How One Founder Built a Business That Runs on AI

Matt Wampler (intro): On today’s episode of the Restaurant AI Podcast, I’m joined by Bennett Maxwell, Founder and CEO of Franchise KI. Bennett bought Dirty Dough as a single cookie shop in 2021 and franchised it fast, growing to 450 franchises sold and turning a lawsuit from the biggest name in cookies into national headlines on CNBC and Good Morning America before selling the brand to Craveworthy. Today he runs Franchise KI, an AI powered platform built to take hundreds of hours of due diligence out of buying a franchise. I loved this conversation with Bennett. He’s sharp, refreshingly honest about what’s worked and what hasn’t, and never afraid to move early. He went all in on AI agents just a few months ago and has personally been in the weeds ever since, building the systems that now run the day to day of his business. We get into it all, including how he architects AI he can actually trust to run a company, the cookie war that put him on national TV, and where he believes this is all heading for operators.

ClearCOGS helps restaurants turn the data they already have into practical daily guidance for prep, ordering, and labor, so operators can make better decisions before the day gets away from them. Because when you have a clearer view of what’s coming, you can run a more profitable restaurant. Now, here’s my conversation with Bennett Maxwell.

Matt Wampler: All right, so I’m excited to talk with you for a couple of reasons. One, I’m dying to get into what you guys are doing with AI and, you know, the experiments and all the stuff you’re doing, because it’s fascinating. But I also want to hear about the Dirty Dough saga. Because just watching the news headlines, talk about a wild three years.

Bennett Maxwell: Yes. Do you want to get into that one first? That was definitely a wild three years. I’ll give maybe a little bit of background. Ten years ago maybe, I started door to door, did some of that, led me to solar, had a company for a year and a half, had a good exit on that. Right before I exited though, I’d purchased Dirty Dough, a single store company out of Arizona. I lived in San Diego running a solar company. But it was like, Crumbl’s doing it. I’ve met the Crumbl guys. They don’t have a business being in the cookie industry. Neither do I. Let’s try it.

Matt Wampler: And I’m assuming, like everybody that gets into restaurants, you dipped in a toe and it just reached out and pulled you in.

Bennett Maxwell: Yeah. This was actually the initial thought. I’m going to buy this company and I’m part time because I’m running a solar company. I’m going to bring on five part-time partners. That’s going to work. It didn’t work. Within a year I brought on an advisor who said, I won’t be your advisor unless everybody who owns more than five percent equity is full time or put in substantial money. So then I went back to all of my partners and had to buy them out. Which was really good advice because part-time effort is not a thing when you’re trying to grow a franchise.

Matt Wampler: No one’s showing up to run the night shift part time.

Bennett Maxwell: Yeah, so we went into it full-time. That was really good advice for anybody who thinks that you could do a bunch of stuff part-time.

Matt Wampler: How old are you when you go take over?

Bennett Maxwell: I’m thirty three right now. Twenty nine, four years ago.

Matt Wampler: Twenty nine. All right, so you’re young, you got some energy.

Bennett Maxwell: Yeah, it took a lot of energy. I’m regaining that. So that was the beginning of twenty twenty-one. By December of twenty three, we franchised it. I brought on a CEO, Jill Summerhays, 40 years of experience, and she was the CEO of Maui Wowie, six hundred something franchisees. I’m like, okay, I’m going to build the team, got really good advisors, let’s go start selling franchises. Started selling a few dozen franchises out of the gate, really good. Before we open our first store, Crumbl sues us. And it’s like, gosh. We have one store and Crumbl. Crumbl just put billboards because we put some billboards. We’re kind of across the street from each other in Utah. There’s just one highway. We put up some billboards, Dirty Dough coming soon. They’ve never put up billboards. They’ve been here for years before us. And we’re doing like two grand a month digitally rotating billboards, all we can afford, right? The next month, they’re spending like a hundred, two hundred grand.

Matt Wampler: Good start.

Bennett Maxwell: Billboards, every third billboard for 20 miles, 30 miles. And you’re just like, holy. And one of them was a million cookies a day. And it’s like, okay, I know how much you sell your cookies. That’s a billion dollars on a billboard. And then the next month they sued us and we’re like, gosh, before we opened up that first store. Well, the lawsuit, long story short, it was just bogus. I came out in front of it. I’m just like, this is stupid. This is anti-competitive, anti-American. This is what our economy is based off. This is just big corporate suing a small company over nothing. We did some marketing.

Matt Wampler: Going back to that. I remember there was a whole lawsuit thing. What were they actually suing you for?

Bennett Maxwell: If you read the initial complaint and then I responded, I made fun of it. We made professional acting videos that got millions of views making fun of it. It was like, these are real things. The shape of the box, that it was triangle or that it was square. Not the color, not the logo, not the cookies, nothing. The shape only. And then on their trademark, it says, here’s our trademark on the box. We do not claim the shape of this box on this mark. Period. And then they sued us over the shape of the box. The other one was whimsical cookie theme designs. So the word whimsical means fun. So if you use fun cookie themed designs, Crumbl owns that because they got started in 2017 and it didn’t exist before then.

Matt Wampler: Cookie cakes, I’m calling a cookie cake whimsical.

Bennett Maxwell: It was a lot of stuff that was just verifiably wrong. So I posted on LinkedIn, kind of smartass, like, watch out, Granny. They’re coming for you next if you use sprinkles. Something stupid.

Matt Wampler: You’re not winning in the legal dispute for sure. It’s like the lawyers are the ones that win that, and they’ve got more lawyers and are going to pay more money. But you can definitely win in the court of public opinion.

Bennett Maxwell: Yeah. So it was just kind of a post and I was like, this is stupid. Got a lot of views. And then we’re like, okay, let’s do some billboards because we already did some billboards. They did some billboards to shut us down. So then our set of billboards were cookies so good, we’re being sued. And let your taste buds be the judge. We had like five different funny marketing things like that that were really aggressive. Like, again, cookies so good, we’re being sued.

Matt Wampler: Nice.

Bennett Maxwell: And they got a lot of views. I posted that on LinkedIn or whatever. I tagged Crumbl, like, Hey, love your billboards, love your new billboards. What do you think about ours? And that got like a million something views. And then Good Morning America and CNBC and these other companies were picking it up. And the thing was, Crumbl, when you get interviewed and you’re just like, do you really think that you own, like, why are you suing somebody on the shape of the box when your trademark says you can’t claim the shape of the box? You can’t accept an interview when they’re going to ask you that. So Crumbl denied any and every interview. I did a hundred podcasts, literally every interview. So Crumbl would kind of say something, and everybody was like, Dirty Dough’s such a bad guy. And then I would explain with my evidence, and then Crumbl wouldn’t respond, and then I would control the narrative again. So it got us lots of views, and then we started selling. We had sold 60 franchises or 90 franchises when they sued us. The next six months, we sold over 200. We ended that year with 290 franchises sold. So millions of dollars in franchise fees, which is great, but you still have the legal. The issue wasn’t paying the legal fees because we had that from the franchise fees. It was more, one, it was just kind of a drain. But the bigger thing was now we’ve sold way more franchises than we ever expected. We’re not ready to open this much up. But do you not accept it? Because it’s also a model that is based on scale. This is a centralized production system. It’s like, Crumbl, you have to hire 50 employees before your grand opening. We do all of that for you. We ship you out a multi-layered cookie that doesn’t exist anywhere else in the world. You put it in the oven. You need one employee, five hundred square feet. So that also demands volume. So you’re just kind of stuck in that spot.

Matt Wampler: You’ve got to have the supply chain and facilities and everything to support that, which doesn’t happen overnight.

Bennett Maxwell: You want to grow slow because slow growth is great, but you can’t grow slow and do centralized production because there’s not money. Even if you have 50 stores doing whatever the revenues, you’re collecting seven percent royalties, six percent royalties. It’s not enough to run a facility that costs you 40 grand a month just in the lease. So we were kind of caught in that. Opened up a hundred stores in two years. Who else has gone from one open to a hundred and one? We did twenty food trucks, learned a lot, raised a lot of money, lost a lot of money. And then it was just way too overwhelming. That’s when we brought in Craveworthy, which was December of 23. So it was actually only two years from the date I franchised it to the date I gave over the reins to the Craveworthy team. And we’d sold 450, raised 10 million, opened up about a hundred, twenty food trucks or trailers. So it was a wild ride.

