a founder or marketing leader can face. From determining the right ideal customer profile to breaking through established, legacy competitors, seed-stage companies are forced to make high-stakes go-to-market decisions with limited runways and tight budgets.
On this episode of FiredUp!, we sat down with Jared White, CEO of Matey AI, to address the explicit pain point of startup marketing decisions in complex B2B sectors. We discuss how to navigate legacy competition, position your product around clear customer outcomes, leverage hyper-targeted ICP profiling, and optimize your engineering and marketing workflows in the age of generative AI. This week, episode 140 of the FiredUp! podcast is about marketing decisions at a LegalTech AI firm!
Download the Multiplier Marketing Megapack today. Exclusive offer for all our listeners — get all our Startup Guides in one go! Over 50 pages of advanced tips and advice that dive deep into content marketing, search advertising, and marketing attribution.
In this episode of the FiredUp! podcast, Jared White shares the importance of building trust with AI systems and actionable steps you can take right now to target the right audience for your product.
Jared White is a technologist, serial entrepreneur, and the CEO of Matey AI, an innovative software company using advanced language models to solve complex data challenges in legal tech and eDiscovery. Coming from a background in quantitative trading where precision and accuracy are paramount, Jared approaches software design with a relentless focus on data trust and hallucination-free AI outcomes. Under his leadership, Matey AI established a foothold in criminal defense and public defense agencies, providing powerful, accessible discovery tools for legal teams.
Jared and Morgan discuss:
- Don't Fear Entrenched Competitors: The presence of large legacy players in a market is often a strong indicator of demand and capital availability. Modern technological shifts (like foundational language models) allow early-stage startups to completely leapfrog legacy architectures.
- Lead Generation vs. Brand Building: At the seed stage, building an immediate, measurable sales pipeline must take priority over broad brand awareness campaigns. Focus on getting into the room, solving urgent pain points, and letting your product's performance build your brand trust organically.
- Frame Value Around Outcomes, Not Capacity: Avoid alienating potential buyers by over-advertising extreme technical capabilities. Instead of positioning your tool purely for massive, rare enterprise use cases, emphasize how it solves everyday, high-frequency problems affordably and efficiently.
- Use AI for Hyper-Targeted Lead Prioritization: Supercharge your go-to-market strategy by deploying AI to analyze CRM data and score your target personas. Profiling prospects based on background, specialization, and firm size allows small sales teams to focus outreach on buyers who are most likely to convert.
- Keep Core Teams Lean to Move Faster: Scaling headcount too quickly can introduce unnecessary logistical overhead. Leveraging modern AI development tools enables small, highly focused engineering and marketing teams to produce massive output while remaining agile.
Building a successful startup isn't about having a massive budget or a massive team—it's about making sharp, disciplined decisions and executing on real customer needs. Are you positioning your product to solve an immediate, acute pain point for your buyers today?
Thank you for listening! Tune in to all the episodes for practical tips on crushing your startup marketing goals. Don’t forget to follow, rate, and review the podcast, and tell us your key takeaways!
CONNECT WITH JARED WHITE:
CONNECT WITH FIREBRAND:
Firebrand is a startup marketing agency. We help tech startups secure outsized marketing outcomes on their path to growth.
X (Formerly Twitter)
5.5.2026
At a C-stage company, there are a lot of marketing decisions that need to be made. So today, we talk to the CEO of Matey AI, which is a legal tech company, all about their go-to-market and their decisions that they're making. Hello, everyone. Welcome to Fired Up, the podcast for marketers working in early and late stage startups. Hello, everyone. Welcome to Fired Up. My name is Morgan McClintic, and today I am joined by Jared White, who is the CEO of Matey AI. Jared, welcome to Fired Up. How are you?
Jared White:I'm doing well, thank you. Thank you for having me.
Morgan McLintic:It's great to have you here. You are in the legal tech space, specifically eDiscovery, which is highly competitive. There's a number of larger late stage companies in that space. In fact, eDiscovery was one of the earliest adopters of AI. And what made you decide to even target this sector in the first place.
