A.I. Native Auto Hospitality - The Future of Customer-First Auto Repair

What's up, players? This is Joe Adams with Adams Automotive, the number one shop in America. This is Master Tech to Millionaire, presented by Auto Shop Answers, where we take master technicians and turn them into professional business athletes, CEOs. We're here today with our Chief AI Officer, Michael Floyd of the Auto Hospitality Group. We're going to talk about a lot of amazing stuff today, but I guess, well, let's get into it. How are you doing, Michael? Good. Thank you for having me. Okay, so I want to get into automation. I want to get into kind of some of the projects you're working on, where AI is at today, if it's going to steal my job one day. But I guess a little bit about your background, like how you got into AI and then how you made your way into our business. Yeah. No, I'm happy to dig into that. So I grew up in Austin and went to college for computer engineering at the University of Texas, Hookham.

Really didn't know what I wanted to do out of school was kind of debating between big tech and, you know, being behind a computer and coding all day or doing consulting and like getting to see different kinds of businesses, different industries, travel. So I ended up going into consulting and got to travel, got to go to New York City, and then COVID hit. And I realized very quickly that I didn't love making PowerPoints. I really missed like building and really thinking through like technical and logical problems. And so I decided to leave consulting and go into an AI company in Austin that, you know, was pre-ChatGPT. So this is like a different kind of AI than we'll be talking about for most of today. But just, you know, joined an AI company as a software engineer and just really enjoyed it, but felt like, man, you know, there were so few companies at the time that could really take advantage of AI.

You could think of the ones that have famously done it, like Netflix with their algorithm and Facebook with their and TikTok. Those are the companies with, you know, billions of dollars to pay the best engineers and all the data points. Like we all joke how, you know, Facebook can hear every word you say and sell things. So, you know, there was very few companies that were actually getting value out of AI. And it was even hard to see with the company I was at, like people really getting a ton of value out of it. And so I kind of started to get a little bit disenfranchised with it. And then ChatGPT came out and I was like, oh, my gosh, this is free on the Internet. You know, it's a new form of AI called generative AI. And you don't have to train your own model. You don't have to spend millions of dollars. And so that's when I was like, this is going to be like our generation's version of the Internet.

Like this is going to change the world. And, you know, that was a little bit of a crazier take a couple of years ago. Obviously, today, it's the only thing you see when you open the news or LinkedIn. Like we're filming this right now. And, you know, the government shut down the latest model from Anthropic. And so it is like the in the stock market is at a record high. So it's pretty clear that it was it was life changing. But, yeah, three years ago, I I took the leap and I started my own business and started to consult for companies in different industries, kind of married both the things I love, which is seeing in different businesses and actually creating an impact. And then being technical and really understanding AI. And so I got to do a ton of cool work for, you know, commercial real estate firms, energy trading firms, just Fortune 500 all the way down to Series A startups and worked with some of my best friends and built a company.

And then that's kind of what led me to you guys a year ago. And you can kind of talk about how that how that came to be. Yeah. OK, so. OK, yeah. So I we hung out at the beginning of January 2025. And I was reading in an earnings report for some company that they were like using AI to automate things and like workflows. And if you're if you've been to Houston or seen a bit of our concept, we talk about Moto visuals a lot. And we had this manual process for like we literally had an app that everybody paid twenty to ninety nine a year for or you could like download the Moto visual, download the video that we send to the customer. And you had to like compress it and splice it. And we do it with every single video. And so that was like a three or four minute manual process. And when you got really fast at it, it wasn't that bad. But, you know, we had one hundred and fifteen thousand repair orders last year times five minutes.

You know, I was like, man, I wonder I know Michael's doing I talked to him last month about doing some AI stuff. I don't know anything about AI, which I want. I want you to explain a little bit more about AI, like to people who may be ignorant to it. And so I called him. I was like, hey, could you like this is like a repeatable thing that we do every single day. Could you use AI tools to like automate it? And he was like, yeah, we could we could kind of explore that. And then one thing led to another. And now we have the button. And I guess I'll we'll back up to kind of talk about how you got into the company as well. But let's switch gears and let's talk for a little bit about AI in general. So like you said in 2022, ChatGPT came out. How does that have anything to do with like automating a workflow at my family's shop in 2025? And then now we're in 2026 with all the codecs and cloud stuff going on.

Like how was artificial intelligence a thing at the companies you were working for before ChatGPT? And then after ChatGPT came out, like what were the differences? And then how did you know that it was going to be a big thing? And just generally talk about what it is and help people understand that are listening. Yeah. No, I'll try to kind of give the I mean, I could do this for hours. And we actually teach a class on this. So definitely come next month. But my like very dumbed down version is so generative AI is AI that generates text. Right. So most people probably listening to this have at least seen ChatGPT by now. You know, you could type in a prompt, you know, plan me a Super Bowl party. And it's going to shoot out a list of 20 different steps of, okay, you need nachos and queso and you need to invite people. And so it kind of answers your questions. It's like a Q&A engine and it's generating that text.

And I think the best kind of analogy, it doesn't work exactly the same, but you can almost think about these AI models like ChatGPT as kind of like the human brain a little bit. You know, we ask someone a question and a bunch of things go on in their brain and then they have an answer that is kind of based on their life experiences and their things they've learned and their specialty. And, you know, and that's what kind of gives them whatever that answer is. And so these models are trained on basically all of the data on the Internet and so much more and can answer all of these questions that you need. Oh, yeah. So basically I was reading like the earnings reports for this AI company in 2025. And they were talking about automating workflows and how it's like a cheat code. And the key word was creating the self-driving company, which I thought was really interesting. And then I had talked to you about a month before about some of the things you were working on, your company that you'd launched and how you were like providing services, AI services to other companies.

And I was like just thinking internally, like, what are some things that we do every day that's like just a series of hitting the same button over and over and over again that we could theoretically automate? And one of them was splicing MotaVisuals onto the end of our videos that we send to customers. It was like an app that we had on our phones. It took four minutes every time. And we had 100,000 repair orders last year. You can do the math on how much time we were spending on just like this one task. And there's not like it's hard to like measure an ROI, like if we could automate that. But it was just like the idea, right? How can we get AI into our business? And so I reached out to you. And, you know, one thing led to another and you ended up coming into our business. But for those that are not informed, I definitely want to take a step back and kind of explain how does ChatGPT coming out in 2022.