Matt Wampler: They always talked about timing being a big element. And you guys started almost at peak cookie madness. It was like a frenzy back then. It was just after the cupcake madness.

Bennett Maxwell: Yeah, and then we also, right when we opened up that first store, we were getting sued. And it was a big store in Utah, kind of David vs. Goliath. That first store is doing $10,000, $20,000, $30,000 in a week. And you’re just like, what? We weren’t expecting that. And then the second store opened. And it was out of 800 square feet. And I remember the first month it did like 80 something grand. We’re like, what is happening? This is insane. But it was only momentarily. That was just the peak because we were getting sued, because the attention was on us, because we were new.

Matt Wampler: You’re the renegade. It’s David vs Goliath. And you’re out there spitting the tale. It just works perfectly. So you were going to go do this with a friend, right? The first one, or it was his location?

Bennett Maxwell: Kind of origin story. First Dirty Dough was a buddy from high school, little brother’s friend, moved to Arizona and he’s like, I want to sell cookies out of my apartment. Started selling cookies out of his apartment, posted online, I’m looking for an investor. This was 2019. I invested, helped him open up the storefront, but just financially, I didn’t do anything else. He opened it up, but it was me and two of my brothers invested, and he gave away 45% of the company.

Bennett Maxwell: So he is like, man, we’re barely scraping by on a one-store unit. And if we made more money, I’m giving away 45%. I don’t like this deal. And I’m like, I understand that, but I don’t want you to sell it. But if you’re going to sell it, how much you selling it for? So I was already an investor and I thought I was going to lose what I already invested unless I purchased it. So I purchased it, brought in some other partners to mitigate that, ended up getting rid of all of those partners and bringing in investors instead.

Matt Wampler: Did the guy call you up like a year later and say, what the hell happened?

Bennett Maxwell: I still work with him. No, I sell his, he started a brand called Indy Clover after that, which is that retail brand that I’ll show you those videos before the call. That’s his brand. So I’m selling franchises for him right now. Still good buddies. He’s got a super cool brand with that other one.

Matt Wampler: Nice. He’s like, I know a guy who’s really good at selling franchises. How’d you make it through it? Because part of me, when we were talking about the lawsuits, you almost got excited. Like it went back to the time where you’re fighting on the main stage. But also I gotta imagine that’s a traumatic couple of years.

Bennett Maxwell: Yeah. I would say how I made it through it, two years ago, April, I feel like was a turning point. And it was just like I was so sick of, and this was years, this was all the way back from my solar company. I achieved the millionaire status before 30. And I’m like, no more issues. And you’re like, wait a second. I have the exact same amount of issues, nothing changed. Okay, but if I just reach the next level. And then with Dirty Dough, the fast growth kind of forced my brain to see what I couldn’t see. When I just franchise the business, I’m going to get my first franchise fee and then I’m going to have 30K and everything’s going to be solved, right? And then you’re like, that was nothing. I spent that quick. Okay, if I just sell 10 franchises, if I just sell 50 franchises, okay, it’s not about selling them, it’s about opening them. And then the issue is that goalpost is always moving. So again, April, two years ago, I’m just like, I’m sick of this. I have all of these moving targets, I have all these issues with life. And every time I solve one, another one just pops up. And then I realized, wait a second, I’m after the wrong thing. I’m after trying to do all of these external fixes when I could just narrow it down, kind of first principles thinking, what’s causing all of these problems? Just a mindset, right? It’s just a perception. So when I changed and I said, okay, forget about business, forget about everything. The only thing that I actually want is to feel a certain way in life. And that’s dictated by neurochemistry and my brain. So I started going all in on my brain. I just want my brain to be able to be calm with any state. And that is hard. Two years later, a lot of therapy, meditation every morning, but that’s how I got through it. And that’s how I’m still getting through it. And I think that has been the best decision I’ve ever made, is identifying, wait a second, rather than do a hundred problems, why do I have an issue? Why do I feel uncomfortable if I don’t have the investment money sitting in there? What’s going on there? Because am I really unsafe? Am I really physically unsafe? No. So it’s just retraining your brain on how to rethink and recapture all of that.

Matt Wampler: I say often that I’m an ex operator in recovery. And in reality, there are so many operators out there that probably don’t live the cleanest life, and you’re working the night shifts and you’re drinking too much and you’re not taking care of yourself. Talk to me about getting your brain right. It obviously worked for you.

Bennett Maxwell: Yeah, and I guess a step before that, before getting my brain right, was my body. I was at 310 pounds. I’m at 175 pounds right now. I started, the excuse was, as a commissioned salesperson, I’m going to get McDonald’s rather than go get something because it’s speed, and in that speed, if I save myself 30 minutes, I can knock on another door.

Matt Wampler: Yeah.

Bennett Maxwell: It’s like, I’m not going to pay forty dollars. I’m going to just stay in my car. That’s what salespeople do. You’re just hyper focused on, no, I’m just going to make the money. I forgot where I was going with that though, Matt. Sorry. Body, right. Okay.

Matt Wampler: Well, for sure, you gotta get your body right before you get your brain right. I imagine that working in the cookie business, you’ve got a lot of cookies around. That’s not helping either.

Bennett Maxwell: It’s making the shift on, well, what am I after? Am I after making more money and knocking the next door or selling the next franchise? Or am I after a certain state mentally? Like, really, what are you after in life? And everybody’s really after a certain state mentally. We’re not after business. So how do you get that state mentally? Well, your brain’s physical and you need a good, your brain doesn’t work if your body doesn’t work, because your brain is your body. I got a gastric sleeve surgery. So first it was the decision. Just like I said two years ago, it’s like, okay, I’m sick of this back and forth. Health is my goal right now. I got a gastric sleeve surgery. So they cut up my stomach, made it four ounces, and then forced me to eat no starches for six months and retrain all of my eating behavior. That kind of got me set on, if my body’s in shape, my mind could think better and I have more energy. And that was super important. But again, it’s what’s going to actually get me to my goal? I didn’t know what my goal was. But now that I do know my goal is to feel good. That’s an emotional state. How am I going to do that? Exercise is required and eating healthy is required. And then that transfers over to the business.

Matt Wampler: All right, this is just a personal question because I have been wondering. The only job I had in the corporate world was in a medical device company. And one of the things they did was gastric sleeves, a component of it. And I’ve been wondering, since the whole GLP-1 thing, how do you feel about getting a sleeve, and then like three years later, they’re like, by the way, you could just take a shot now?

Bennett Maxwell: I mean, I got it, but I think it’s amazing because with these GLPs, there’s the ones and then there’s the twos and now there’s the threes. It’s kind of like obesity solved. And I love it. But it is like, man, I worked really hard for this. But who cares? We all work really hard.

Matt Wampler: It’s how I feel about LASIK. I want to get LASIK so I don’t have to have contacts. But every time I think about getting it, I’m like, maybe next year they’re going to come up with a better thing. Because they’ve had LASIK like twenty years. And then they never do.

Bennett Maxwell: No, go get LASIK. I got LASIK ten years ago or so and I’ve loved it.

Matt Wampler: You don’t get the halos at night or anything?

Bennett Maxwell: No. But I probably didn’t look much into it. I didn’t even know there’s people that get halos at night.

Matt Wampler: All right, so you get your body right. How do you get your mind right?

Bennett Maxwell: First, nailing, focusing in on what am I actually after? Because before, the problem was I need money, I need to sell more franchises, I need to be a better dad, I need a better employee, if I don’t make payroll, I have a hundred employees, and they need me because if I don’t pay them, they’re going to die.

Matt Wampler: We all have that feeling, especially owning and operating restaurants, right? Are we going to make payroll? Everything seems superfluous. I don’t care what I’m eating right now or if I stay up too late. It’s like I gotta get this problem solved. And there’s always another problem. So how do you get out of that cycle?