Jared White:So wonderful question, and thank you again for having me. I really like to say candidly that legal found me more than I found legal, and it is true. There's lots of legacy players we call legacy players in this space, and they're one of the earliest usages of AI NLP type linguistic AI, if you will, and I tend to say that all of the things that they built for the application of AI for discovery, all of those things exist because language models did not yet exist. So I think you could entirely replace them and completely discard them, and no one would lose any sleep. And we would have much better outcome by using the modern version of what we know as as linguistic AI language models, foundational models. And I really, from my perspective, it wasn't that daunting that there were a bunch of legacy players because there's really no industry that has escaped software at this point. So everyone is going to have a bunch of legacy players, and why not legal as well to apply language models to it? That was
my thesis going into it:is that I shouldn't be scared of the fact that there were some entrenched players. In fact, that's probably a good indication that's a good market to be in. There's plenty of money to be made,
Morgan McLintic:right? And I guess traditionally in e discovery, it's the AI vendors were classic machine learning pre-gen AI vendors. And when I worked in that space, I remember there was a lot of sort of effort to make sure that this can stand up in court. Right, so you're using a computer to look through all the discovery documents, and then comes along Gen AI and language models, which are probabilistic, and I guess add an extra whole layer of sort of uncertainty to it. Did that sort of daunt you about getting over that bridge? Because I certainly remember getting talking about court cases to try and stand up machine learning, and here we are with that extra level of probabilistic,
Jared White: I think of it this:it really wasn't that daunting, to be honest with you. And perhaps I think the as a founder, like your best and your worst qualities are how optimistic you might be. So tend to overestimate probability of success and underestimate the difficulty level of problems. But but the notion of defensibility came up very early in, yeah, and I said, I like the analogy I literally gave to someone was like, yeah, we used to have wooden bridges too, and we would be concerned about driving trucks over them, but now we have concrete bridges that we don't have to worry about this with, and so here we are with the concrete. Let's build the new bridge and then not have to worry about this. I do think that it stands to reason that you would have these these questions raised about the existing or what was what I consider legacy ways of killing things like that. That's a very fair question to ask. Is should we trust this? I think it's actually a less fair question to ask about language models. Perhaps not like GPT 2.0 but definitely it's a fair question to ask about those early models, but as the models get better and better, we can trust them more, and people have more experience with them. And you don't have to spend this like inordinate effort to tune a tar model or a cow model to determine if something is relevant, and then still put a human in front of it. And what you're really just trying to do is like winnow down from some large number to a smaller number that's more reasonable. That's very fair. I think that at the end of the day, you know what we at Matey, what I say is that we're selling trust at scale, so you can trust our systems. And so, yeah, I think at the end of the day, we have better technology now, and that's the point of it, right?
Morgan McLintic:So it's a neat outcome and gradually. There's there are court cases that that use this technology and it becomes the norm because it's best practice and it works. Okay. Certainly.
Jared White:Yeah. By the way, humans make mistakes too.
Morgan McLintic:They do. That's right. So that's a good point. So you're going okay. I'm going to go into this space. There's been a technological shift. We can apply that to to this particular issue. Every sector's got incumbents. Okay. Fine. As a sort of seed stage company, you have to make decisions around who I'm going to target, who's my ICP, and in this instance, am I going to go for in-house, or am I going to target the law firms? Am I going to target small law firms? Am I going to target enterprise-scale law firms? I'm going to be on the plaintiff side or defense because those impact the sort of workflows that you're automating, if you like, how did you go around picking criminal defense as a niche?
Jared White:Great question. Again, whenever I go back to that answer, like legal found me more than I found. There was some individuals I was in contact with and became in partnership with early in as sort of design partners for this company, and some folks there had a lot of experience in the criminal defense space, and so we were introduced. Our very first contract was actually with a state government agency that provides public defense. So very rare startup gets this first contract with a government agency, right? I wouldn't say that I regret that at all. There's other challenges associated with servicing a government versus servicing an individual or a small small business. But end of the day, that was like I think of it this way: if I've got to build this much software to service like the broader entire market, the criminal defense space was a much smaller section than the software I needed to build. In particular, because you don't have like production requirements, for example, and there's also very little competition in the criminal defense space because they don't have any money.
Morgan McLintic:Right. My next question is like, okay, this is a niche, but it's often a niche that people wouldn't go on because they don't. It's not well funded.
Jared White:So, and again, to your earlier question, like, hey, are you daunted by existing players? It's like existing players are the ones indicating to you that where the money is, right? At some point, there are no untouched markets, and if there are untouched markets, they may be untouched for a
reason. I would say this though:like we found great traction in certain sections of the criminal defense space, and then others are not so great. How we found it is basically opportunistically, and I think that as a startup founder, you have to find your customer before you build your product. Gone are the days of building the product first, and then hoping that you eventually find a way to monetize it. Is
Morgan McLintic:there a story behind this? Then you identified the problem, and then because you've founded companies before, and so you're a technologist and not an attorney. Did you know how did you come across this this particular problem?
