Consulting Rainmakers

The AI setup that replaced his inbox and his CRM

Bennett Maxwell ·

Bennett Maxwell is the CEO of Franchise Key. This Consulting Rainmakers episode, hosted by Mikulas Zach, is about one idea, in Bennett's words: The AI setup that replaced his inbox and his CRM.

Can AI agents replace my email inbox and CRM?

Bennett Maxwell: He connects AI agents to his email, call transcripts and CRM so they schedule calls and update records on a timer. He rarely opens his inbox or CRM himself and checks the agents' work with a second AI.

Watch on YouTubeSpotifyApple Podcasts

About Bennett Maxwell

Bennett has not opened his email app in over a month and has not logged into his CRM in two or three. He dropped out of college after making a quarter of a million dollars knocking doors in five months, sold his solar company, then sold four hundred and fifty cookie franchises in two and a half years before exiting that too. Now he builds agent systems that do the parts of his job he used to do by hand, including booking this interview. His argument is that anything you know how to do on a computer is now a text file you can sell, in any language, at almost no marginal cost.

Expect to learn what an agent actually is next to a chatbot, how to stop a model lying to you using deterministic hooks and a cross model audit, what door knocking taught him about closing, the eighty twenty of sales, why detaching from the outcome is the real CEO skill, why he is not worried about what AI usage will cost, and the exact email system you can build today in about three minutes and much more…

Chapters

  1. 0:00Pest control, solar, cookies, and two exits
  2. 3:13What an AI agent actually is
  3. 7:21Killing hallucinations with hooks and cross model audits
  4. 10:20What door knocking taught him about selling
  5. 13:27The eighty twenty of sales, and detaching from the outcome
  6. 16:44Turning his own knowledge work into a product
  7. 18:55Why he is not worried about AI costs
  8. 23:50The one thing you can build today
  9. 26:44Where to find Bennett

Transcript

Machine transcript of the conversation between Bennett Maxwell and Mikulas Zach, lightly cleaned. Spelling of names and terms may be off.

Mikulas Zach

Welcome back to the podcast. today I'm joined Bennett Maxwell, the CEO at Franchise KI. Bennett, it's great to have you here today.

Bennett Maxwell

Yeah, thanks for having me. I'm excited to be here.

Mikulas Zach

So so tell me how did you get into this role and what do you do?

Bennett Maxwell

Sorry, I'm gonna pause for a second because I realized I'd I like to make the screen like small up to my camera so it looks like I'm looking at you at least a little bit. And then I'll go back to okay. Sorry, I was gonna, so my eyes kind of looked up. and then your question was, what did I how did I

Mikulas Zach

Yeah, what w how did I how did you end up here?

Bennett Maxwell

Yeah, very random career path, I guess. I I've done a lot of different industries, starting with like pest control. The idea was you can learn so much from each industry and you're gonna learn the most from that industry in you know it's always always front-roaded, front loaded. If you've heard of Pareto Pareto principle, the 80-20 rule. So it's like you're going to learn 80% of the knowledge with just 20% of the effort. So by working for a different company, different industry every year, you're going to advance a lot more. If that's what's like, I'm gonna learn 80% of what there is to know about selling a pest control product in the first year. Now, if I stay there for five years, maybe I'm gonna go from 80 up to 95%. We have those diminishing returns. So that's kind of my strategy early on. Like I dropped out of college and I'm like, I'm going to work for a different sales company every year. So I did that for a number of years. and it was really good at like making me an expert. communicator and then it allowed me to sell different products and kind of learn a lot. So that led to then I started my solar company in 2020 because I'm like, okay, I'm already like building sales teams. I'm already doing the sales. Let me bring it in and just do it myself. I partnered with my brother and that went really well. Did it for about a year and a half and then somebody bought it. I didn't even know that somebody would buy that company like selling a company wasn't even in the cards for me because it was just like you're we're hiring 1099 contracted salespeople were knocking doors selling solar contracts like what are you really buying but they were buying a system and we had a seven figure exit when you know didn't really plan on that. That led to Dirty Dough guess I was already involved with a cookie company called Dirty Dough I purchased it one store and I'm like I want to franchise this and Crumbl's doing really well right now with franchising never done anything with food never done anything with franchising but I'm like hey same thing like I could figure it out What what's Crumbl doing? What's Mrs. Fields doing that I can't figure out how to do? Or hire somebody that already knows how to do it? So it's kind of the mentality, and you know, it's a good mentality for some things, bad for others. It causes you to Grok way too fast. we sold 450 franchises in like two and a half years, opened up a hundred brick and mortar locations, raised just shy of ten million dollars, closed twenty-something stores, like it all went so fast. I brought on the former CEO of Jimmy Johns and his he had a restaurant group, they came in and managed it for a while and then ended up purchasing it. That was about two years ago. And then I started my current business Franchise KI helping existing franchises scale. Rather than you know taking Dirty Dough from one location to 400, let's take something for that already has 100 locations and you know double it in size, which is what I did with my first client last year. They had 100 locations. We doubled them, sold 123. And then this year brought on a few other clients, but I'm really getting into AI agents. So, anyways, that's kind of the long story, but that's how that's how I'm here.