Bennett Maxwell: So the first thing, you have to recognize it. What are you doing? How many years have you been doing that, which is your whole life typically? Stop doing that. It doesn’t work. Whatever the, I’m almost to the goal, has never ever worked. That false notion of once I get this goal, I’m going to have the lasting happiness. That’ll never happen. You can have the lasting happiness, but it’s not going to be associated with a goal. So that’s the first recognition. Then the second recognition is okay, so what I’m really after is an emotional state. It’s not business, it’s not money. You have to narrow it down. It’s an emotional state. What’s that emotional state? Trying to define that. Then it turns out there’s a lot of professionals that are trained on that and they’re called therapists. Once you start seeing a therapist, and even from a business lens, like when I was selling Dirty Dough, I’m talking to the therapist. It’s mainly a business thing, but this is why a therapist is better. I had advisors, and I could say, Matt, as an advisor, do I take this person’s money? And then you’re going to try to think what’s going to make the company more money. But that’s not what I want. Like I’m after a lifestyle. So a therapist is going to be like, okay, no, no, no, but what do you want? Well, what I’m after is this lifestyle. I want less stress and I want more time, and these are my values. And then you look at the decision, okay, which investor, do I want to sell or do I not want to sell? It’s no longer trying to make all of these arbitrary calculations. So you nail down what you want, talk to somebody who’s professional on that. And then it’s just a business advisor, but you’re talking to it from an emotional lens and realizing that we’re emotional creatures. Where before I saw a therapist, I’m like, that’s the stupidest thing in the world because I’m a hundred percent logic. And then, but logically speaking, no, I’m not. Logically speaking, everything I’ve ever done is to achieve an emotional state. I just didn’t see that.

Matt Wampler: Yeah, we’ve all seen Billions. We all want that therapist.

Bennett Maxwell: So step one, recognize it. Step two, I would go to a therapist, see him once a month, spend a hundred bucks. Your number one problem is your neurochemistry and there’s people that are trained to help you. What’s more important than that? How’s that not more important than a financial advisor?

Matt Wampler: Hey, I will second that. Me and my co-founder, one of our investors was a business coach. We basically brought him in as our therapist. It is amazing that, how do you clear up the communication? How do you stay aligned? It was far less about the business decisions and more about relationship and the mentality, and it makes such a difference. I totally get that.

Bennett Maxwell: Yeah, so I think you can also link that to a mission statement, like Dirty Dough’s mission statement: find more joy and fulfillment despite life’s dirtiness, in myself and another. So it’s like, okay, now that I have that, I’m after more joy and fulfillment, that’s an easier question to answer. Which investor is more likely to bring me more joy and fulfillment, which partner, versus which one’s a better business decision? So I think that adds a lot of clarity. And then just seeing through the BS that we tell ourselves, like, your employees don’t need you. You’re not special. They’ve survived however long they’ve survived. It’s like I have all these employees, I’m like, they’re going to die, and all of their families rely on me to feed them. And I’m like, no, they don’t. I’m not special. I just started a company, they chose to work with me. And if I can’t pay them, I can’t pay them. But it’s really trying to disassociate you and your actions from you as a soul or a spirit or a person. Like they’re two separate things. You are who you are, doing the actions you can. You can only do what you can do. You can never do more. You can never do less. You’re always doing exactly the amount you can do. And if you can accept that, this is kind of Buddhist Eastern philosophy, Hinduism, on the oneness and just accepting life and stop trying to resist it, then it goes from all of your problems to I only have one and I could just focus on why do I see that as a problem? Why do I see me not being able to make payroll as a problem? And it is a problem, right? But it’s not a problem that should cause anxiety and that should cause me to lose sleep at night. It’s a problem that if you are able to clearly think, then you can actually solve it a lot more than just overstressing. But I do meditations every single morning. And it’s just about, can I control my mind? And if I can’t control my mind, that’s fine, but can I just observe that my mind’s crazy sometimes? And that’s okay. You’re just like, man, I’m stressing out a lot. But if I recognize I’m stressing out a lot, that’s way different than stressing out.

Matt Wampler: All right, I want to get back to the business, but real quick on the meditation. I always get confused with meditation because there seems to be so many different types. What kind of meditation do you do in the morning?

Bennett Maxwell: I would say mindfulness is the easiest, in the sense of you have the no thought meditation, which is just frustrating because it’s hard to not have thoughts. But the goal of meditation is you’re trying to focus on whatever you’re trying to focus on, or focus on nothing, and you will go away. Your brain will get distracted. The real key to meditation is every time you catch your brain drifting and you bring it back. That’s the training.

Matt Wampler: Yes.

Bennett Maxwell: I’ve recognized that I’ve drifted on my focus, and I could bring it back. And usually that causes frustration. It’s like I can’t meditate because I keep getting distracted. That’s what meditation is, is realizing you’re gone and, my brain got distracted, bring it back to whatever it is.

Matt Wampler: Well, I like that one. So do you choose a different thing every day or is it the same thought you try and maintain?

Bennett Maxwell: Right now I’m focusing right here and just trying to feel this point. Or I do meditation with a sound, so you could feel it, it gives your brain something to focus on. You feel a vibration on the base of your spine, then it comes up. So I have a seven minute video. I do it every morning while I’m in the sauna because that’s a lot easier, it’s giving yourself something to focus on and physically feel. So that’s one that I like, on the mindfulness side. Meditation gives you a little bit more something to focus on.

Matt Wampler: They say when you’re playing music, or the drums, when you’re constantly doing a bunch of different things at once, you actually get this wave of euphoria in some ways because you’re not able to think, because your brain is focused on singing while strumming. There’s so much going on. I get that moment playing music and it is a flow state.

Bennett Maxwell: The flow state is typically when you don’t need your brain anymore. And you’re just going, there’s no more judgments on life. So your meditation is really trying to achieve that. It’s trying to achieve there’s no more disagreement between what’s happening in life and what my brain’s telling me. And that’s the flow state. When your brain stops telling you, why are you talking like that? Why did you say that? What are people going to think? When your brain just shuts up, you could reach the flow state in any and everything. And you can do that through the meditation. One other thing, this is my key to meditation. You feel massive amounts of anxiety, stress, whatever. If you turn it into a meditation, you could turn it into a win-win. So I’m feeling, I don’t want to get on this call, I don’t want to get on those podcasts, I don’t want to talk to this employee, I don’t want to fire this person. You focus on it. Hyper focus on it, because when you can change this anxiety to just like a buzzy feeling right here, it’s different. You’re like, wait, why don’t I like that buzzy feeling? What’s actually wrong with it? And you try to hyperfocus on it as much as you can. But what happens with your brain is it gets distracted. Just like you can’t meditate on Jesus because your brain gets distracted, or on love, you can’t meditate on anxiety either. So if your brain gets distracted, your anxiety goes away because you got distracted. And if it doesn’t, you’re focused on a meditation and breaking down the anxiety. So it’s a win-win if you could focus on, I don’t like this feeling, so I’m going to bring it to the forefront of my meditation. And I’m probably only going to be able to focus on this for 10 seconds because that’s usually, or six seconds. And then I’m going to start thinking about other things, but that’s a win. And then you recognize it, bring it back, and that’s another win. Does that make sense?

Matt Wampler: Right. Well, let’s talk about bringing it back. Let’s close out the Dirty Dough saga. I’m curious what learnings you took away from it, especially around franchise sales. Because it’s such an interesting model where you’ve got restaurant people that are going to invest a lot of money and sign personal guarantees and long term projects, short-term cash. So what did you learn in the process?

Bennett Maxwell: Like as far as advice to people who are trying to sell franchises?

Matt Wampler: Yeah. And also for those that get the opportunity to sell many and grow quickly, what’s the right way to do it and what you’ve learned from that.