Jared White:In effect, the company I'm actually building is eventually much larger than just a legal company, and I knew that fundamentally. You know, whenever we started using language models, and everyone worries about this whole hallucination thing, and these, to your point, the probabilistic outcomes of the next token, the stochastic parrot, as it was said, those would become major problems to solve, and ultimately, none of us want a piece of software that we can't trust, and trust is everything. And I won't do this in my next startup, but in this particular startup, I still diluted enough to say I should solve the hard problem first. And so, what I set out to do was basically solve the hallucination problem on a petabyte scale dataset. And if I can get a petabyte of data into a system, I can ask it questions and know that there are no hallucinations in the outcome, and be able to have like a a way of proving that there are no hallucinations in the outcome. Then I would be able to apply this technology to essentially anything. And legal happens to be the first and most rigorous use case that we defined because it has least room for error. So that's that's why we got into the legal space first,
Morgan McLintic:right? And that's why you say that this company is going to apply beyond the legal space because having certainty in the outcome applies to lots of different applications of AI. But you still needed to know about the legal tech space, right? The legal space, and so early on, did you hire legal people and then teach them how to build a product, or did you hire engineers and then teach them about law? What was your approach to that?
Jared White:I'm definitely more interested in hiring engineers and teaching them about the law, because at the end of the day, this is a deep technology problem to solve, and you can your customers are essentially telling you what product they need if you're listening, right? And if you're asking that question, the sales and the go-to-market are the that's the side of the organization I believe that should be dictating essentially what the product team understands is the the feature set that or the the solutions that people will pay money for to solve their problems, and then turning this into an engineering exercise is something that we've been doing for generations at this point. And you never want to degrade that part of the process in mind. It's you can make do with first-time engineers on very early stages of the product, but whenever you have a large-scale problem, knew this would be a large dataset issue. Whenever you have like an accuracy problem and a trust problem that you're solving, definitely don't want someone who's learning on the job for that. I come from a background of quantitative trading. You lose money. That's like literally less dollars that you have tomorrow than you had today. Right? It matters immediately. Matters and not. A lot of room for error, so especially as competition increased over the years. So from my perspective, is like that's the one thing that we're never going to sacrifice, because engineers have been turning business problems into software for a really long time as part of the process. We don't have to worry about them knowing anything about the law to start off with.
Morgan McLintic:Okay, and then talk to us a little bit about the name Matey. It's obviously an unusual name in this space. How did you settle on that? And then, does it ever work against you with trial lawyers or in government procurement because it's a sort of more friendly kind of name?
Jared White:I think it has the opposite effect. I think some fun anecdotes I can't immediately share, but where folks, whenever they start to realize, whenever I tell them, yeah, it's we're pirate themed because it's nautically themed. Like this is maybe like a pirate would say, we have, for example, our flagship agent system is called Ahoy, and we've got some marketing around that. It's time that you had Ahoy moment. So it's really it's a lot of fun. I felt like I wanted to be part of a company that had a degree of levity around this the name, and nothing like truly esoteric. Like I was part of this one company called Alien Vault once, and what is that about? And I don't know. It was just like two fun names that they put together, I think. But I was like, okay, that's maybe a little bit too far in that direction. But I wanted something closer, and then I didn't have a name. One of the folks I was working with, whenever establishing the company, he actually never became part of the company, but he had bought the domain name, and
Unknown:it
Jared White:was Wordle of the Day. And I think April of 2023, there was like some day that was the Wordle of the Day, and that's where the name came from. And I and so then it worked for so many different ways because everyone was using the word copilot, and mate is something like you know it's a different word for that. And being someone who loves the water as much as I do, loves the ocean as much as I do, like it just I was very drawn natural. Yeah. Additionally, what I would say is to the earlier question about the applicability of this product to the legal field versus other fields, like legal mate happens to be the first mate in the ecosystem, if you will, and we can configure our application for other types of workflows and to do other types of things. Those might be another type of mate, sales mate, executive mate,
Morgan McLintic:right,
Jared White:so forth.
Morgan McLintic:Okay,
Jared White:it works for it works on three or four different dimensions, including being humorous to talk about, and but it's not so extreme. I think that you don't get credit for it in in a government procurement process. Literally had a government procurement process where we was like an RFP, and they really wanted us to win the deal, which we did, and they're like, "Yeah, maybe I'll just put into the RFP that like it has to be a pirate-themed company or something. So people ultimately will buy your product if they want, if
Morgan McLintic:they want, if they want. And I guess because it's not overly legal, it can apply outside when you expand into other sectors. So you're feeling okay. This levity and this sort of yeah, I like the copilot side. We can apply it elsewhere. And fortuitously, you had the domain name. I just think it's interesting the way that companies come up with their names, and you've post rationalized it, and it's working for you. So that's great. So we've got our proposition. We've got our target ICP. We're going after criminal defense. We've got our name. All right, now tell us a bit about your go-to-market approach. Obviously, founder-led sales at the beginning, but tell us about the approach. And then, did you try anything that just didn't work?