Mikulas Zach

Okay, and we talked a bit before hitting record about AI agents. So can you talk more about that and your process of creating AI agents?

Bennett Maxwell

Yeah, so I started, I I heard the concept at the beginning of this year in February, because yeah, maybe for some definitions. You think like AI, most people think of an LLM or a large language model, something I'm gonna talk with. What's the difference between that and an agent? Well an LLM, they're like they're getting smarter and they can connect to some of your tools and they they're actually getting better at memory, but ultimately they're just good for that one conversation. What an agent is, is that same LLM typically. put into a what's called an agent harness. And it's like now you're getting that same brain, but you're giving it access to a lot more tools. And you're giving it access to be able to plan its work, execute the work, verify its own work, and then if it doesn't, you know, if it doesn't meet the criteria, do it again while it's using different tools. So that's an agent. And agents are capable, you could build that agentic loop to do anything a human can do on a computer. And I'm like, that blew my mind away. And this was, you know, in February. And it seems like we're in another world since February with AI agents. you know. So I'm like, I'm gonna dive into that. So for anybody listening, conceptually to manage people is a great skill, right? Like if I can manage people and have them do you know my effort, but I could spread that across 10 people, great skill. Instead of managing 10 people, if I can manage 10 managers and each of those 10 managers now manage 10 people, now I have an influence of 100. right? Like that's a really good skill to have. Now think of it. How many agents can you manage though? You know, like if I'm getting proficient at agents and I could manage an agent, now I can manage 10 different agents. Well, I don't want to manage 10 different agents. I want to manage 10 different looping agents. And each of those agents are managing their own 10 sub agents. That's one's doing the research, one's doing the coding, one's doing the quality check, one's poking holes into it, you know, you know, and you can kind of set it up through that. So that's terminology, like that's what an agent is. Grok with xAI, Claude Codex or Codex. You download any three of those out of the box, and like in 10 minutes, they will check all of your emails. They'll learn exactly how you've been speaking over the last five years or 10 years, because it could just analyze all of your emails. It could build a decision profile on how you make decisions. So it's not like this AI sounds like a human, because that doesn't mean anything, like humans sound different. This AI sounds like Bennett, because it was trained on 10,000 Bennett emails, you know, and that's how easy it's becoming. So once I saw that, it's like, okay, well, what does my business do? What tabs do I have open on my computer? What can I start automating? You know, so like this podcast was scheduled. I've never emailed you back. It was fully scheduled by my bot because I haven't checked my emails, physically like my own emails probably in over a month. I haven't logged into my CRM in probably two or three months. because I just talked to my agents and my agents have all of that access. So it's kind of like if you have a a super personal assistant that's super quick at everything, you know, like you wouldn't log into your CRM because you're you're just telling your assistant, hey, go go read my call transcription, summarize it, and and and go update, go high level for me. And then, you know, in there I told you when the next call would be, just go complete everything. That's what my claw's doing. So as long as I'm talking to a prospect and like, okay, just rattle off, you know, five cities that you want me to run a demographic report. That's all I have to say on the call. I don't have to trigger anything after because I have a trigger. You have, you know, you can you can have your agents work off of schedule tasks. So every hour it scans everything in the company, any sales calls, any every call, and then it it completes all of the items on there. So you can kind of start building it out. And if you tried this two months ago. Like legitimately it's a hundred percent better than it was two months ago. And if you wait another month to try it, it it's probably gonna be another fifty to a hundred percent better. Like the the exponential curve on how quickly it's going is is getting really.