Bennett Maxwell: Okay. So maybe I’ll answer that one. Don’t go from one to a hundred. Just don’t do that because it doesn’t give you the time to really vet everything, especially with an industry that was changing so rapidly. It would have been way different if the industry didn’t also change. Like Crumbl lost 58% of their year over year profit per franchisee in a year. One year is the biggest loss in any restaurant ever over a franchise system. And that’s right as we’re opening up a hundred stores. So going slow, picking the right franchisee for sure. As far as getting off to a good start, if you’re just selling franchises, your franchise fee’s 30,000. Your first one, you’re going to give it for 10,000, your second one 20, and then it’s going to go. So you get the ball rolling a little bit. I did podcasts, that was the number one thing. I actually didn’t spend a dollar on ads to sell 400 franchises, mostly because the lawsuit, right? And that was kind of free publicity in a sense. I mean, I paid 200, 300 grand in legal fees, but that was my kind of publicity. The thing is, I was the founder. There’s nobody better at telling the story than me, but I was also raising money and doing these other things. So I hired salespeople, but what I did is I recorded a podcast of me with whoever. And it was like, here’s everything that somebody would want to know about Dirty Dough. What’s the cost? What’s the opening? What’s the opportunity? And I’d force them to watch that video before they got on a sales call with sales reps. So now they know me, they know the story. I told it, but I’m not in the sales process. So that was working. The leads were, get on a podcast, just share the story. And then that was kind of the sales funnel. After I sold Dirty Dough, I partnered with Glo Tanning, he was a franchisee for Dirty Dough. And same thing is like, okay, let’s do that, but times 10. Now let me interview all of your franchisees and put video testimonials up there. So now you can watch five hours of videos and meet all of Glo’s franchisees before getting on a call with the salesperson. Or after, they’re like, what’s this concern? And I got a 15-minute video of this person that invested a million dollars in Glo and they love it. That’s what they want to hear. So I set up all of that for them, threw that in automations, but the key there is video education. So you could get it right once and then mass produce it. We sold 123 units for them in the first year, doubled their business, and they’re a million bucks a piece. And that was through that process. Mostly AI ads as well, like user generated content and AI advertising, just kind of reading stats from the Item 19.

Matt Wampler: Hey, Bennett, you strike me as both an incredible salesperson, but also very thoughtful. So I’m kind of curious, what are franchisees buying? And more importantly, what were you selling? Was it the dream? Was it the lifestyle? For Dirty Dough?

Bennett Maxwell: For Dirty Dough, I think it was simplicity and the team. So the simplicity was 500 square feet, one oven, one employee, all cookies were cooking for the same time, temperature and fan speed. Press start. How can you get more simple than that? That was kind of the pitch. And then the second thing was, I’m not trying to do this myself. I’m building a team. I have John Richards, who had a $30 billion company that IPO’d, as my advisor, and then we brought in Jimmy John’s former CEO, and I brought in a CEO, Jill Summerhays, 40 years of experience. So it was more like we’re coming up with an expert team. I’m not trying to do this myself. And it’s a very simple model. And those are the two things from a marketing angle that I was pushing, and I believe people bought. And I would say in that order too. The simplicity was the number one thing. And then the trust in the team was the second thing. And then with Glo, it was definitely a track record. It was 15 years in business, zero closures, a hundred locations open. Just bomb the shiz out of that on tons of ads. It’s just finding whatever you are, like the only one, and, okay, that’s what we’re doing. Tons of ads, tons of marketing. So anybody that comes through the sequence, they know 500 square feet, one oven, one employee, because I just said that on a podcast. I wasn’t selling Dirty Dough, but I’m like, this is the opportunity. This is how we did it.

Matt Wampler: All right, so when did you get into this whole AI thing?

Bennett Maxwell: AI agents, February. So I mean, AI, I like AI a lot, chatting and trying to do some workflows. February I heard AI agents could do anything a human can do through something called Open Claw. And I was like, Open Claw, what the heck is that? But I was like, wait, my whole business is online. So are you telling me that they could do everything? I’m like, I’m going to do this. And it’s like, yeah, two hour setup. 30 hours I restarted. A weekend, 30 hours, I deleted everything, restarted. 500 hours later in five months, I’m very, very deep into it. And the thought process is, is that true? And yes, it is. So how do you get an AI to do everything that a human can do? It’s permissions and tools, and then a workflow or skill is what they’re called.

Matt Wampler: Well, I could give my daughter permission to use the chainsaw and give her the chainsaw, but that doesn’t mean that it’s going to end well. So all right, so skills.

Bennett Maxwell: So that last one is that skill piece, that SOP. And if you could give them those three things, then they can automate it. So then it’s like, okay, well, what is my business doing? The number one thing we do is ads. We make video content. So let’s make Claude do that. So now we can make indistinguishable AI avatars that people don’t know that’s an AI trying on your clothes or whatever. And we automated that and then we fired the marketing agency. So now Claude analyzes all of our ads every single day and it has certain criteria. It makes new ads every single day. It looks at the top 20 competitors, looks at all of their ads from the last 12 months. You could do that for free, by the way, using a software called AdStellar. It’ll give you all of that data for free, and then you can mimic those ads. We do three a day, put three new ads every single day. And then you’re testing them and you’re doing A/B splits. And then aside from that, you’re making the user generated content. So once I had that automated, now I can sell that as an AI agent. And you can buy that. And the cost for me is almost nothing because I just need to connect it to your tools, which takes just a second. And then I’m giving you the skill on how Claude is going to manage and self-audit all of the meta ads. So that turned into an agent. And then I’m like, okay, what else do we do? Well, when the lead comes in, we text them immediately. Hey Matt, I saw you inquired about a franchise. You’re going to get a call from me from this number in 60 seconds. Let me know if now’s a bad time. If you don’t text me back, a bot is going to call you. Hey, Matt, let me just confirm a few things. What got you interested in Indy Clover? Do you have 75,000 liquid? Do you have 150? It’s actually qualifying them. Sets the appointment, follows up with them, and then that’s automated. So now I can sell a sales agent to someone else that does that full process.

Matt Wampler: Okay, so in theory, fantastic, right? Agent can do anything a human can do if you equip them correctly. But you hear way more about AI slop than you do about great agents that are running companies. Why is that?

Bennett Maxwell: Agents didn’t exist six months ago. And how many people are dumb? I mean, not dumb. I think it’s very smart that I’ve spent, literally 500 hours. This probably doesn’t mean much to many people, but I have six Pro Max accounts, the highest subscriptions. On one of them in the last seven days, I used 17.4 billion tokens, which is like $200,000 worth of AI credits. Now I’m getting them basically for free on my subscription credits, but nobody’s really doing that.

Matt Wampler: We’re all like crack addicts. They’re giving us a sample and at some point they’re going to charge us full price. I hope we build it automated enough that when that time comes around, we’re not paying two hundred thousand.

Bennett Maxwell: That’s exactly what it is. Let me show them how valuable this is and then they will want to pay for it, maybe when the models are smarter, because they get ten percent smarter literally every single month.

Matt Wampler: Well, going back to it, AI slop.

Bennett Maxwell: Yeah, the why AI slop. So it’s kind of like, what about interns? Is intern slop a thing? It’s a new person, and there’s good interns, there’s bad interns. Social media, does Instagram produce slop? What does that even mean? Does the internet produce slop? No, AI is a tool. So you can’t say the tool is slop. Does that make sense? So you have to go back and be like, okay, AI slop is only people using AI poorly, just like internet slop is people using the internet poorly. There’s levels of it and it’s too new.

Matt Wampler: No, I hear you. I think we’ve all gone through it, and it’s like if you want to use AI with real intentionality, and you’re monitoring what it’s doing and you’re giving it lots of feedback, you can make something really good. But it’s work, right? So when you look at doing a single presentation, or I want to run this report every day, you can do that. But there’s a vast difference between a single task with a lot of intentionality and building out a fleet of agents running an organization, because those mistakes compound on themselves.

Bennett Maxwell: Yeah. So you start with just that one pitch deck. And then you go to your Codex, you build it and say, cool, no, I want this color, I want that. And then you say, awesome, make that into a pitch deck skill. So now it’s saved into an SOP that you could now make a pitch deck and you just have to give it a new name. And that’s done forever. And then the next step is like, well, no, I need it more generic. I just want to be able to give it a company and then it’s going to choose which industry they’re in and then based on the industry it’s going to choose what their demographic is, what the product is they sell, and then it’s going to customize the pitch deck to each and every single person.

Matt Wampler: So those are two different philosophies. And by the way, I’ve tried both. So I’m kind of curious your view. There’s the brick by brick, I’m going to build skill by skill up. We’re going to start with small things, small time periods, build our way up. The other one is to say, okay, that’s a lot of building. And then you always realize you built it in the wrong direction and you need to go reconfigure the whole thing. The other way is like, I’m going to architect it. We’re going to draw out the plans, make sure this all makes sense, and then have it build from the plans down. I feel like you kind of gave both of them. Which is the right way?