So you said you worked in AI before ChatGPT and then ChatGPT came out and ChatGPT is generative AI, whatever that means. I need you to help me understand what that means. And how does that lead to 2025? I can automate buttons on a screen for putting videos into TechMetric at my mom and dad's shop. And then all the codecs and claude stuff we're doing now. Like how did you, one, know that it was going to be kind of like a category defining step change in technology and to bet your career on it? And two, I guess just provide some color on like AI for people that are ignorant or maybe don't have as much context that we do in our company. Yeah. So, you know, pre-generative AI, you really had to have so much money to build these models. And then you would build the model once and, you know, pray that it was good at predicting something or sending you an ad on Facebook or something like that. And then when ChatGPT came out, you know, you just get to use the best model in the world for free or for $20 a month.

And then what I realized is every three months, it just gets better. You know, they just release the next version of it and it's a little bit smarter and it's a little bit smarter. And that's when after I think two releases, I was like, OK, I think, you know, this seems like it's not going to slow down. And eventually, right now, it's fun to use to plan a Super Bowl party or something. But eventually, this is going to be the most important thing for a business to use to make their business better. And so, generative AI, you know, I can kind of give my two-minute spiel on it. We do a whole training day on this. And I've done this for tons of different auto repair shops as well as other, you know, companies. But my two kind of two-minute slimmed-down version is essentially, you know, I like to think of a generative AI model. You know, if you've seen ChatGPT by now, you know, you type in a question, it sends out an answer.

So, it's generating text. That's all that it means. That's all generative AI. It sounds so fancy. But it just means, you know, it's generating some sort of output from your input. So, if you've used the image models, too, it can generate an image. But it's, and how it works is it's essentially trained on the entire internet. So, I like to give this example of, like, think of yourself, you know, taking a test in high school. And the AI model, so when we say, like, the new model from ChatGPT is out, it's basically like the brain. And so, when you take a test, you know, you might study, you might go to watch YouTube videos, you go to class, you go to office hours. And you're learning more and more. You're not memorizing everything, but you're learning more and more. And almost like your IQ is getting better at whatever task you're learning. And then that is basically what you use to get a perfect score on the test.

You're not memorizing every word of the textbook and regurgitating it. But you're basically getting a better proficiency around it. And then you can do better on whatever subject that is. So, they're basically training these models or these brains on the entire internet so that it gets better at everything across the board. From diagnosing a car to, you know, something, you know, planning a party or something like that. And then basically, these models have what's called a context window, which on a test, you can think of as a cheat sheet. And so, the more information we can put on the cheat sheet, the better we can, you know, have it memorized and go and remember on a test. And so, when you combine basically smarter models and bigger cheat sheets and then access to things like the internet or, you know, for example, for us getting access to things like Techmetric or Google Ads or anything like that.

You know, really, the goal is to just set these brains up for success. And once we kind of saw it getting better at using the internet and using different tools, we saw like, okay, we're going to be able to, you know, really make people's lives easier and use this as a part of our business to make it more efficient. Okay, that's super interesting. So, one thing that I've found helpful for me to understand AI and how it works in general is like the context window thing. Yeah. So, if you just ask ChatGPT like a question about how to run an auto repair shop, it's just going to spit out junk. But if you give it all the context of our business, it's going to give you much more sophisticated answers. So, like the easiest, you know, for those that may not have as much knowledge about AI or how it works, like I love the human brain analogy because it's like, you know, an adult, if you've got a hot stove and it's red, you know that it's hot.

And it's like, don't touch it because it'll burn your skin. But like a four-year-old, you know, or a three-year-old, they literally don't know that. And so, like, they're going to, everybody's done it. Like, when you're a young kid, you touch something and it burns you. And then it's like in your context window, in your brain, you're like, okay, never doing that again because that was a horrible experience. And it's like, check for the rest of your life. Another example is, you know, and this is in your section in our AI training, which everyone definitely needs to come to in July. It's on the Friday before Key2Key in Houston. You know, basically, your section on how to get more out of AI is, it talks a lot about giving as much context as possible. And so, the anecdote that we use is, for a shop owner, is everybody's got a shuttle, probably. You know, you take customers to and from their house. And so, you know, a human who has like a 20-plus-year context window adult brain, if you told them, hey, we take customers to and from their house at the shop,

they would generally understand that that doesn't, like, there are edge cases. Like, if you in Houston, Texas got in the shuttle and you were like, hey, my house is in Florida, the shuttle driver would be like, okay, well, I'm not going to drive you to Florida. That's like a 20-hour drive or a 10-hour drive. But an AI, you know, like a super brain AI, it might be like really excited to work and it might work 24-7 and it might be like the smartest thing in the world. But it would just take the command and it would be like, okay, yeah, I'm taking the customer home. And it would get in the car and drive. And so the more that we, as far as I understand, and correct me if I'm wrong, the more that we can basically point AI, which is like, think about it like a human. Like, they work 24-7. They're the smartest person in the world, you know, and they just, they want to work, you know, it's like, and you can have 100 of them.

The more we can point them at, you know, a business model or at like data or at a, you know, like Todd's concept is how we like talk about it, the more effective you can be. So do you want to talk a little bit about, you know, Todd talks about this a lot. He's like, everybody's trying to use AI, but they don't have a business model, you know? So do you want to talk about how that kind of clicked in your brain before you decided to join us full time? Like, you know, it's not just AI, it's fine. It's marrying AI with somebody with a lot of like industry knowledge or wisdom or experience or stuff like that. Do you have any thoughts on that? Yeah, no, I think another analogy we use is just think of it like an employee, you know, it's your smartest, most eager employee, but it's their first day. Yeah. And so you have to really guide them, you know, they have to get access to the tools you use. And that is how you can set them up for success.

They can't just come in and magically read your brain. They are getting smarter, but they're not reading your brain. But before you go on to the next one, it's like opening a ticket in Techmetric and like closing the books or like closing a ticket or checking in a customer. Like a human, like you have to teach them a couple times how to do it, like just like an AI. Like you wouldn't get a human on their first day and be like, oh, they suck. It doesn't work. Yeah. You know, like it's not working. Like I see people all the time with ChatGPT, they're like, oh, it's wrong. You know, like it's not working. Yeah. It's like if you hired somebody and you just told them to go to work, they would do stuff wrong. It makes no sense. But after like literally one day or two days of them opening a repair order or closing it or uploading a picture into the inspection part of Techmetric, like they get really fast at it.