Jared White:Yes, we tried a million things that didn't work. Good lord, where to begin on that subject? I'll tell you. There are certain segments of the market that we found that are very interested in using a product like this. And one of the challenging parts I think of selling into law firms is that, for the most part, the computing systems that we've been selling for 80 years now have the promise has been you'll be able to save time. Lawyers have an many of them at least have an inherent desire not to save time, and I hear them say so. Why would I want fear affordable hours? It's a fair question if that's how you make your living. What I have seen is that these are all just my opinions. Lawyers tend to be more interested in buying products that they use in terms of the technology that they'll purchase than they're willing to buy products for subordinates, and the challenge has been in e-discovery. At least is that lawyers are essentially intermediated from the discovery process. However, there's a first-year associate or a paralegal or another company altogether that is doing that work. So why do they care? I just get my answers and then I do my work, and so lots of legal AI companies out there that are sold successfully to lawyers because the lawyers I'm going to be able to use this, and then I can go to my kid's soccer game. But I don't care; they're not going to theirs. I want more billable hours on their plate anyway. So that has been a huge problem that we had to work around and figure out avenues. Yeah. To navigate, and there was a lot of trial and error on that, and I think we have cracked that nut. What I will say about this is that at the end of the day, this is the area because they're intermediated from the data. Then what we have to show them is that they will essentially make more money by saving time on this, because then that lawyer themselves, their client, is not spending as much money on things that are not legal work, so we're like, hey, stop spending your clients' money on the arts and crafts. You know that you have more energy you could spend on their case. Certainly, would spend more energy on their case if you had immediate access to answers as opposed to those that took you two or three weeks to resolve. And so then you can have there's more available hours available or more budget available to you to bill, and that messaging works out. But then there's another entire class of lawyers of the million or so million two lawyers in the United States. About 400,000 of them are not billable hours individuals, so they're in-house counsel. That's the salary. I think that's about a third of them. So then there are government agency attorneys. That's another 10% or more, and so you have all those classes of folks that don't necessarily make more money if they spend more time working. That actually want to spend less time working, and start once we figured that part out, that became a lot easier. Other sorts of problems in terms of who the buyer versus the user is, and so forth. But I think at the end of the day, I was talking to some friends about this problem, and I'm like, yeah, I'm still going back and forth on whether or not we're selling competency, or we're selling efficiency, or we're selling time saved, and just depending on who you're selling it to, I think that's where you where you have to figure out the messaging, and it's certainly like I said at the beginning, like this is there's going to be a bit of imposter syndrome for me here because I still don't know what the hell I'm doing, and we're just figuring it out in this brave new world. But
Morgan McLintic:you get into the sales opportunity there. We're at a law firm. Law firms are owned by the partners, so this person is inherently interested in the profits of the business and not spending money on technology. And as you say, this is it's quite common for for other startups. Hey, I want you to spend money to save somebody else time so that they can go home and watch their stock and match. But that person doesn't matter to me, right? As much as so that that's a fairly that's a fairly common problem. And the way that you solved it was okay, but the client has a fixed budget, and you're eating quite a lot of it with this process, and you could have a better outcome if you could spend more of your own time on this case rather than eating the budget on eDiscovery as a fixed cost.
Jared White:That's it. Your client has a half million dollars to spend. Otherwise, makes no sense to fight this lawsuit, right? Just settle. So if we're going to fight this and we have a half million dollars to spend, then let's not spend 250,000 of that on e discovery software. Let's spend 150,000 software, and get a better outcome because they're going to call you as opposed to the other law firm down the street next time if you're more efficient with their time. And also, time isn't just time saved; it is the amount of time that the lawsuit is ongoing. If you're able to in courtroom make arguments that don't delay the case or that do delay the case what you'd like to do, and you're able to establish those arguments in the basis of fact and evidence that's in the case, and you don't have to go through two layers of folks to get to that answer, which might take way longer than that you have standing on your feet to make arguments. Then ultimately, you're going to get more cases, and what I think will happen ultimately is that you'll start to see more fixed fee arrangements, and you'll start to we're going to move back into the direction of lawyers billing for events as opposed to billing by the hour. None of us are going to fix that problem. That's going to be the American Bar Association, and so forth, and we'll see what they do, but those are conceptually arguments that would make sense to rational players, because those rational players are going to essentially collect more of the business. They're going to collect a larger piece of the pie. It's also going to grow the size of the pie because there's a lot of legal services that or legal representation that I think is underserved. People that I can only take a matter, or I can only deal with the matter if it's a $500,000 event.
Morgan McLintic:Right,
Jared White:I'm not going to deal with the $18,000 event, and so not even going to hire a lawyer for that. The insurance companies just pay those kinds of things. You can
Morgan McLintic:lower the bar on it because you can. Yeah, it's just not worth it. It's going to cost too much to fight this case. But actually, if I can reduce the cost of that, then I might seek justice for it. That's
Jared White:right. That's right. I think you'll see a lot of that over the next 10 years.