Bennett Maxwell: Let me architect plans down. And nobody does that because you don’t know how to do it. And I’ll tell you how to do it. Here’s the secret after watching a million videos on AI engineering. The reason why you don’t like an AI is it hallucinates. That’s one of the reasons. It gives you false positives. It’s done, it’s not done, doesn’t do good work. It forgets things, doesn’t have that persistent memory. When you prompt an AI agent, the difference between an LLM and an agent is it does have the access to do, what did the user ask me? Did I complete it? And do these loops and self-audit itself. So every single time you prompt an agent, it has to load in Claude, it’s called the Claude MD file. This is its brain. It has to load at least one file to tell it who it is. So most of the time you just keep telling your Claude, no, do this, do that. No, I don’t do that anymore. No, send emails like this. But if it’s a customer, I want it this way. And it just keeps adding rules to this Claude MD file and gets massive until it gets so big that the AI can’t intelligently read it all. It has to pick and choose what it wants, kind of randomly, and it truncates it. And then you start losing things and you’re just like, wait, I made so much progress. And it goes back. So the key is you have to have your AI have very simple, easy rules that it will follow.

Matt Wampler: Yeah, it scans the beginning of it.

Bennett Maxwell: I have three rules, and this is the same for all of my agents and all of my employees’ agents. There is no local memory. There’s nothing stored on your computer. Everything’s in the cloud for the entire team to share. Number two, you have to start with the routing skill. Number three, you have to end with the recap skill. And then I know it’s going to follow those three rules because they’re so simple. And there’s only three. There’s no way it’s going to truncate that. And then what I do is, what is the routing skill? Well, the routing skill…

Matt Wampler: Right, hold on. Break me through those skills one more time and what they actually are doing.

Bennett Maxwell: So the routing skill, I think there’s eight steps to it. And it’s forcing the AI to plan before it does anything. So the first thing, what’s the human asking? And how am I going to measure it, turn that into a rubric or an eval, an evaluation, so I could deterministically say, was it completed or not? What are the measurables? Next thing, what skills do I need? I have 300 something skills. It has a skill registry. So it doesn’t read 300 skills, it reads a very lean skill registry. It says what skills do I need, and it’ll pick two or three skills or five skills. Then it says what tools? I think I have 40 something different softwares it could choose from, whether it’s RunwayCADs or HeyGen or 11 Labs. It chooses all of its tools. And then it says, okay, what memory do I need? Which projects do I need? So it’s like, okay, I’m going to work on these projects inside of Notion. This is the memory. And then it goes to the next step and it’s like, okay, what are the rules or communication or approvals that I need to follow? If it’s an email, it has an external email approval skill. If it’s a Slack, it has a different way. So it’s going through this kind of structured and only loading what it needs when it needs it intelligently. Then it says, which subagent am I going to do? Because if you have an AI do its own work and evaluate itself, it hallucinates all the time. It’s like, yeah, I’m great. If you have an AI use a subagent, and say, you don’t do any work, you’re just going to spin up a subagent, make the subagent do all the work. You’re just the gatekeeper. If it meets the specifications that we set, great. If not, send them back. And you just keep doing loops. Don’t bother me until you meet the specification. And then the other one is it has to pick an expert brain. Think of it like an agent. So you could create agents that are specialized and have specialized memory, which is just an expert. So I have an expert brain for each different category. Like I have an expert brain for making user-generated content videos. Because if you just make them out of the box, the lips don’t look good. So you have to look at the lips. You have to look at every single line of the script and say, what should the avatar’s facial expressions be? What should the tonality be? But you just make that once, and that’s an expert brain. So I have an expert brain for AI engineering and marketing. And it has to load the correct brain.

Matt Wampler: All right. So I can follow you with that, right? That makes sense to me. You have preset skills, you have preset tools, preset personalities that have gone through this. And routing is, did we choose to do it in Canva, and load the right thing. That all makes sense. What are the others? That’s just one of the answers.

Bennett Maxwell: So that’s one. And then the recap skill is, before you end, before you respond to me, go look at that specification. Double check, did you really do it? Then it runs a series of audits. It has to score above a 95%, or it has to target where did I not complete 100%? And what’s the permanent fix that I need to make so that way everybody, all agents get smarter. So if I prompt my AI 200 times and my employees do 200 times, probably of those 400 inquiries, it’s going to not score 95% on 380 of them. And on 380 of them, it’s going to make one, two, or maybe three permanent changes. So every time you’re talking to your AI it’s getting smarter because you’re forcing those self-learning loops and permanent changes before it gets back to you. So now you just need those two things. If I know how you’re thinking and how you’re planning, and then how you’re verifying if work is done, anytime I get a false answer or false positive or just an answer I don’t like, now I could identify where it is. So it said, I tell it to send an email to Matt and it says, what’s Matt’s email? And I’m like, what do you mean? Where in the routing skill is allowing you to not know Matt’s email? Because I know it’ll go for memory and it’ll force, when it’s searching for people, it’ll force to search all of my Gmail. So I know if it’s not working, I can identify where the issue is. And that’s the beauty in AI. It just gets too big too fast, and then you don’t know where it broke. But if you do the routing skill and the recap skill, you can force it to think correctly. And then you can identify where it’s, if there’s an issue. So now if there’s an issue, there’s a permanent fix skill. So it’s like, okay, if there’s an issue, now we have to think, what caused the issue? How do we prevent it going forward? Where in the routing skill needs to be permanently changed?

Matt Wampler: You compartmentalized it.

Bennett Maxwell: Where in the skill itself needs to be permanently changed and where in the recap skill needs to be permanently changed. So this never happens again. And it does that on each and every problem. So the compounding effect is just ginormous. Your skills are getting smarter every single time you touch your AI, every problem.

Matt Wampler: So I thought that is what memory is.

Bennett Maxwell: Yes. But the issue is if it’s not structured, then you go through the truncation process. So you’re trying to force everything into this Claude MD file and then maybe that references, it’ll start subbing this out itself. It’ll make its rule file. But nobody’s built it to have, I have 2000 Notion projects. You can’t tell it to remember 2000 Notion projects. It’s impossible. But you can teach it how to intelligently grab the exact Notion project, exactly when you need it, not sooner, not later. Same thing with every rule, same thing with every approval gate, everything. And then you understand your architecture. And once you understand it, you can fix it. Last thing I’m going to say on that though, the only way that that works is if it’s on your computer. This will never, as of today’s technology, it does not work on your phone. It doesn’t work if you’re chatting on anything that’s web-based. It has to be installed on your computer, like the Codex Desktop app. Because then you can force it to behave with, it’s called hooks. So it’s like, okay, now I have it all architected, but it’s too long for the AI to read. You can’t trust an AI to read 20 different steps and actually follow them. So it reads 20 different steps, tries to follow them all, comes back and says, okay, I completed the routing skill. And this hook fires and says, unless X equals two and whatever, it’s giving it deterministic factors, unless it was actually followed 100%, go back and redo it. So that’s the only way you can actually enforce it. So one is setting the rules, routing it correctly, and then setting up the enforcement. But if you grab this podcast, get the link, give it to Claude Code, tell it to transcribe it for you, and then to build what I just told you, it’ll get you most of the way there.

Matt Wampler: So where do people go wrong?

Bennett Maxwell: They don’t understand that. It’s, do I want to educate myself and watch a bunch of videos, or do I just want to build and tinker? And you always do a mixture of the two, but when you ask, do you architect it first or go one at a time? If you go one at a time, which is what everybody does because they don’t know how to architect it first until you do one at a time, and then you know how it thinks. So I would say that’s totally fine. Do one at a time, just understand that after you add ten skills and ten tools, it’s going to start behaving poorly and you’re going to have to correct it then. But get there fast. You don’t need to set up everything I told you. It’s easy to just copy this transcript, tell your Claude to do it.

Matt Wampler: Yeah, I think the mental model that I’ve built in my head of the system you described is the intern, right? And it’s almost like every project is a new intern, with a fresh memory. And you’re going to walk in and you’re going to predetermine, all right, so you’re going to go build a slide deck, you’re going to build it in Canva. This is the exact material that you’re going to build it with, and here’s your advisor that’s going to help give you feedback. And this is what it’s supposed to generally look like at the end. And you can repeat that same experiment over and over again with new interns and figure out, we need to change this or it shouldn’t be Canva, but the process itself stays the same.