Yeah. Because that's how it works. So anyway, I hate to interrupt, but I love that analogy, thinking about like a human. So continue in talking about how marrying it with industry wisdom and experience and stuff. Yeah. So, I mean, what we've found, especially as the models get smarter and better at using tools. And when I say tools, I'm saying the Internet to look things up that they don't memorize. I'm thinking of Techmetric. I'm thinking of, you know, Google Ads and all the different RingCentral, all the different things we use. It's tools and then it's those steps. And so the fact that, you know, this business, you know, has the auto hospitality model and it is just, it is so well trained and so well kind of laid out for how we run this business. It actually makes using AI so much better and it, you know, kind of increases the probability that we'll go from, oh, this is a helpful tool at helping me automate a certain task to this can fully automate a certain task.

Because the way we coach on, you know, doing things is like very detailed and very disciplined. And so by having that and having access to the data and kind of the standard operating procedure is a perfect marriage for AI because every three months it gets a little smarter and we'll get closer to automating more and more tasks that, you know, we don't want our people to be bogged down with like the button. Okay, so a couple things. People generally, people get kind of scared of AI because they think it's going to like take their job away or, you know, it's going to, they're just, they just don't really know what it is. So it's like, I don't even want to dip my toe in the water, but I want you to, so continuing on the thought of like marrying to, like you couldn't just go open an auto repair shop because you have software engineering experience. Right. Like you could use AI to teach you how to like file the right paperwork to open an auto repair shop, but like you're not going to learn how to fix the car or, well, I don't know how to fix the car, but you're not going to, you don't have the operating procedure.

Okay, so I like to tell anecdotes because it makes more sense. So like what are a few, a handful of workflows or use cases in our business that you've utilized? Like, okay, here is how the executives and the management team think about the business. Here's how they think about the data. Here's the questions they're trying to answer. Here are the tasks that we do every single day in the business. How now that we have this superpower AI, chief AI officer, how do we point your talents into the operating model to make it more efficient? Like, do you have any use cases you can think of? Yeah, I think like the simplest one that kind of touches on both of those. And this was one we implemented immediately a year ago and we're kind of way more advanced now. But it's just, you know, what makes our model auto hospitality so good is just the way we answer the phones and the way that we sell jobs. And we have very specific scripts on how to do that.

And so we built, you know, basically a tool in ChatGPT that can take any sort of, you know, thing you're selling, a water pump, break job, whatever, and put that into, you know, because we have those scripts, actually tailor that script to whatever that part is and then to whatever that customer is. So we're kind of combining our standard operating procedure plus the data that is, you know, what we're trying to sell. And that has allowed us to have consistency across all of our shops, you know, and we've grown since we implemented this to basically answer the phones. Like, it's just even better, like even more consistent. And so that's a very simple example of kind of marrying the data plus the standard operating procedure and AI just, you know, reducing friction there. Yeah. So that's kind of like where it started. And you said it last summer, like very simple, like, okay, we read scripts to our customers when we're presenting work, total investment.

And then we were able to create kind of out of the box ChatGPT tool to help new service advisors be able to like build a custom script immediately. Okay. So fast forward to kind of some of the things we're doing now, and maybe we'll talk, maybe, maybe this isn't the best setting to talk about Codex and Claude and all of like the tools we're using now. But one of the things that we are using those tools for is connecting basically, actually, let's just rip the bandaid off. Let's talk about it. So Codex is, as I understand it, it's ChatGPT, but it's on your desktop. It's on a laptop or a computer. It's not on your phone, like in the cloud. What most people that are listening to this probably understand is ChatGPT on their phone, correct? So it can be, but I think the best way to explain Codex is it's ChatGPT, but it can access everything on your computer. And it can run for longer and longer amounts of time, which allows it to do harder and harder tasks and validate its answer, which is the big thing.

Because, you know, it basically checks its work before it gets in front of you. So it wastes less of your time. You know, everyone hates to get a response from AI that's wrong. And so by Codex being able to run for this long amount of time, we're just getting much more reliable responses to whatever sort of task we're trying to do. Okay, so let's go backwards to try and explain how Codex works or something like this works and where it's going. So, you know, two years ago, three years ago, you asked ChatGPT to write a poem for you. And it's like, that's pretty cool. And now, but if you asked it, like, if you just asked out of the box ChatGPT, like, what's my best record? What's my, who's my best service advisor in the shop? Like, obviously it wouldn't know. Like, it would be like, well, you'd have to tell me who your service advisors are. Yeah. But now we have Codex that is basically ChatGPT advanced mode and it's on your laptop and you can use it to write code to connect to your tools is what you're saying.

And your tools, think about a human, think about a human and what tools they use. They use TechMetric. They use RingCentral. They use Gmail. They use Ramp as our credit card. They use QuickBooks or NetSuite or whatever our ERP is. And you can basically ask ChatGPT, hey, go connect to that tool that has all of that context and data in it and report back to me what you find. Instead of having to, like, dump, you know, what I was doing two years ago is I was downloading spreadsheets out of TechMetric and putting them in ChatGPT and trying to, you know, make decisions out of it. So, an anecdote to tie it all up is, like you said, the phones. Where we're at today versus last year on the scripts is I can point the AI machine gun. I can point it at TechMetric and at our phone system. And I can actually have it download all of the phone calls and transcribe them and run them through ChatGPT. Which one of these are, you know, customer opportunities?

Which one of them are calling for a break opportunity or an oil change or a state inspection? And how well are we converting them? And then it goes into TechMetric and it says, did this phone number show up? Is it a customer that spent money? And I can actually very fluidly get data out of both of those systems now where I literally couldn't do that two years ago. Two years ago, you know, and for 40 years, Todd has taught the very, the number one thing in our business is the phones. Listen to your phone calls. Make sure your customers are experiencing customer service, hospitality, and make sure your service advisors are answering the phone with the anytime presentation. Well, what that looked like even two years ago was go into your VOIP software, download like 10 MP3 files. You know, that takes 10 minutes because the internet's slow or whatever. Manually listen to them and there's no transcripts.