Morgan McLintic:Interesting. So you had some challenges with the messaging and dialing that in, and realizing different messaging is going to work with different subsegments within your sort of customer profile, depending on what their objectives are. And so I think that's brilliant. Right now, you're see funded. We've got one sort of main product that we're selling. Tell me, there's always a sort of a push and pull between investment in the brand and, from a marketing perspective, and investment in sort of driving pipeline. Is brand a deliberate investment for you at this point, or is it a bit of a distraction?
Jared White:Right now, we do not make direct investment to our brand. Probably could make a bit more. You know, love to have time to update our website, for example.
Morgan McLintic:Yeah,
Jared White:and would love to get our presence a bit more widely known on Twitter and or X, as it's called now, on LinkedIn. We do a fairly good job on LinkedIn. Everyone in the sales organization and the executive team basically spends time on LinkedIn posting, we have content that we generate and so forth. That's about the extent of it. I think that we could do better. We don't have enough time to do much more than we're doing, and I tend to think of brand is there's nothing better than them thinking about you because they're using your product, and so then associating you with a solution to a problem or a certain pain that they have is the next step of this, and then being able to associate your solution to a problem before they know about you, before they're using, is the desirable position to hold. At this point, whenever we do our next round of fundraising, I'm very much interested. Both there's about there's an allocation that we've made towards that, and I do want to make a big splash, and you'll hear more from us in the coming months about that. I think, though, that in doing so, we're definitely going to need someone with more of your skill set than mine, who's done this before successfully. It isn't just guessing at it, which is what we would do. I have
Morgan McLintic:to profess a bit of a bias towards building the brand, but I'm very aware that yeah, because it's good to know if people know who you are in the room or they're building their shortlist, and they don't know who you are. You're not on the list, right? But at the early stage, you've just you got to build pipeline, and otherwise, there's not going to be a Series A, right? Yeah, it's great to build a brand. I'm going to need to do this through the lifetime of my company, but I have to have a company, and I have to have revenue, and in my my timeline is short right now, and so the product has to sell. I just have to get bash my way into the room. It's nice to be invited, but I have to force my way in and let the product shine in that sense. Exactly aware of
Jared White:it. That's exactly right.
Morgan McLintic:Okay, at this stage, what would you say are the biggest marketing challenges that you face as a sort of a challenger in the legal tech sector.
Jared White:Great question. This may or may not be the answer to the actual question you're asking, but I'll just reiterate some of the bigger challenges that I'm facing from a go-to-market perspective.
Unknown:Yeah.
Jared White:In terms of everyone has heard of the legacy players, right? And so they're like, "Oh, my. What I hear a lot is my client says I have to use whatever.
Morgan McLintic:Like the big one in this space that's been around for decades, yeah, is running. I had someone
Jared White:say to me, "This is a Fortune 10 company that they're handling a lawsuit for, and they're like, 'I desperately want to use this product instead of this other one. We shall not be named, but they have their own agreement with that company, and so then I have to use it because that's how I get it from them. There was a light bulb that went off for me, which was like, okay, that's who I need to be marketing to, not you. So more on that to come. But so at the end of the day, like that's the bigger challenge. Like, and also to a lesser extent now that we've been around for three and a half years. But to your original point about defensibility and is your software proven enough to be able to actually do this? Can we trust it at all, and which was another reason we went into the criminal defense wedge market, which had no one really servicing that market because they essentially have to use you if they're going to use anything. Then you can say,"Hey, we have 500 court cases that have been tried that have used mating, and so there's at least 500 people, some of which have argued in federal court, and others that decided that this was a value to them, right? Here's the proof in the pudding, and so forth. But we need to be better at publishing those results to the extent that we can. Other bigger challenges that I face are like, what do you actually do? How does this actually work?
Morgan McLintic:Right. Market education, right?
Jared White:Just market education, and like, how should I think about you in the sea of other people that are selling to a law firm
Morgan McLintic:of any
Jared White:number of different types of law firms and wedges and niches that people have. So, what do you actually do? And then, I'm a deep technologist, and I tend to build software that is more widely applicable to it's more widely applicable to use cases than the one that we're typically selling at that moment. Very tempting for me as a technology guy to say it does whatever you need it to do because that's why I built it this
Morgan McLintic:way.
Jared White:Not a good answer. No,
Morgan McLintic:because you need to make it solve a specific urgent problem right now, like the classic painkiller, not vitamin. Oh, this will make you better. All software makes things. It's going to save you some time and make things a little bit better. There's a million things I could do, but actually, as a criminal defense, like I'm facing this case, and there's a mountain of stuff to go through, and I don't have the time. I swore I would fix this next time, and can you do that?