Mikulas Zach

Hmm. And how do you go about AI hallucinations? Because it's something that I've been dealing with a lot, is you know, AI just I ask it to do something and then it says that it will do it and it doesn't do it, or it does it in a different way.

Bennett Maxwell

Two answers. One, I'll keep the the the best way is you have to do what's called deterministic hooks or deterministic behavior. You can't rely on the model to think and to reason. Like what you need is, hey, as soon as I talk to you, you have to plan exactly what success means. and you have to create a rubric for yourself. And then you don't want the AI to do its own work. You want it to have another AI. So the easiest way is you can have your Fable 5.1 just spin up a subagent and you just tell it, do a sub agent, have the subagent do the work, you grade it. Just by telling it that, it'll implement that, you know. and you'll get a way better response because one AI is way better at poking holes and you know, another AI. Now, what I do, because I pay for all three of the those ones, Grok, Claude, and Codex. I do what I call as the CMA skill. Like I'm like run it through the CMA skill, CMA skill, which is a skill that I put together, cross-model audit, CMA. you have to grab two other models. So it could be if you're Grok, you could grab Claude, you could grab Quinn, you could grab Chemi 3, like you know, I give it five models top tier. You have to grab two of them that look at whatever you're working on from a dozen different opposing viewpoints. And then you have to to grade it zero to a hundred. So it's like, okay, you score to 60, make these changes to score higher. And then you have to go make that. And then I say do a loop until you can't improve after two rounds, or you hit 10 rounds, you know, so it doesn't go on forever. But usually it only goes like, yeah, two to five rounds. And it, so it's now you come up with your plan or your finished product, and then you go, Claude like, this is great, I made it, you know. And then you go ask Codex, and Codex is like, no. That's not great, you know. They're really, really good at poking holes in each other. So setting it up that way for me reduces a lot of that. And then the other thing is again, if if you're structuring it correctly, like you can't tell me it's correct unless you've provided proof and you've checked that proof and you've put that artifact inside of Notion and you check that and then you know, you kind of through trial and error you start building more on it. And then if you can make it behave deterministically. And you just talk to your claw and like, no, I need to install hooks. So you can't ever say it's done when it's not actually done, it'll walk you through how to do that.

Mikulas Zach

Hmm, very smart. You you said earlier that you were doing door-to-door sales, right? what have you learned and did it transition into something that's that you are doing right now?

Bennett Maxwell

Yes, I think I'm good at podcasting and conveying my ideas because I did door-to-door sales. Podcasting sold hundreds of Dirty Dough franchises, which was, you know, millions of for sure millions of dollars just on like franchise fees, you know. When I so I was in I was in college, I had about a semester left. I wanted to be a pediatrician because that sounded like a cool title. I liked kids, made decent money. It was 250, you know, you can make over $200,000. And that's just the number that I wanted because I don't know why. So I started going, you know, I going to school. I had about a semester left to graduate with pre-med and I did my first full summer. Moved out to North Carolina for about four and a half, five months, recruited a People to go out there with me, knock doors, and I made $250,000 in five months. And I'm like, this is silly. Like, what why am I going to school? Like, unless I'm really passionate about being a pretty nutrition, which I wasn't. I didn't even like blood. So what that I guess changed for me is like sales is a professional career if I treat it as such, you know, because most people treat door to door as kind of like, especially in Utah. That's where all the door-to-door companies are. It's just like, go do it for the summer so you don't have to work during the year. But I saw that as much more. I'm like, okay, what if I drop out of college? But the same hundred thousand dollars that I have earmarked to pay for college and the same four more years of med school and three years of residency, what if I allocated that same effort into being a salesperson? And then I'm like, well, it's not really a salesperson. I mean, it is whatever you want to define it, but it really is a communicator, right? Can I be a professional communicator? And then what does that not translate to? Like when I'm hiring, who's gonna buy a Dirty Dough franchise from Ben at Maxwell? He's never done franchising, he's never been in food, but I probably had the most rapid growing franchise like from zero to a hundred for sure in the last, you know, probably five years, maybe even a decade, on like actually opening that in the entire world. Like it worked, but it wasn't because people are buying franchises from me. It's because I recruited the right team. And how could I recruit the right team? Because I was able to convey my ideas and also ask questions and listen and try to, you know, find something that's mutual mutual beneficial. And then I had to go raise money. I've never raised money. I was raising an average of $200,000 every month for two years. I didn't get any big checks really. It was just like little checks here. And if I didn't make my $200,000 raise, you know, that month, or maybe, you know, a million over however many months, we were going under. So again, like all. Everything in business, I think it it it conveyed like when you're looking at it as I need to be a professional communicator, translates to everything.