So you just have to listen to the first one. Oh, that's a vendor call. Listen to the second one. Oh, that's a really good customer looking for an update. Listen to the third one. Oh, that's a sales call. We can coach on that. Let's save that for later. Oh, here's somebody listening to the anytime script. And you get one data point. And then you like email it to your manager and it's like, okay, we can coach on this tomorrow in our morning meeting. Whereas now, I can literally have AI listen to 10,000 calls all at the same time. And like, hey, Codex, go and find me 10 calls where we dropped the ball and didn't convert the customer. You know, and then it just works for 30 minutes straight, which it couldn't do two years ago. And it brings them back to me. It's like the most, it's the most crazy thing in the world. So I'll get off my soapbox and hand it back off to you. But that's like one use case that I'm seeing a tremendous amount of value.

You know, do you have any other use cases that we're kind of automating that might open the minds of some shop owners to AI? Yeah, I think I love what you bring up with the kind of the calls all the way through to, you know, the money in our bank account, essentially. Like a click on Google turns into a phone call, which turns into an RO, which turns into, you know, a mix of approved and declined jobs. And then that hits the bottom line. And so for us to be able to kind of connect that vertically with a tool like Codex has given us so much more kind of insight into the leads that we're generating, which, you know, one of the biggest things about the concept is, you know, you get more people to show up and you increase your ticket average. And so I think the inbound call center story is just super powerful where you had listened to some calls manually and just had this thesis around like, man, you know, our service advisors are amazing, but they love to sell jobs.

They don't love as much to get cars to show up for an appointment. And, you know, if we could kind of train people to be specific for just that, you know, could we essentially increase that ticket app or could we increase that show rate? And so we took that thesis and had Codex run for pretty much days at a time, accumulatively, and kind of prove out that theory that, man, if we transcribed 10,000 calls across our shops, then from those calls, we figured out who actually showed up and then what was their average ticket and on different types of jobs. And so we were able to see, man, we think there's meat left on the bone. We're well above the industry average already, but there's meat left on the bone. And so that allowed us to build basically a platform that our inbound call center team uses to essentially, you know, raise our, I mean, it raised our booking rate by 50%, which is just mind blowing.

Yeah. We could not have done that without AI and Codex. Yeah, it's actually like Charlie, Charlie said to me on the phone the other day, he's like, Joe, I don't need to look in a computer. There are cars everywhere. It's like, it's working. Anyway, but the number you're talking about is, you know, the calls that roll over to the shop floor have a 42% arrival rate. And the calls taken in the call center have a 61% arrival rate. And so it's like, bang. Like that is data that we are, that is coming into our systems that we're pointing AI at and we can validate with the tools. It's like, it's gotten so granular that now we've got it broken out into buckets of lead type. I don't even know if you know this. It's like, how many calls across our platform were oil changes or state inspections? How many were tire estimates? How many were breaks? How many were alignments? How many were second opinion from dealer? That's a really good one.

And then connect it to, so how many were there? Our number one lead source is definitely oil change. It makes sense. How many of those showed up? And what is the ticket average? And so here's a good one. Second opinion from dealer. If you're a shop owner listening to this, you know that that's like a money ticket. That's like, you want to get that customer in. And the data suggested that we had like a $1,700 ticket average, which makes sense. It's a broken car. They're a customer who's accustomed to spending money at the dealership, who has higher than average prices. And they're looking for a second opinion, which means they already are having a bad experience with another company. And we had the opportunity to earn their business. And we found that we had a low booking rate. And so it's like, okay, what can we, we had a low booking rate, but a high ticket average. So what kind of initiatives can we put in place in the business to get that, like you want to get that as high as possible.

So we implemented, we just do 15% off, you know, just as a first time customer. We'll just give you 15% off that dealer estimate and then we'll get you in the door. And oftentimes it's not even that. It's a completely other thing. But what we found is it doesn't matter. We just have to get the customer in the door so that they can experience hospitality and experience the systems in the business. So it's really, it's, it's, it's really like alignments. You know, we have a lower booking rate on alignments. So it's like, okay, we're looking at, are we charging too high for our alignments? Like our customers just more price conscious and they push through the anytime presentation harder. Should we look at that? And it's, it's less about, you can get analysis paralysis and looking at all the data, but it's like, think about the business a lot. Or this is how I use it at least. Think about the business a lot.

Where are there areas to improve? How can I use AI to get unstructured data out of the phones and out of tech metric and convert it into like an actionable thing that drives car count or drives ticket average or whatever. So, oh, I love it. I'll pause there. Is there anything else you want to touch on, on, on just systematizing the business in general so that it becomes AI native? You know, like Todd did such a good job systematizing literally like every objection, you know, every type of car breaks the same way. Like, how do you handle a new customer introduction? Like that allowed it, you to infect the company with AI and make us AI native. Do you want to talk about like why that's important? Yeah. No, I think what we just talk so much about is we're, we're not, we don't think AI is some silver bullet. That's just going to replace fixing cars or replace the hospitality aspect of the business. Like what differentiates us from the shops down the street is our outrageous customer service.

You know, it's the white glove service. And so we're not trying to, hey, how can we save, you know, some money on the bottom line by automating away the conversation with the, with the customer. Like that is not something that makes any sense for our business. And so where we see so much value out of AI and being AI native is just adherence to the concept. Yes. So it's extracting all that data. For example, you know, some other low or like, you know, use cases we have simple is just like AI call scoring. So not only do we generate the script with AI to be specific to what they're selling, but then we can score that conversation to know, did they talk about our three year 36,000 mile warranty? Did they do positive negative selling? Did they do the things that Todd has proven over 40 years? If you do these things, you will be kind of a, you know, a, like a very above average shop in America. And so by us being able to extract all this data and put it in the hands of people like you on the management team or shout out Mike Quinn, you know, or Glenn as GMs or even the service advisors who are using these tools on a daily basis as well.

You know, we can give them access to this data so that they can adhere better to the concept. Yeah. So another example is, um, so like the concept is actually just a series of steps. Yeah. And if you follow the steps, you will have a higher ticket average is the idea. And Todd has spent 40 years trying, you know, desperately and successfully getting people to take quality pictures and videos, you know, read the scripts and do positive negative selling. You know, did you, did part of the script is, did you watch the picture and video? And so how are you using AI to like, you know, Todd's got this thing called the board and bag. Yeah. So one of the things that we always look at is like the quality of the video and the quality of the little text that you sent to the customer. And there's just a hundred thousand repair orders. Now I can't look through every single one and there's context to every single one, but we can, you know, what we would do previously is literally click through random ROs that didn't sell.