Jared White:And so, this is actually in terms of the biggest marketing challenge. I don't know actually what the hell I'm doing in terms of marketing. The whenever I my initial reaction, even the one that I know to be a poor reaction to this, and it's an internalized one because of coming from a technology background. Whenever I hear you say things like, "Yeah, you have to be that your painkiller, you're like, "Stop your bleeding, not just this is a health and wellness vitamin, and I'm like, "Well, that's really underselling the product, and it's really not. It's actually generating." More revenue to do that, but and I know that conceptually. But again, like the other thing in terms of biggest marketing challenge that we have is like we were selling whenever we go in. We're like, hey, you can put a petabyte of data into this platform, and it you can ask it questions, and it will not hallucinate. We're very proud of this. Okay, and then they're like, "Okay, I'll let you know whenever I have the right case for your platform. So then now they're going to wait for a massive case. Like if you have five paragraphs of data, you can also put that into this platform, and it's only going to cost you some number of dollars. It's not going to cost you 10s of 1000s of dollars. I likened that as like the guy who's like wearing a million dollar watch and driving a blue Ferrari on dates and complains that all the women that he goes out with are gold diggers. We're like, you're advertising the wrong features, brother, and everyone else is turned off by it for that matter. But we started saying to people, you, it's not that we're petabyte scale and that you need to wait for a petabyte sized case. It's petabyte scale and that you just put everything that you ever do in here, and then you're not worried about that, and so sell that horizontally, not vertically. So that was a recent revelation of mine.
Morgan McLintic:Yeah, interesting because from a technical perspective, you are trying to build this for enterprise scale, and the technical challenge gets a lot harder up at that petabyte level, but those cases are few and far between. And what you're trying to say is, hey, no matter what scale you get to, you can use this all the way through. But by anchoring at the petabyte scale, they discount you because that's not. They're like, okay, yeah, I don't
Jared White:have any cases that size. Yeah, you do 1600 cases over the course of your career, and all of them are small, but you should just put them in here.
Morgan McLintic:Great insight. Okay, you're an AI company, and I'm assuming you're all in on AI. How do you think about the use of AI in your own marketing? Where are you leaning into it most?
Jared White:Well, I'm going to actually make a plug for a friend of mine, Josh Payne, who runs CoFrame, so shout out to Josh. And CoFrame is a company that's think of it as they're going to optimize the messaging on your website version. Cool using AI, and so that is a basis starting point of how you can use AI to improve your messaging. I think is one of the areas that we explore, I think we use from a marketing perspective. We did a lot of paid digital last year, tended to not really generate returns for us because we wound up getting clients of lawyers that were finding us as opposed to lawyers that were finding us, and there was that problem. But we did a lot of AI. We we used AI to A/B test a lot of things, not just the quantitative AI, but the qualitative AI in terms of the the marketing messages and so forth. We use we we use AI, for example, in determining who we believe to be someone more likely to convert, so that we can essentially rank our ICPS hundreds of 1000s of contacts in our CRM. We've enriched them. We know things like where they went to law school and when they graduated, because they have that on LinkedIn generally. We know what kinds of law that they practice, so we have a sense from that of whether or not this is a AI forward demographic and whether or not they might be a decision maker at their law firm. And so, because we know where they work and how big their law firm is, so we use AI to essentially find those needles in those haystacks. I'm interested in the first 400 customers, not in the first 400,000 customers.
Morgan McLintic:Right,
Jared White:and I want to figure out who the first 400 people I need to contact because my conversion rate might be one out of X, and if I need my number, my Y, to be some large number, then I need to find the right people because I have limited bandwidth in terms of reaching out to them. And then thereafter, using AI to figure out when they are likely to convert. And so, even the ones that tell us no, oftentimes we find or it's not a no forever. It's just I don't need this right now. And so, imagine you buy new tires on your car every two or three years, and they're marketing you tires throughout this process. I watch a lot of Formula One. It's likely going to be a set of Pirellis on my next course, right? And I don't need those right now. And if someone came to me with a set of Pirellis, I'd be like, "That I don't have a place to put them. Not going to be of use to me at the moment. So then, figuring out when they need new tires is, I think, is something we've been spending a lot of energy on from an AI perspective.