Mikulas Zach

Hmm. What is kind of the eighty-twenty in sales now that you have a lot of experience?

Bennett Maxwell

Yeah, putting in 20% honestly, I think it's mostly confidence on like not overthinking and not caring and not having like the I need to sell you something. because you could get on you could get on calls very unprepared, not know anything about that person, but Be there and be present with them and just listen to them and be genuine and don't try to bullshit them on like if they have a valid concern that discredits what you're selling and you agree with that, then agree with them. You know, like I I think those are the the the little things that most people are like, Yeah, well I don't know if I want to buy a salad house because we would be the first one in California and that's a lot of risk. And I'd say I that's a absolutely a lot of risk. What do you think about that? Like, why did you apply for that? You know, knowing that it's a lot of risk, why did you still apply anyway? And they're like, Well, I know it's a lot of risk, but there's also a lot of benefit because then I'm first to market. And I'm like, okay, cool. So what do you think is more important to you? And then if they say the risk isn't there, I want to know that because I want to resolve it or I want to get off the call because. You don't force somebody to buy a half million dollar salad franchise, you know. It's just like if it's a fit, it's a fit. And my job is to see if it's a fit and help them see it. Because a lot of times they'll look at a franchise and they're just like, the turns aren't good enough. And I'm like, compared to what? Compared to what? And they're like, like real estate. And I said, I have real estate. What returns are you seeing on real estate? This is a call I literally had last week. Because the girls like, I show her my amazing franchise. I'm like, They make $68,000. It costs $180,000 to open up. You're making your money back in three years. You don't have to quit your job. Yeah, just the returns aren't that good. I'm like, show me any fucking property that you've bought that has better returns than that. Like it doesn't exist. But most people also wouldn't be honest and like a little bit more combative. So I think all of that, like leaving all of the sales stuff aside and just be like, no, I'm just gonna get to the root concern and be honest about it. And if it's a real concern that I would also have that same concern, I would be like, yeah, I don't, I don't, I don't recommend this for you. And that's totally fine, you know. Wh where I think too many people are are too worried about losing a cell.

Mikulas Zach

So being detached from the outcome of the call is helps.

Bennett Maxwell

Yeah. And then extend that to what's probably the number one business advice as a you know, a CEO. It would be the same thing. Just in regards to the entire company, like to maintain a good mental state and be a good CEO, in my opinion, that's probably the number one thing as well. You have to disengage from the outcome. like you you have goals and you're shooting for it, but you're not attached to the outcome at all. Because you're attached to whatever, you know, you're what you can only do what you're what you're doing, nothing more or less. And otherwise it's leads to kind of that hyper obsessiveness and control, and then you're stressed, and then you can't sleep. And then when you can't sleep, how are you gonna go close a deal and go raise money and go lead your team? It's hard because I've done that too.

Mikulas Zach

Yeah, and now with Franchise KI, what is the goal in the next year?

Bennett Maxwell

Yeah, I wanna continue to bring on right now we have three brands. So last year I just did one brand. three brands, but the bigger, the bigger play here is what I'm selling is a lot of value, but it's also because it's like, it's me. Like I'm really good at selling franchises. So not many other people can like, hey, I could go sell, you know, I've sold 500 something franchises in the last few years. But it's very limiting because again, it's me. But now with AI, like that voice agent that I was showing you before the call, like that's trained on my calls. So it's gonna be pretty similar to me. And every call I'm on and every time the smart the model gets smarter, it's gonna be more and more like me. so crap, I forgot where I was going with that, man.