It's like, you came in for a check engine light, you know, you came in for breaks and you didn't get your brakes fixed. Why? Like Todd, most people are, they have external locus of control. It's like, well, they didn't have any money or, oh, well, he wanted to talk to his wife or, oh, well, you know, he's just, you know, he gets paid next week. And, oh, he just didn't want to do it. Or like, he's going out of town on a trip. Like they just say stuff. Whereas Todd has always been like, no, no, we dropped the ball. Like, where's the issue? And it's probably a series of steps that was missed in the presentation. So you want to talk about how you're using AI to kind of be a second set of eyes for people like Mike and Glenn. Yeah. In TechMetric. Well, I think we definitely want to use out of the box technology. And I think if you're listening to this, not everyone is going to have enough shops to have a chief AI officer and be actually building a lot of these tools.

So even for us, we would rather just use something off the shelf that's amazing and gets the job done. So there's a lot of great out of the box, you know, technologies like Rilla who help us with call scoring and call tracking and call coaching. But, yeah, we've been able to really like, because, you know, of the investment made and just making this a really AI native business, we have people building things across the company. Our head of our call centers, Gary and Josh, you know, they basically build tools that fit their workflows. And so, for example, the board bag audit, which Joe talked about, this is really the kind of the 50,000 foot view of are we adhering to the concept across our shops? You know, we're not shy about the fact that we are a growing business. And so our best people that have made this business what it is can't be in every store every day. And so we're able to essentially plug into all this data across all these stores and surface the things that, you know, need to be looked at.

Like you mentioned, so we, you know, this was something we built just in this last couple days is basically, okay, we score all of our calls, but what about our videos? And so now we're transcribing all our videos and then figuring out which ROs did we have declined jobs on? And those videos get surfaced to Mike and to Todd so that they can actually see very quickly, you know, again, we're not taking the human out of the loop here, but we're trying to make it less or reduce the friction to them being able to see into that business and see, okay, you know, we tried to sell a leak and the car wasn't even on, so you can't even see it leaking. So of course, the close rate on that job is going to be lower. And how do we essentially, you know, do that with everything, whether it's how we fill out tickets, how we build the videos, you know, everything in adherence to the concept. Like how do we surface that information so that Mike can then go and coach or Todd can go and coach.

If you were a shop owner listening to this and you're like, oh my God, I can't keep up. Like what, what would you do? Like if you're just a shop owner with one or a couple shops and you just, all you do with AI is like you have Chachupiti on your phone and you're feeling probably like, okay, I'm behind the eight ball a little bit. How do I keep up? Like where would you start? First of all, yeah, I would say it's definitely not too late. I think so much of LinkedIn and so much what's said online is just, you know, they realize if you just say AI, their stock price goes up and it created this kind of ecosystem of like everyone just has to say they're doing everything with AI. Okay, let's say you're a shop owner right now and you're listening to this and you're like, man, I have one shop. Like this is scaring the crap out of me, you know, like, how do I, how do I catch up? You know, how do I, sorry, I lost my train of thought.

Let me, let me, let me start over. Ready? Okay, let's say you're listening to this and you're a shop owner and you've had one shop for 20 years or you have a handful of shops, but you're, the extent of your AI knowledge is like you have Chachupiti on your phone and you play with it sometimes and you use it as like a pseudo Google in a way, but you're not really applying it in the business. You feel like, you know, you should. I feel even with as adept as we are and having you on the team, I feel overwhelmed sometime. Like, are we doing enough? You know, but a lot of people don't feel that way. Like, how would you, if you were in their shoes, how would you kind of like attack learning the language of AI? Yeah, no, I'm really glad you asked this because I think we're talking about all these concepts and models and all these big words. And I think there's always, like, it's not as scary as you think.

And I think also just the world has kind of realized like, oh, if you talk about AI, you know, your stock price goes up. And so, it just created this mass, like, everyone's just saying how much they're doing with AI. And I think the world really isn't that far along to like, you know, like we haven't automated a huge part of our business and we use this, you know, religiously. And so, I would say don't be discouraged. And, you know, I would honestly say take the transcript of this podcast. It's a good practice of taking that data, put it in chat GPT and say, hey, where can I start? But my answer would be, yeah, I think just looking at the out-of-the-box tools, I think, is a great place to start. And, you know, whether that's Arilla or what Techmetric has or RingCentral has AI, just seeing where you can like turn things on to start to get more value out of the AI. And I would not expect you to go and build all of the tools and things that we've done.

I think what's been so amazing about this company is it's just, it's growing. And we really, you know, we're not kind of shy about, you know, our desire to grow and to have our concept be used in more and more shops across the country. And so, the reason we're doing this is, you know, because we want to be able to support hundreds of shops, you know, down the line. And how do hundreds of shops adhere to the concept? And I think that required us to, you know, kind of go all in on AI and really make sure that we're building the things that we're supposed to. So, kind of a shameless plug, if you want to use like the best AI technology in this industry, like, you know, come to class, come to AI training next month. And just, you know, reach out to the auto hospitality group and kind of start that conversation of getting to know us. Because so much of what we think about with AI isn't, oh, I can, you know, automate this manual task.

It is, how does this allow us to make, you know, everyone's heard of the Blalock location and the crazy record numbers they've done. How does every shop in America have access to the same or every shop in the auto hospitality group have access to the same tools, data, and also like the ways that Blalock or some of our great shops are running so that they can run things the same way. And we're really building toward this, okay, right now it's a lot of, okay, get the data to adhere to the concepts. But eventually, as we, you know, we keep betting that AI is getting better, the money is still being poured into the big companies. They're about to IPO and we'll see what happens there. But we believe these will continue to get better and we'll be able to make life easier and easier for service advisors and technicians and shop owners who are on our platform. Yeah, the thing about, like, if I was a shop owner and I just had little literacy and AI, I would just, if I were in your shoes, what I would recommend is, you know, information is free now.

Like, it's like there's no friction to getting information and the cheat code is to just ask the AI again. You know, just ask it again. Just keep asking it. Like, for example, 40 plus years ago, Todd, you know, he had to get the knowledge out of his father-in-law's brain. Like, it was kind of gatekept, you know, like how to have a business model and how to run a business and how to, like, what an earnings multiple was. Yeah. He tells that story all the time. Like, you know, get audited financials. Like, if, you know, the earnings multiple thing, like if you leave your money in the business, it's worth a multiple of your earnings. If you take it out, you know, you have to pay tax on that, you know, to go spend it on stuff. I like to keep as much as I can in the business, yada, yada, yada. That was literally context that was, like, you know, just in his father-in-law's head and he had to get it out in order to start his business.