Morgan McLintic:I love that. It's a bit like if you buy a new mattress, instantly you get bombarded by all the other mattress companies trying to sell you a mattress. I literally have just this is the thing I just bought. I need the least at the moment because I can only sleep on one. But I like the use here of okay. To our earlier point, lead generation is really important for us right now. Prioritization is important. There's a large target market in theory, but really, what I'm after is that first, yeah, 400 the next four, and so I've used AI. I've looked at the customers who have converted, and I'm building up not just an ICP profile. Of the kind of company that they are, but also a sort of a persona profile of the people. Where did they go to college? What did they agree with? What do they specialize in? How old are they? All the different demographic factors that might then say, "Hey, this is the person to reach out to, rather than this person, because just any little edge like that, you can do that exercise once, and it will save you save you time. But rather knock on every door, let's just knock on the doors when people are home, and that will be great. I really like that. And what I also liked is you didn't go to hey, we're just cranking out a bunch of LinkedIn content and blog posts and stuff. You pressing the button on AI because sure I
Jared White:feel like that's dilutive of the brand at the end of the day.
Morgan McLintic:Agreed.
Jared White:I want to see like highly. I don't mind the folks I follow on X, for example, or Twitter. I don't mind seeing their posts every day if it's valuable to me. But if this, if I see them posting the same nonsense and AI generated slop, then they're going to get an unfollow within a couple weeks at the most, and so I think that being attentive to your hey, we have a discerning audience here in legal tech.
Morgan McLintic:You really do,
Jared White:and they can smell AI slop a mile away. So none of our content is AI generated from that perspective. Yeah, I might toss an article through Claude and say, tear this down and edit this or whatever, make it sound a little bit different, and then I'll go re-edit it again. But it's going to come from me, or it's going to come from one of the people at the company, the human being at the company, and then yeah, you use AI to maybe post it or to tell you which LinkedIn threads it might do well in. Sure, but someone then has to click the button, but don't use it to generate the content. That's just being lazy.
Morgan McLintic:Yeah, I love that. I I agree. So I guess related then, what do you think that many companies at this stage do wrong?
Jared White:Again, this is an imposter syndrome question for me because I feel like I've done everything wrong. If you're not making mistakes, yeah,
Morgan McLintic:that's fair.
Jared White:I'm not at this stage anymore, right? What are we doing wrong? So honestly, I have found fascinating enough that our company does a lot better with fewer people than more. So, like at some point, we had a lot more software engineers, and for one reason or another, they moved on, and we didn't replace them quite deliberately. And so, we have three software engineers at this company now, and myself being one of those, so I'm part time as a software engineer, and we generated half million lines of code in the last four weeks, and that would not have been possible if we had eight engineers because just the additional logistics,
Morgan McLintic:coordination, overhead, and just prioritization and the management of that. So a smaller team can go faster.
Jared White:Yes, and so at this stage, it's tempting to say,"Okay, we're going to reinvest in the business, and we're going to basically grow the team. In the age of AI, I don't necessarily know that's the right move.
Morgan McLintic:Right.
Jared White:I'm really glad that I wasn't a startup three years prior that got really successful because I actually spoke to a friend of mine a few days ago, who is in that category? I think he founded his company in 2018, and they're huge in comparison to us, right? They're huge, yeah. And they are 75 people now. I asked him how many of those were engineers. I think 30 or 35, and they're going to generally produce less software than my team of three, myself included.
Morgan McLintic:Really,
Jared White:and so it's almost like there's a disadvantage to having been earlier, and there might even be an advantage to having started yesterday.
Morgan McLintic:You just have to unpick the process and reinvent the software development process with this large team, because a small team can move faster, and you've already decoupled the notion of success from headcount. Right? I'm bigger, therefore I'm stronger, and I'm better, and I'm moving faster. That's just implicit human nature. But your experience in terms of measuring the output is not that you've decoupled those two things, and you don't have to you don't have to redo all that. But yes, so an element of luck just being founded at the right time.
Jared White:Yeah, and by the way, it wasn't true for us. I don't know that this would have been true for us a year ago because it wasn't. A year ago, we had eight engineers, and we were looking for more. And because to do more, the models for software engineering got best got worth using for me personally in like the q3 q4 of last year. And by the time q1 all around, it was like this is 80% of my production now, and prior to that, it was still faster for me to write most of it, and then send off bits and pieces to AI and like. But grew the models developed Opus 4.5 I think it was the one where I was like, okay, this is good enough for me to give it whole write from whole cloth, and then I will edit. And prior to that, it was like me writing and letting it edit, if you will. Now, like of that half million lines of code that we wrote, like 250,000 of it is code that is in a language I don't even know, and so I don't necessarily have a lot to offer in terms of right. But I can tell it how to edit its own. I bring that up. I know we're talking about marketing, and I'm bringing some software engineering because software engineering just is a couple years ahead of everyone else, and this is going to be the reality of everything. There's no career that's escaping this. No career, and so look at everyone around your company and
ask yourself the question:Is this person going to use AI in their job, or are they going to resist AI in their job? I think if I were to say at the seed stage, the biggest mistake you can make is growing headcount. That's probably the greatest mistake I made.