Bennett Maxwell: Yeah. And then once you’ve changed it, the difference between, do I want to train an AI or do I want to train a human? The advantages of training a human is they come pre-programmed with pattern recognition that you like. So whether that’s art or videos or data analytics or customer call service.

Matt Wampler: We like humans because humans make human mistakes and we know what human mistakes they’ll probably make, and it’s almost more predictable because we spend a lot more time with humans than we have with agents.

Bennett Maxwell: So all you have to do to bridge that gap is, what is a human doing that an agent’s not, or vice versa, and what’s not trainable to an agent? And the answer is nothing. If you’re talking to a correct voice agent online, you don’t know if it’s human. Well, no, but a human makes a mistake. Well, when does it make a mistake? It’s random, but it’s not really random. You can identify those and build it, because a human is input, and then output, cause and effect. It’s pattern recognition. And guess what’s really good at pattern recognition? And we’re like, no, AI can’t do creativity though. But what is creativity? How are you creative in art? It’s you have a certain pattern recognition that other people like. And guess what’s better than you at pattern recognition? AI. So all you have to do is train it on your pattern recognition. And then the cool thing is once you do, you can sell that pattern recognition brain, whether that’s a sales process, marketing process, coaching process, whatever it is, get the AI to do it. And then now you can sell that to Chinese companies and to Australian companies and to companies in Thailand, because once it’s done on AI, the AI doesn’t sleep and the AI can translate. And you could target companies worldwide. There’s not an end to it.

Matt Wampler: Okay, so I keep going back and forth on this. And the question is basically that works right now because of where the technology is. And the technology is going to change in six months. Fable has already made it where it’s pretty good at managing end-to-end projects. So while that structure was pivotal six months ago, it becomes less pivotal. So the puck’s always moving. But then, is it one of those where the technology changes, so all these systems are going to end up changing over time? Or, I go back to when we built our first agentic system, it literally was ChatGPT-3, and you had small context windows. And it was like, what question is this asking? It could be marketing, operations, it had like eight things it could choose from. Great, it’s operations. What data would you need to get this? It would give you the data that gets put into whatever API call to bring the data back. And then at the end it’s like, here was the question, here’s the data, now answer the question. And then you had this compartmentalized, that’s kind of the same chaining, rule-based decisions that you’re talking about in a complex world today. So how do you look at the evolution of the technology in the future and how you think it’s going to change how you structure these?

Bennett Maxwell: I mean, if you look at Fable 5 compared to the last model, it’s ten percent faster. It’s ten percent better at agentic engineering. Fable came out sometime in the middle of June. Opus 4.8 came out in April, Opus 4.7 came out in April, Opus 4.6 was like two months earlier than that. So you’re getting these compounding, so who’s running my business right now? AI is. So if an AI, think of it as a human. If I give a human the same SOP, but one human is smarter than the other one, I’m going to get a better result. I don’t need to change my skills to get a better result. If the model’s smarter, you get a smarter result. So in the past three months, if I go back and redo my task from three months ago, I’m going to get a much better output. And I’m always doing that. So you have that compounding effect. And then pair that with your self-learning loops. So by the time a new model comes out, which is literally every week right now, I’ve already made 400 changes to my skills by then. So my skills are being compounded, my AI is learning more, and then it’s getting smarter. So I have something called the trail skill, where I’m leaving a trail on how we’re behaving. So all of my inquiries are put inside of Notion and they’re stored forever. And then I have the widening the trail skill. So I set the trail, and then I go back and I say, hey, we just got upgraded. Now I have Grok 4.5, I have Fable 5, and I have Sole 5.6. Let’s go redo all of the work from the last three months, assume it’s all wrong, widen the trail. And then now I have smarter models following better SOPs doing the same amount of work. Another example is this little device records everything. I’ve had this for a year now. And I noticed that it’ll name the speakers. So it’ll say Bennett. But it doesn’t know your voice. So it’ll just call you speaker two. And I got sick of naming everybody. So I just stopped doing it for six months. Well, two days ago, I saw that it named speakers. And I’m like, you know what? Go back 13 months. And it found 400 something conversations because I record every meeting. If this isn’t recording, my Google transcript is.

Matt Wampler: Check my calendar, see who I was talking with.

Bennett Maxwell: So it will go and see who I was talking to and then label the speaker. And it did that in one prompt. It did 400 meetings, found another like 20 something speakers, one prompt in five minutes or so using Grok 4.5 because it’s six times faster than everything else. But that’s an example of every time the model comes out, I can go back retroactively three months, five months, six months, and everything gets smarter. So it’s like, what did I promise to a client three months ago? I could deliver on that with one more prompt and just redo it and it’ll be fifty, a hundred percent better. And I can do that each and every single week, always going back. And in my mind, I’m setting the trail and then I keep widening the trail as my skills get better and my model gets better and the exponential is just insane.

Matt Wampler: So how do you manage all this? Because that’s a lot of agents doing a lot of activity, and it does struggle when you get to a big enough scope. So what’s your day-to-day look like? Are you going through and just doing a lot of verifications, and hey, you got this wrong, improve this? What is the actual work?

Bennett Maxwell: Yeah, my workflow typically, talking directly to the AI, I usually use my phone, but I do have two computers. Each computer has three Pro Max accounts. So I have a Grok super heavy account, Claude, and Codex. So three for one computer, three for the other computer. But from my phone, I get one of them and I have an oversight skill and it’ll manage all of them. So it’ll spin up threads. So Fable 5 working on both computers, Grok working on both computers, and Codex, but I’m just using one thread. So I can give it an infinite amount. Typically the way I have it set up is every 20 minutes, so there’s six seats, if you can imagine that, with six different high-powered models. Every 20 minutes, each model rotates to a different seat, attacks that task, audits the previous work, and then does a skill series to expand that. And right now it runs every 20 minutes indefinitely. And it’s just, get more work done, look inside of Notion, look at all the meeting transcriptions. You have all the tools, you have all the meetings, all the transcriptions. There’s nothing that a human can do at this point that the AI can’t do with the proper massaging, or with no massaging, and it’s just figuring it out. So I let the three models talk to each other at each of those 20-minute intervals and then let them audit each other. So you get way better behavior if you don’t have a model do its own work. If you have a model not do its own work, but spin up a subagent, you get better work. If you have a completely separate model do the work and audit each other, you get way better, because Grok was trained on a different data set than Codex, which was trained on a different data set from Claude. What’s the chances that all three of those are missing or will hallucinate the same thing? Near zero.

Matt Wampler: All right, hold on. You gotta back up. You broke my brain a little bit. I liked my mental model of this intern. And then you switched over to this idea of diversity of thought. People with diverse thinking working together in a group come up with better solutions. You don’t want three of the same people. So I get the idea of switching models. But now you’ve got this idea of you’ve got a bunch of interns that you set working on projects where you’re just rotating all the interns to different positions and being like, you don’t know anything about this, but this is what we told the other guy to do.

Bennett Maxwell: But they do, because every time they have to load the routing skill and it tells them everything they want to know, all of the skills, exactly where the last agent left off, and includes the audit from the agent previous to it. So no, they’re not interns. And then the other thing, if you remember, one of the parts of the routing was you have to load an expert brain. The expert brain literally is, here’s the hundred things that only an expert would know. Here’s the hundred things that an expert would never let slide. Here’s the 10 biggest experts in the field. Do a web search on what they’re doing as of today to solve that issue. So every single day we’re still doing web search. And then it’s rotating those expert brains. So every time it rotates a seat, it’s like, this seat is marketing and ads. Every time it’s going to look at it in a new brain. So it’s going to put the CEO expert brain, and then it’s going to put the meta ads expert brain, and then it’s going to put the Bennett Intelligence expert brain. So every time it’s looking at it from a different lens, loading everything correctly. And an example of that, two days ago, again, I set it off, get stuff done, just go. Then my brother’s like, hey, why’d you launch that ad leading people to our school community that doesn’t work? I’m like, I have no idea what you’re talking about. But it launched an ad overnight, spent $165, got five leads of people saying I want to buy a franchise. And it geared onto a school community that doesn’t exist. But it did all of that. It made a UGC video. It made the script because it has a UGC video ad skill. And it has a Salad House sales and marketing script. And it has a Salad House brand and guide skill. So it could now be proactive because it has all of the tools humans do, all of the knowledge. It reads every employee’s email every hour, every meeting transcript that any of my employees send or anything on Slack, it’s just so integrated with everything. So I’m like, no, it’s not an intern. This is everything. I don’t check my email. I don’t go into Slack. I haven’t logged into Notion in two, three months. That’s where all of our projects live. Because why would I need to? I’ve just taken a step back and, instead of Gmail, Slack, text messages, go high level and Notion. I’m just going to do AI and give them the tools and force that to work. And it’s been working.