But now it's like we had the internet before, but you couldn't, like, talk to Google for three minutes about some complicated issue in your business like you can basic chat GPT. And so I would say, you know, and following up on the platform thing, like there are, like 40 years ago, there weren't best practices groups out there. And now Autoshop Answers literally exists. And you can, like, one, you can just keep spamming the chat GPT button. Like, keep just using the tool as much as possible and trying to teach yourself. But number two, like, shameless plug, like, that is kind of the whole idea of the platform. The whole idea of the platform is like, hey, if you partner with Auto Hospitality Group, like, we're going to lift your numbers. Yeah. And that's it. And we're going to experience arbitrage as a platform. That's, like, the whole playbook. So what happens when you come on the platform is inbound call center immediately.

So you're going to get more cars. Like, we have the data internally to suggest that, like, if we're answering the phone with a consolidated service professionally, not only is the phone going to ring less at the shop floor to allow us to be more productive and talk to customers in person and get better service, but two, you're going to have a higher booking rate because the guy answering the phone is a pro and he's not busy talking to seven other customers looking for updates, you know? So you're going to get more cars. You're going to get more access to visibility in terms of operationally, are my shops, is my shop or are my shops executing concept? Like, do I have access to a tool where I can look and see that all my team members are putting videos on the inspection reports? Like, are they adhering to the scripts? Are they selling in the way that we want them to adhere to? And then the last thing is just learning from people like us.

So I would just stick around the concept as much as possible. Stick around people that are, like, that's kind of the cheat code, you know? Like, having a consultant. Todd was a consultant six years ago. Like, trying to figure something out by yourself is always going to be harder than just, like, going and finding somebody else that's figured it out and copying them. You know, one of Tommy Mello's rules is success leaves clues. And so it's like, we've stolen a lot of stuff, including the inbound call center from his company, from their industry. And it's like, just clone them. Just learn from somebody that's doing it in their space and try to do it in your space. So let's get back on AI. It's moving really, really fast. Like, we were talking about in the car on the way over here. Three years ago, it was like, write me a poem, you know? And then, like, words come out. And then a year ago, the deep research function kind of was getting popular.

And it's like, oh, my gosh. It can, like, Google 100 websites and, like, bring back report, like, data. Like, that's cool. What can we use that for? And I use it for a lot of stuff, you know? But it's, like, you know, it's, like, kind of growing in the direction. And then in December, right? Yeah. Do you want to talk about how it, like, learned how to work for 20 minutes straight? And now it's working, like, overnight. Like, I'm sure you have prompts running right now. Yes, I do. I want to hear what prompts you're working on right now. But do you want to talk about, like, how fast it's moved in just three years? And then dovetail that into, like, where you think one, two, five years from now is, both in the world and in our company? Yeah. No, I love this question because this is, when I'm done with work, this is what I go read about. I'm not on social media. I'm, like, okay, what came out today and where can we use it in our business?

And I think, yeah, it's a blessing to love what you do. I think I see that a lot here. But, yeah, to answer your question, you know, it's always kind of you could see just, you could see the vision where, you know, if it could generate all this text. Like, you could have asked it three years ago, like, tell me how to run an auto repair shop. And it can kind of, like, hit the right, you know, bullets. But it gets too many things wrong and it's just, you know, it's too surface level. And so, but you could see the kind of, the, like, you know, the vision was there. And so, over time, it's just gotten more and more reliable and just more and more thorough. Like, it's almost like, you know, your IQ, like, as you're growing up, you just get smarter. Like, some of it is the learned experiences and the access to tools. But some of it is you just get smarter. And so, these models have just gotten smarter and understand just, you know, the world better and how to interact with it as a, like, in business.

Which has allowed it to get all of these new capabilities. And that's kind of how we think about it as a business is, you know, we track all these different use cases across our business from, you know, turning the customer concerns into, you know, the call script and all these things. All the way to more, you know, complicated use cases like doing a three-way match for accounting. And, you know, not every use case is at a good point to where it's automated. But as every three to six months when a new model comes out or, you know, a new way to use the model comes out, we see some of those use cases become, they get to that threshold where we're like, wow, we can actually offload this. This makes our service advisor or, you know, the, like, data analysis that we talk so much about, that makes Joe's life so much easier. He can look at the business in so many different ways with the same amount of hours because of that threshold that was hit.

And so, you know, in 2025 at the very end, like the kind of breakthrough that happened was the models from OpenAI, which is ChatGPT and Anthropic, which is Claude, became reliable enough to run for 20 to 30 minutes. And we've been seeing this in software engineering for even longer. And that's, again, because that's my background. I kind of see that and I'm like, man, how can that, how can, you know, it run for 30 minutes coding? How can it run for 30 minutes on Todd's concept, on some sort of workflow in Todd's concept? And so, you know, every time a new model comes out and, you know, again, we're in that, we're in this phase where we're filming this, where like the government just shut down the next big model, which is called Fable, that people are raving that is the next step change. And so we're about to find out whenever, you know, they unban it, does that then allow us to get to a place where we can fully automate some of the, you know, AP tasks that are manual and way down on our controller and our accounting team?

Because, you know, across the business, we're so AI native. We have people using, you know, it's not just me as a software engineer using Codex or Joe as someone who gets so much time with me using Codex. It's, you know, people across the entire business from the back office to the front of the house to the back of the house to the call center using this every day, figuring out what's it good for, what's it not yet good enough for, and really just building like that. It's legitimately a new skill, like riding a bike. You know, you fall to the left, okay, lean more right, fall to the right, and then eventually you get it. But it's a big feel thing with these models of, I can kind of tell like, oh, this is probably not the right, you know, we're not sending our P&L, it's not doing our P&Ls right now, because we know that's way too many manual steps to give to AI. And there's great systems out there that can help us with those P&Ls.