Mikulas Zach

Yeah, no no worries. My my question was: what's the main goal with Franchise KI?

Bennett Maxwell

Okay, that's what I was going. I think I could replace everything. The main goal is I think now that I could replace I it doesn't need me to be involved to sell your franchise in a sense. 90% of it can done be done without me. So I could sell that in the form of an agent now or a group of agents. So it's like now you can buy Bennett franchise system, which has an agent that will produce a hundred ads for you every single month and manage all of them. And You know, it's like one of those ads that I showed you. Like they're indistinguishable from real life. They're AI avatars that will do look and say anything you want with any accent. And you'll show it to 100 people, and a hundred people would say, That's a cool sweater. Nobody would say, Wait, that person looks weird, you know. so I could now sell each aspect of that business calling kind of into one. And I think that's that's where the world is going. is the AI agents are giving you the ability to take your knowledge work, whatever your specialty is on your computer, and turn it into a freaking SaaS product. So now it's like free because it's just a markdown file with a set of instructions. I could go sell that to all these other franchises and it's marginal costs. And then after you've exhausted your market here, then you know you could go to Canada and then you can go to China because AI doesn't have language barriers, you know. So whatever. you're selling like you can sell it worldwide with any time zone and any language because the AI can do it if you're training your AI to do it for you. If that makes sense.

Mikulas Zach

Hmm. How do how do you look for or how do you approach usage credits? Because I feel like you know the subscription models that we have right now are very genuine and it they allow you to spend a lot of usage for a relatively small cost. Like I was just looking at my cloud usage the other day and it said something like I don't know four thousand dollars in the last one month on a two hundred dollar subscription. So with your model being heavy into AI, what if the usage costs, you know, jump

Bennett Maxwell

Yeah, I'm not worried about that actually at all. I was because I'm so heavily dependent. And and and they'll give you weekly resets, like especially if you're a power user. And one week I've done seven billion Codex tokens on my $200 a month Codex account. So I have two ProMax Codex accounts, three for Claude, and then two for Grok. And then with Grok, it also comes with Grok Bot and it comes with two Cursor Ultra accounts. So it's about, yeah, I don't know, 12 accounts that I paid the $200 a month. So I'm spending two grand a month.

Mikulas Zach

Uh-huh. huh.

Bennett Maxwell

But I'm getting hundreds of thousands of dollars. And that one week of 7 billion tokens, that was $200,000. I'm like, why would they do this? And that freaked me out to be like, I can't sustain this. Once they start charging and seeing what they could do and seeing how reliant on this, I'm screwed. Then the next week, this was who knows, you know, a month ago. Then the next week, Opus, so Opus five dropped, which is a really good model, but Opus 4.8 was like the best one before that. Then Quinn came out with 3.8. It's a Chinese company. And it's 95% as good as that, you know, that Opus 4.8, and it's free. It's free. And you're like, wait, how do they do it for free? You install it, I mean, you have to have computer, but you install it on your computer. The entire brain lives on your computer. So even if the internet is out, it still works. That's free. So your models, your free open source models are only about three months behind in intelligence. So whatever the cost is today, if I just want to go back three months worth of intelligence and say, hey, what was the best model back in June? That's free. That's free today. Because I have the computer. And you don't even have to have the computer. Nvidia hosts, you could get a bunch of free models online as well. There's a bunch of other things that you could do to get free models. So today's models, Fable 5.1. You're gonna have a Kimmy or a Quinn product here by November. It's gonna be as smart or smarter than Fable 5.1. So for me, it's it's very, I don't worry about it. I'm like, I'm just gonna get whatever tokens and I'm trying to token max because every single inquiry I'm hitting it with, it has to record what I'm saying and record like what I'm asking it and its response. So that way every time a model gets smarter, we can always go back and say, hey. Look at the 7,000 inquiries that me and my team asked you, you know, three weeks ago before Fable 5.1 came out. Go redo all of that work. And it can and it can keep doing that. And you can go back two years. So I call that the trail skill. It's like every time I say something, you're recording it using the trail skill. And then I could go back and widen the trail skill. And then it says, okay, now go analyze all of the conversations, what went well, what didn't go well, how can we permanently change behavior because of that? What else can be done, XYZ? So and then you also can program like recursive self-learning loops. Like before you respond to me, you have to grade your own work. If you're not at above 95%, go fix it permanently for the entire fleet. Because if you have your bots rules living in like Google Drive or anywhere in the cloud, then all of your bots are just reading the same cloud and it doesn't matter which platform you use, because I have all my employees just working on cloud as well. But we all work through the same skill. So every time my Claude does something, if if it finds an issue, it'll fix it. So if my employee's Claude does that exact same task tomorrow, it self-improved because of mine. And that happens on, I don't know, probably 70 to 80% of times we ping it, it'll make one or two little changes to a skill. So you're getting you're getting hundreds of improvements a day.