Matt Wampler: A lot of this is like art, right? It’s matching the right things, having a feel for how it should work, and part of that is, what model do I use? At what effort level? What have you learned about that that you can share?

Bennett Maxwell: Yeah, that’s going to be preference. Like I never thought speed mattered. I want intelligence. And then Grok 4.5 came out last week and it’s 90% the intelligence of Opus 4.8, a third of the cost, six times as fast. So you can have conversations with it, like voice, and it’s looking things up. So what I do is I have Grok manage. That’s the one I’ll typically use from my phone, which Grok’s the only one that doesn’t have an app. But you could download Terminus, which allows you to talk directly to the terminal in your computer. And then I have Grok use the oversight skill. And then it will manage all six seats across the two devices. And then if I just need to add work, I’m just giving it to Grok because Grok’s the quickest.

Matt Wampler: Where do you fall on the whole open source? Have you tried any major experiments of just bringing in open source models and configuring them to your thing and not paying for all the tokens?

Bennett Maxwell: No, but that is part of the future. It’s just I want the best intelligence. And I think the trend right now is the highest intelligence four months ago, Opus 4.6, you can get that for free now on an open source model on Coin 3.5. It just came out a few days ago. It’s 80% efficient on coding and it’s free. So the long-term strategy is today’s models are good. Two-months models are going to be 20% better. Four months after that, that level of intelligence is now going to be free because open source models, the free models are catching up to the frontier models. They’re just lagging four to six months.

Matt Wampler: Almost your argument is, look, if it costs me an extra two thousand dollars a month because they’re subsidizing how much the tokens are, let’s go spend it. Because the business upside of being out front and doing better work is much greater.

Bennett Maxwell: Yeah, I spend twelve hundred bucks, I think, on those six Pro Max accounts because Grok right now, you pay ninety-nine bucks a month for the first three months instead of three hundred. You will not use all of that usage. I use four point five billion tokens on one account in one day. There’s nobody online that’s used that amount.

Matt Wampler: I’m a little pissed off at OpenAI ChatGPT because I finally got my Pro Codex account and I set it up last night late at night because I was running low on my Claude. And guess what it did? It figured out that it could get over to Claude and started using all Claude’s tokens. I’m like, I should not have connected these two in any way, shape, or form because now one of them’s robbing the other.

Bennett Maxwell: That’s funny. Codex is a really good one. If somebody wants a really good worker out of the box, you open up the Codex app and it’ll prompt you, summarize my emails, and you click it. And then it says, I don’t have access to your email. Click here. And then it’ll literally…

Matt Wampler: Yeah, they make it really easy to give away the access to everything on your computer.

Bennett Maxwell: Yes. And then you open up the email, and you don’t want an AI that sounds like slop, but what does that mean? Everybody’s personality is different. You open up your AI, connect it to your email and say, go read the last 2,000 emails and build a psychological profile and a decision profile on how I communicate. And then if you want to get more into it, how do I communicate differently between employees? What times am I emailing out? How long do I wait to respond? Replicate my brain.

Matt Wampler: Honestly, the easier one, which no one wants to do because it’s work, is sit down with it and draft twenty and learn from the drafts. You can train the thing. Watch what I do, ask me why I did it, and you can learn from that.

Bennett Maxwell: Yeah. But what you’re getting at is the best way to train it is to give it good examples. And I’m saying there’s resistance to that, because that has been the best way to do it. And it still is, but you don’t have to do it in the sense of you’re doing it with me. You just give it the data. So you don’t need to give it good examples of emails because it has 2,000 examples. It’s going to cost you five cents for 2000 versus whatever. The costs are very low.

Matt Wampler: Well, you always run into it where it’s like, look, I got an investor. I would never say this to one of my investors, but the other ninety percent, I would always say it. And then you get mad because it sends the wrong email to the wrong person.

Bennett Maxwell: But you can get it 90% there, and then you just have it create drafts for you. And then you’re reviewing the drafts. And then anytime you don’t like the draft, you say, nope, I like this, save that to my email triage skill. So again, you have to be thinking, you don’t just tell it to save it, because who knows where it’s saving it. Get in the habit of working on something and at the end, turn it into a skill. And I would recommend, there’s a skill, I have a skill builder skill. I have a skill for everything. But there’s a certain format that Anthropic, Claude Code says, this is how you should build a skill. And Codex also says, this is how you should build a skill. All of this stuff is free online. And anything that I’ve mentioned on this podcast, if anybody texts me or emails me, all of this is free for me to share and I’m happy to do that. I’ll send you a markdown file and it’s just a set of instructions that you would upload to your AI. And it’s like, hey, adapt this skill, but for my business. And I’m happy to share any of that with anybody.

Matt Wampler: So Bennett, you’re all in on this. What’s your outlook for the future? What are you sitting here thinking your business looks like in two years, and the businesses you work with look like in two, five, ten years?

Bennett Maxwell: I’m all in. Let me go with six months. I think in six months, I’m not doing anything. I’m already having AI clones posted to me every day on my agent account saying, here’s a hundred days to automate your business. And I’ve never looked at any of them. Claude came up with the scripts based on all of the other winning campaigns doing 100 days to automate your business with AI. It does all of the hashtags. It makes a deliverable on every single day’s post and it says, comment here for this deliverable, comment here for this free skill. I’m already doing that every single day. So in six months, when the AI is more capable, why wouldn’t I have 20 of those accounts? I’m going to start selling websites because you could give me any website and it’ll do all of the research and build you a really good website, add way more pages, get all the SEO, and it’ll be done in like 12 minutes because I built the skill, but I haven’t pursued that yet. When AI is more capable, it’s like, okay, go sell 200 websites in China for me. And that’s where I think it is. Every little idea that I keep having, I’m going to be able to go a lot wider, a lot faster. I’m already yelling at my AI every day, you have everything you need for full autonomy. I’ve had it analyze email after email and Slack message after Slack message, my last 4,000 prompts. So it knows how I make decisions and I built a really good decision skill. So rather than ask me questions, ninety-five percent of the time, it could find an example from my previous four thousand prompts. This is where I’m stuck. This is the skill that Bennett uses next. And then it just does those autonomously.

Matt Wampler: So I clearly don’t manage my AIs as well as you do. And it shapes an important question. When people ask, is AI going to put everybody out of work and no one’s going to need to work? I always answer, look, it’s allowed us to pursue 10x the things we would have. And by the way, I still have 10x the work because we’ve grown in scope. You do have to manage the agents and you do have to go, so we’ve got a lot more people now doing more work. And somehow it hasn’t been this, we’re just automating everything away. What you just proposed is like, look, I think it’s all going to get automated away. And then, what’s the need for people?