And so, you know, where I see AI going is just continuing to get smarter and continuing to get better at using the same tools we use on a daily basis. For example, we're not just giving a spreadsheet to AI and saying like, okay, make our, you know, match this with all the tech metric data so that we know every dollar that came into our business is accounted for. We're having AI connect to RAMP, which is a great AI company. We are having it connect to NetSuite. We are having it connect to QuickBooks. We are having it connect to TechMetric and use those tools the way that we use them. And that is really what I think a point that I want to get across is the AI is not necessarily just a black box that just outputs, you know, the automated workflow. It doesn't just output the P&L, but it uses NetSuite to use the P&L. And you'll start to see that. I think that'll be more and more mainstream, you know, and people will start having that aha moment that they'll remember of like, oh, I see what he meant by, you know, we're not asking AI to do the impossible.

We're just asking AI to use the same great systems that we use on a daily basis to make our lives easier, to essentially be our, you know, interns or kind of new grad employees. It really is crazy. Like, it's like a, it's like a person. It's like raising a kid. I've never raised a kid. So I can't speak from experience. But it's like, when you're a baby, it's like you have to be like constantly monitored 24-7 to like make sure that everything's okay and everything's working properly. And then you get a little bit older and like you can leave them for a little bit, you know, and then they still are there. It's a great example. And then it's like, you know, they go through grade school and like you can, you know, go to baseball practice and you can leave the kid and you know that he's got like the skills to like survive at baseball practice and he like knows how to hit the ball and like engage. And then it's like they get their driver's license and they can kind of like go drive around town.

And I remember being 16 and like driving to College Station for the first time and it was like an hour and a half. And it's like I wasn't allowed to like go further out, you know, because I probably wasn't like the model in my brain. It wasn't like developed enough, you know, to be like where it's guaranteed to my parents that like he knows how to get home safe. Yeah. You know, and then I remember like being 18 and like flying out of the country for the first time. And then like eventually you become an adult and your parents like it's they don't they don't they don't monitor your activity. You know, it's fully automated. Yeah, it's like totally fully automated. And it's like that's a 20 year process. But like think about all the context that's in my brain from 20 years. It's like how to take care of yourself, how to make food for yourself. Like a six year old doesn't know how to do any of that. They can't drive a car.

They can't make doctor appointments. They can't like make full sentences. But eventually they can. So, you know, it's kind of the same thing in accounting right now, like in all the business parts of the business. But you mentioned back office a handful of times. So I want to follow up on that. Like the back office particularly is less of the Wild West because like it's such critical. Like if the AI messes up in the call center, it's like fine. You know, like the service advisor can still like is a human and human in the loop and they can talk to him. Whereas if I'm paying an invoice and I get a decimal wrong or if I'm doing the P&L or if I'm making payroll and I get a decimal wrong and I pay somebody, you know, $10,000 instead of $1,000. That's like a huge mistake. You know, and that's like that's like the flying out of the country. You don't let a six-year-old do that. You might let them like try to ride their bike down the street, but you're not going to let them like, you know, do things that maybe an adult you need an adult in the loop for.

So do you want to talk about more like kind of the risks we're seeing in the back office? And then, you know, where do you think the most leverage is in the back office and where it's going in the next couple of years? Yeah. No, I think, yeah, we really try to assess the risk of every use case, just like you said. And so we can be more aggressive with ones that are lower risk. Like it, you know, it makes a lot of sense. I think the back office, you know, this is so important to us because one, we want to attract amazing business owners who want to join the auto hospitality group. And for us to be able to make their lives easier with a centralized back office that is AI enabled and AI native is a massive selling point. You know, you can talk to Charlie, you can talk to like the Slacks we get and we actually run our entire company off Slack. Why? Because Slack is a very AI native platform. It's what OpenAI, who makes ChatGBT, it's how they run their business.

So that's why we decided to do that. And so we can, you know, send things into Slack, take things out of Slack, because we're trying to mold our business to be as AI forward as possible so that when the next model comes out or when the government, you know, unbans the model, you know, we're ready to take advantage of that. And then the owners that are part of the auto hospitality group have peace of mind that we are doing as much as we can. And so I think just from the back office, I think, you know, this business AP and the amount of volume our shops do and the scanning of invoices, you know, for example, scanning an invoice and actually having like an AI read it like a human as opposed to just like taking the text out of it is something that was literally impossible three years ago. And so now we're able to take the next best model and have it read that invoice. And, you know, it's not perfect.

And I'm sure everyone knows here there are some I've seen some egregious looking invoices. Yeah, it's crazy. Where there's just like like Sharpie, like if you can't read it, the AI is not going to probably do better than you. So it's not going to be perfect, but these are examples of use cases that as it gets better at reading that invoice or making that next decision of is this something that needs to get escalated to the controller or to the CFO or to the service advisor to fix because they clearly fat fingered, you know, a parts number. You know, those are things that it doesn't automate our entire back office, but it makes it much less frictionless and it allows the people in the back office to think more about, you know, the the more important questions that drive the business of, you know, are we getting the best deals for our parts and are we consolidating purchasing well enough? And, you know, there's so many things, higher leverage things that they can do that AI will free them up for.

And it's already happening today where they still, you know, I think a big thing to mention is like a human still has to own the outcome. Yeah. So if AI today could perfectly put out a P&L, our CFO would still have to own that outcome. And if it's wrong, it falls on our CFO. So if we can automate it, great. And then he can focus on other things. Strategy. But, you know, until we can confidently have it to where those humans are comfortable, you know, letting like they still have to own that outcome. And so anything in the back office, which is higher risk and affects our entire business, you know, we really need AI to get to that level where, you know, we can automate things, but still feel confident in the humans who own that outcome that it's just as good as it was when it was fully done by humans. Okay. Just for some context, too, on the accounting and like looking at risk tolerance, like one of the things that I learned recently is Matt, you know, Keys, our CFO, superstar.

He was like, yeah, our business and accounting is pretty much 98% accounts payable. Like we order a lot of parts and we have to pay a lot of parts bills and make sure that they're right. And about 15 to 30 times a month per location, there's an instance where it's a 90% chance that the service advisor put the wrong number into Techmetric. And then the invoice shows up and it's different. And then the numbers don't match. Okay. And so it's like, and so we just were riffing one day and it's like, okay, well, what if theoretically we could put the number in right every single time? And Matt was like, well, then I would, I would free up like 80% of the manual labor, like the fires that we're doing in the accounting office. And so it's like, okay, how can we use AI to, to, to kind of lever, be leveraged in this situation? We first thought about getting codecs to just automatically input the numbers off of the invoice directly into the system so that it would theoretically be a hundred percent accurate every time.