Mikulas Zach

Yeah, self s that's great. The self kind of how do they call it s self-learning cur like loops, those are great. wrapping this up, Bennett want to ask you what is a actionable AI advice that you would give to listeners?

Bennett Maxwell

Self learning. Recursive self learning. yeah, download one of those three devices. pay whatever you're gonna pay the $20 a month or the hundred or whatever. I would start with email. as soon as you get on there, it'll say get started, and you'll click connect my, it'll say summarize my emails, and you'll click it and it'll say, Your email's not connected. Click here. You don't have to leave the screen. You'll just stay in Codex or Claude or Grok and you'll click connect, it'll have access to your emails. The first prompt that I would give it is. Go look at my last 2,000 emails or five years worth of emails and build a decision profile, a voice profile for me, a psychological profile for me. I need you to fully understand how I respond to emails, when I respond to them, how it differentiates between family, friends, coworkers, employees. go build that for me. It'll be done in three minutes, literally three minutes. And then you say, Cool, save that as the Bennett email skill. And then you say, now schedule that to run. once every single hour. And it's gonna check all of my emails and I want you to categorize it in three things. Any spam or unrelated, just whatever, leave it as read, but also unsubscribe me from that. So I don't so we don't get that anymore. Anything that I should might need to be aware of, but I don't need a response, put it in the FYI category, but leave it read because I don't want to be pinged on it. Just tell me what I need to know. And then everything else, leave it unread. Tell me what your recommended response was. And then I just go through and I click accept, or I could put warmer, or I could put, you know, revision, whatever. And then you're just everything that I just told you again, probably the easiest way. If you're listening to this podcast, grab the link from YouTube, wherever you're listening to it, put it inside of your Claude Codex and say, find what Bennett said about the email skill and the email triage and and build that for me and run it for me and set it up every single hour. And then it'll freaking do it. So That's actionable. You can have that done today and it will be better than any assistant you've hired. It's gonna, you will hit frustrations and it will be dumb for a little bit, but I I'm like, like maybe 10 revisions. Maybe you have to go back and forth 10 times. And if you do that right, you shouldn't have to be checking your email to the extent that you don't even open up your email app like I don't. Cause why would I? It's like I I don't like that even like I'm I'm just thinking about all my emails. I'm like, no, ooh. You know, I just want to talk to the clean interface on Claude and tell Claude mean, Claude going to tell me because it also reads the emails and it understands context because it's also inside of my Slack. And it's also inside of my Notion, which hosts all my projects. And it has access to every recording from every meeting that I've ever been on in the last two years. So it's a lot it it thinks better than I can too, 'cause it has way more context.

Mikulas Zach

Yeah, that's super smart. Well you heard it guys. thank you Bennett for joining. My last question is, where can people find you?

Bennett Maxwell

my website Bennett dash Maxwell B-E-N-N-E-T-T-Maxwell.com. I have a lot of stuff on the agents, my social media stuff, franchising. If anybody wants to get like a free blueprint for AI agents, you can fill out like a handful of questions. But most people want to know, okay, now I kind of know what an AI agent is, but what do what's more examples specifically for my industry? And what can I do with my existing tools? So if you just answer what tools you have and and I think you just put it in your website, it'll give you the full roadmap. It's free. It'll build your first three agents that you could copy paste to your Claude, whether it's like an appointment setting agent, whatever, it'll analyze your business, your industry, and do that for you for free. So you could also grab that from the website.

Mikulas Zach

Okay, thank you for listening and thank you for joining.