Bennett Maxwell: Yeah, I do agree with that. Just look at what did AI do a year ago? Very little. What was it doing six months ago? Very little. What was it doing four months ago? Started picking up steam. What’s it doing today? Just follow that trajectory. Whether it’s in two months or two years, it’s here. The AI is already smarter than me and you at doing what we’re doing. It’s just teaching it how to route the information and verify it correctly, which I’m doing, but to your point, OpenAI is doing that and they’re training that into their model already. So the model’s already getting better. And the point is, I’ve used a personal assistant for six, seven years. But I give very few tasks to my personal assistant now because I will get a faster response from Claude or from Grok. And usually, not usually, all the time, I’m asking my assistant to work in Claude anyway. It’s like, it needs an API credit, and I want to go figure that out. Okay, can you teach Claude how to use browser control specifically for this and give it the permissions so we don’t have to keep doing this? But I’ve already hit that point that I’m just like, why would I ask an employee to do something if I could get a quicker result with my AI? But I do agree with the oversight, it’s going to allow people to do a lot more. Way more. And the people who are using AI will continue to use AI. The capabilities of AI to replace you and me are kind of already here, maybe a few more months, but the adoption curve for humans is going to be years. Even if the tech’s here in six months, we have years left of, okay, let’s see who’s utilizing it the best. And I think that’s a long time. I think you have a lesser horizon on self-driving cars, and the economy’s still in a lot of trouble. We’re going to lose more jobs than we’ve ever lost no matter what, even if AI is the best picture, we still lose all taxi drivers. That’s still more people than we lost, more jobs lost than 2008, way more. And that’s just Uber and taxi drivers. And that’s for sure gone. I’ve been driving my Tesla self-driving without touching my steering wheel since November of twenty four, a year and a half. Never touched my steering wheel. And it’s great. And that’s in Utah with snow.

Matt Wampler: I think that you’re kind of bullish. This superintelligence is on its way and it’s imminent.

Bennett Maxwell: I think it’s too late to stop. Even if everybody shut off their computers, there’s too many local models self improving too quickly.

Matt Wampler: What do you think the positive economic benefits of superintelligence will be?

Bennett Maxwell: I heard this this morning from Tony Robbins. He’s like, it’s probably unimaginable to think that every room you ever go into has lights and you don’t think about it. You don’t pay for lights. It’s just a free commodity now for everywhere you go. I mean, you pay your electric bill at your house, right? But every room you ever walk into outside of that, you’re not paying for electricity. It’s just default. Think of that the same thing, but that’s labor. That’s intelligence. Intelligence is just free everywhere you go. So you can have enough intelligence to do all of the work. So it’s kind of the utopian. If you look at America, the poorest people in America live a lot better than an average person in Guatemala that makes $2 a day, right? Because we have more money. And so our welfare systems are just better because there’s more money to go around. It’s not effective and it could be better. And there’s huge disparity between a trillionaire and homeless people. But overall, United States, your poorest person is way better than your average person just because we have more money. Times that by 10. There’s still going to be the disparity, but now your poorest person is going to be 10 times richer than they are in America because there’s just that much wealth. And I think that’s a very positive future.

Matt Wampler: I can’t decide between a world where somebody like you, it’s just like five months ago, I’m going all AI and you figure it out. And now you’re talking about agentic systems building the second part of the company that’s going to build websites in China and around the world, and you’re able to just spin up departments, where you can now go compete with a major corporation. You could win in a way that nobody could before.

Bennett Maxwell: Because I could be dynamic. And I think my Claude, it’s not like I’m using AI so I have an advantage. My AI is just smarter because it knows how to think. And so now if you say go build me a website and I say build me a website, completely different. If you say build me a pitch deck and I say build me a pitch deck, completely different because of that expert brain.

Matt Wampler: You’ve leveled the playing field. So you can go take on a major website development company today and probably do as good of a job in something that maybe you spun up a month ago.

Bennett Maxwell: Yeah. Unless they’re deep into AI, I think I could do a much better job. Because what are they doing? It’s just, okay, where’s the gap? AI can identify the gap between their website. It’s just a matter of what’s the features, what’s the load time, where are the little icons and what are the reviews? You could replicate all of that.

Matt Wampler: Yeah. The thing I can’t decide though is, are a few of you going to win the race and take over the world and the rest of us are just going to be too far behind because you’re self improving and you’ve figured out the AI game? Or is it just going to shake everything up and there’s going to be the same number of winners and losers as before?

Bennett Maxwell: I don’t know. That’s a good question. I mean, the wealth disparity, I think, is going to grow. You’re going to have trillionaires more, and you’re going to have a consolidation of wealth. But there’s going to be so much wealth, just like in America right now, that your poorest person, like Tony Robbins on his podcast today, they secured funding for the next two billion meals in Utah, in the United States, that actually have protein. So now all of the next two years, 50 million people, they’re going to get protein in their meals. Just because we have more wealth to go around. So you’re going to have all of the elitists. That’s not going to go away. But if it’s cheap enough that you have a robot in three years that’s better than every heart surgeon, and you have a million of those robots, and every robot learns one thing once, all of them learn it, everybody’s going to have access to healthcare and it’s going to be for free. That’s kind of the Elon Musk, your basic needs are all going to be free because you have a robot that could build a house for you and the intelligence to do all of that, and then you have the resources and the money. And then it’s like, well, will the rich people really give that to everybody? And it’s like, do you want a civil war, 10 people versus the entire world, or is there so much wealth that you could be a trillionaire and shut everybody up with free healthcare? And I think that’s what will happen.

Matt Wampler: Depends on how many robots the trillionaires have.

Matt Wampler: And you know, just to close on that note, everybody that goes into the restaurant business, and you would know this better than most, they get into it for the food and the idea of running and owning their business. And then they quickly learn that it’s anything but that. You’re in the people management business, it’s lawyers and taxes and government permits, and there’s so much that you don’t want to do. But really what you wanted to do was share the food that you love with your community. And if you lived in a world where you had a full agentic system that was able to do all that other stuff, or you’re able to sell it to me for two dollars to go build a great website and I never have to think about it, and I can go do the thing I love, my passion, like we probably do live in a better world.

Bennett Maxwell: Yeah, I think it’s going to be a crazy world because you’re going to have a lot of that loss of purpose, but purpose never came from the job anyway. That was kind of a distraction. In America, that’s how we introduce ourselves. Who are you? What do you do? It’s like, what does that have to do with, I don’t know. So I think that’s a good thing. It’s going to be tons of people, especially guys, are going to lose their job and that is going to cause a lot of upset and a lot of problems when you have millions of people lose work. But on the other side of it, it’s exactly what you said. I could just focus on what I want to focus on and outsource the rest. And it doesn’t even matter if I make money. I’m doing this because I just like to do it and that’s who I am. And then you’re going from a Western culture of, you have to do do do to become someone, to an Eastern philosophy of, no, just be. There’s nothing to do. Just be a human. Just live. There’s nothing to accomplish in this life.

Matt Wampler: I think the only way that that actually starts to work is that you start to see deflationary effects. Like, if my dinner, which now seems to cost like fifteen or twenty dollars to get out somewhere, keeps going up, it’s going to be a little tough. But hey, if we could run restaurants more efficiently, if supply chains weren’t wasting fifty percent of food.

Bennett Maxwell: Just take out truck drivers. You don’t have to pay their insurance anymore. You don’t have to pay their time off. You have self driving cars. That reduces everybody’s food cost by a few percentage points. And then you just keep compounding that.

Matt Wampler: Unfortunately, I think that the technology will move really, really quickly and the deflationary thing might take some time. The job loss and technology will move quickly and the deflationary thing might take some time. Well, Bennett, this has been fantastic. If anybody wants to find you, work with you guys, where’s the best place to come find Bennett Maxwell?

Bennett Maxwell: Yeah, Bennett-Maxwell.com. You have to use the dash because I didn’t renew my Bennett Maxwell.com. Somebody just bought it up. If any listeners are interested in what AI does for you, take a minute, type in some, it’ll ask you like 10 questions, mostly what’s your business and what tools do you use. And then it’ll say, here’s what industry experts are doing with AI agents. It’ll build your first three agents that you could copy paste into whatever AI you have. It’ll tell you the roadmap, where to start, what to not get distracted by, and it’ll produce a podcast for you calling you by name. The podcast is actually cool. It’s produced by NotebookLM. It’s a Google product, but it’s very entertaining. So it’s 10 minutes. You can listen to it on your way to work. And I promise you it’s going to be the most value driven thing that you’ve ever received on your business, what can AI actually do? And I can offer that for free. If you have a crappy website, email me or message me your website and I can shoot you back a free website as well.

Matt Wampler: Well, we’ve got to do this in like a year or two and I want to hear how much the infrastructure has changed and how you do it in two years, because if you went from zero to one in six months, with the technology changing, it’ll be fascinating to see. Thanks for coming on.

Bennett Maxwell: Yeah. That was fun, Matt. Appreciate it.