And it's not there yet, but we'll get there, but it's like a perfect. So what we're doing is we're literally scanning the invoice and it's picking out the quantity, the cost and. RO number. RO number, right. And then it's inputting all those into tech metric or no, no, no. Then it's a scanning against the numbers in tech metric every day before we close all the tickets. Okay. And that way it's allowing us to basically be a second set of eyes. It's not inputting anything, but it's double checking at the end of every day to make sure that none of the numbers are wrong. And then if the numbers are wrong and something has been posted and it slipped through the cracks, now we've got an agent that can like, that the controllers in the accounting office can ping people on the shop floor or managers. Like, like think about it. They were having to manually type out emails and there's like sentiment and you want to make sure that the people on the shop floor don't resent people in the back office because they're nagging them.

And then the people in the back office want to make sure that they're not like disrupting operations and they get frustrated when operators are not like putting in the numbers right. But it's kind of like a, you know, everybody thinks they're contributing more to the pie than maybe they are like as a collective. And so there's like human parts of that that can be automated, which is like, oh, you could just have an AI agent automatically through Slack. That's why we switched to Slack, you know, stuff like this. Can we automate the poking basically of the general manager? Like, hey, you posted a ticket with the wrong number. You need to unpost that and you need to fix that so that we can reconcile on time. And then it just automatically sends them that thing every day so that it's a thing that's offloaded from Sadie or somebody in our accounting department. So we're just getting started. It's 2026.

We're almost done with the podcast, but I want to hear your, I want to hear your like your, your clickbait doomer or are euphoric? Like, are you a utopia guy? Like 2030, you know, like five years from now plus. Like, where do you think, like, what's the clickbait? Like, what is like your, what do you think is happening, man? Where are we going? Are we just going to automate computers away? Am I not going to need a computer? Like, what's going to happen? Gosh, it's such a hard, I mean, people talk about this stuff all day and I just, I don't know. Our business is so interesting and there's so much to do that I try to like not think too much about like, what is life going to be like in 10 years? And are we going to have universal basic income and all this stuff? I generally am like a positive thinker and optimist. And so, I hope that it kind of can democratize like, it's already kind of democratized information.

Like, I think you're seeing it now with universities and, you know, kind of being like, oh my gosh, we don't necessarily sell the same value proposition we do today that we did, you know, 20 years ago when you had to get a degree to understand the three financial statements. And so, I think democratizing information for the whole world is amazing and I think hopefully we'll continue to kind of raise the floor on like what the average human's life looks like. In terms of, you know, are we going to automate all jobs? I think the argument for it is, the argument against that is basically that every time we've automated something in history, the industrial revolution, you know, we used to have all the factory workers. And a lot of those factories are now fully machines like textiles. And, you know, that actually, we're right now at, you know, 2% unemployment and the average human's life is much better than it was working in those factories.

And so, I think the hope would be, you know, we'll probably find more human connection, like things that humans value, services. Honestly, what our service advisors provide here, like we are kind of, people love this industry right now because it's very insulated from AI. Because we provide a service to people who are scared to get their car fixed and get screwed over and have no clue why their car is like, you know, you know, has like a bunch of sounds coming out of it. And to be able to talk to a human who's like, it's going to be okay, we'll get that fixed for you. Like there's a lot of value in those human-based services. And so, you know, I honestly, I guess I would, if I had to pick a side, I would say we'll figure out more jobs to have. But yeah, we'll see. I think more specific to our industry, I think this industry will look, you know, somewhat similar from, you know, the side of it's still a human relationship.

Like we talk a lot about with the auto hospitality group is like the product we're selling is not fixing the car. It is like the relationship and the peace of mind that, you know, your kids are safe driving back to college or something like that. And so I think that value proposition and that relationship will not change from AI. But ideally, we'll be able to grow our business, be able to support more and more shops at a high level and have that data side so kind of locked in that, you know, so much of the business just becomes how can we have a better relationship with the customer? And really, you know, the fixing the car is kind of table stakes. And so I think this industry will look somewhat similar still, you know, even if AI keeps exponentially getting better. I think we should just be able to provide a better service to our customers. And I think our goal is to continue to provide an even differentiated service on that.

Man, well, that's exciting. I think that's our time. But, you know, the last thing I'll say is you say it in your AI training, you know, when the iPhone came out in 2007, it was hard to imagine a world where you're using that technology to like call an Uber and get a stranger's car or like rent a stranger's house. Yeah. And then fly across the country like you never would have like put that together just because you can put a screen on a phone, you know, and you don't need a keyboard, you know. And the way I like to think about it is, you know, everybody in the back office, everybody in every position, mainly even technicians, like you come to work and you use a computer every day. Like that's just how it is. You use your phone every day. And it would be insane to think about Mac keys, our CFO or a Sadie, our controller using, doing their job, like showing up to work with like a paper and pencil.

Yeah. You know, like no computer. And I think that that's where it's going. You know, it's like five, 10, 20 years from now. It'll be like completely unreasonable to expect that like every single person is not just like totally, you know, has an army of agents that are helping them do work. And I think it'll just free us up to tackle more creative and interesting projects and that, you know, at the end of the day, the car can't fix itself. I don't think maybe robots will, you know, maybe robots will come and fix the car. But I think we've got a while for that for shop owners. And as Todd says, we're going to be the first ones in, you know, if robots start fixing the cars. So with that being said, I think we've got to wrap things up. This is an episode, I think we're on like 28 or something, Master Tech to Millionaire. We've got an amazing weekend ahead. We're talking about accounting. We're doing key to key courtside.

But next month, we've got in July on the Friday before training, we've got our AIX Auto Shop Answers training. We've got a whole eight hour program where we dive into, you know, Codex and Claude and how we're using tools that are off the shelf to automate these workflows that are more specific. So we can't wait to see you there. Please sign up. And then, yeah, we'll see you next time. This has been Master Tech to Millionaire. That's the pod. For more information, reach out to Todd Westerlin at 925-980-8012 or visit AutoshopAnswers.com. You can get more information about key to key to callbacks, courtside. We have a VIP rack attack day where you spend an entire day in the trenches with our team learning this perfected business model. We offer leadership classes. We have an AI academy. And also get more information about Auto Shop callbacks. We have auto tech training. We are literally your one-stop shop.

Once again, that number for Todd Westerlin is 925-980-8012.