Morgan McLintic:Yeah, and I agree. I mean, we're all seeing the sort of early stages of it. You know, and certainly in marketing, perhaps the the bit the the pre Opus 4.5 it'll do bits of it, but it won't do all of it in in marketing certainly, and it's something that we're watching very closely or always experimenting with. But there's not been that sort of you can't outsource it all. In fact, just in our episode last week, we literally just covered can you outsource PR to AI? And there's some elements you can, but you certainly wouldn't do the whole thing. But elements of that are going to constantly change. And then, just as we wrap here, then what's one trend that you've got your eye on that most excites you at the moment?
Jared White:There's a few. I'm inherently a technologist, and so I'll tell you, I'm very excited about being able to run models locally. I'm very excited about inference technology because I think that there's not yet an upper bound on the amount of inference model inference that we want to run, and so there's a really interesting company. I forget the name off the top of my head right now. I'll remember it as soon as we close this. Yeah, they just came out of stealth and they raised$800 million and they built an entire rack from start to
finish, like everything:networking, memory, compute, everything is brand new from them, and it's essentially meant to be a low voltage, meaning low power and low heat inference, because we at this point have no concept of how we're going to deliver all of the tokens that we need to deliver that we project that we're going to need, and I think that we don't yet. I think as we generate more tokens, we're going to want more of them, and it's going to be a it's going to be a positive feedback loop for a very long time, and maybe forever.
Morgan McLintic:So you mean people running their own models locally, and this grok like low voltage inference sort of data center in a box, if you like, could be in your home, or it could be somewhere fairly local, and so you've got a more edge, disparate inference, and that just means that demand for that is going to go way up because the token cost comes way down.
Jared White:Yeah, you think of it this way, right? You had a desktop computer, and then you had the Nokia, what was it, 5100 or something like that, indestructible, and then there's obviously very little computing power on that phone. Now you have an iPhone that's more powerful than your desktop was in that era. That was 20 years. Much, I think that's like a three-year problem now, not a 20-year problem. And then you're going to have inference on every device that you have. There's going to be inference going on in your dash of your car. There's going to be inference going on in your phone, obviously on your laptop. There could be inference going on in your glasses if you have AI glasses,
Unknown:which
Jared White:definitely will at some point. If Neuralink and Elon Musk have their way, maybe there's inference going on in some competing devices baked into your own neural system. That's where this goes. At the end of the day, that's probably a very exciting trend for me because I think that the models will get smarter, will get better at using the weights, will have fewer of those weights. We get better at understanding how to get value out of them, and then it's going to just continually explode. The availability of these tokens will continue to explode. That's exciting to me because I don't think that there's an upper bound on where we're going to take it just yet. The number of decisions that can be made, the human intelligence enhancement that you can do. I'm also very excited. One other thing I'll say: this is completely orthogonal to that particular mindset. But I've been very excited for many months, if not years, about using AI to teach kids. So I have a four-year-old and a six-year-old. Six-year-old just finished kindergarten, and even whenever I was a child, they were telling us that you amount the amount of data in the world doubles and information doubles every seven years, and I was like, "Wow, this is a problem because I'm pretty sure that I've been in school longer than seven years, and y'all don't plan on teaching me twice as much while I'm here, so I'm behind versus where I started in terms of the possession of knowledge that that I have, that I read about a third grader who went through like a customized AI generated coursework for them, and they're already finished with like AP Calculus and like the first year of college calculus. Third grader, eight years old.
Morgan McLintic:Wow! And
Jared White:so I'm very excited about this because of what that does. That's
Morgan McLintic:amazing. Yeah,
Jared White:be a good thing. Again, we probably would not have imagined that we'd use tokens for that, and that might have taken a billion or 10 billion tokens to generate the coursework for that one individual. And it's worth every minute of that. It's worth every token, in my view. So that's why I say there's no upper bound on the tokens that we're going to use.
Morgan McLintic:Yeah, interesting. That's fascinating. So Jared White is the CEO of Matei AI, which is currently a legal tech company. But as you've just heard, has ambitions to be so much more. Jared, listen, thank you so much for sharing your journey and telling us a little bit about the company and your vision and the sort of way you think about marketing. I really appreciate your time here. If people want to get in touch with you, where should they go? My email address, Jared at matey.ai. It's easiest way to get in touch with me. Go to our website and please reach out. I'm also on LinkedIn. You can find me there pretty easily, I think. Yeah, say hi. That sounds great. And of course, all of Jared's contact details will be in the show notes here. So yeah, once again, thank you, Jared, and thank you all for listening. If you like that, please subscribe or drop us a comment, and we'll see you all again next time.
