Friday May 08, 2026
204 - Part 1: Using AI in Your Shop to Increase Performance
204 - Part 1: Using AI in Your Shop to Increase Performance
May 6th, 2026 - 00:59:50
Show Summary:
Artificial intelligence is reshaping how auto repair shops operate in practical ways. Jonathan Seitzer shares how AI can improve communication analyze data and save time on daily tasks. He explains a simple framework of rent it feed it and put it to work to help shop owners get started. AI is positioned as a tool that multiplies performance not replaces people. Real demos show how shops can create better customer messaging and gain insights from their data in minutes. The conversation also highlights the need to review AI outputs and use it responsibly. It closes with a look ahead at AI agents and how owners can begin experimenting today.
Host(s):
Jimmy Lea, VP of Business Development
Guest(s):
Jonathan Seitzer, Owner, Dempsey’s Service Center
Show Highlights:
[00:00:00] – Introduction to AI use in daily auto shop operations.
[00:02:35] – Background in finance and technology applied to auto repair business.
[00:06:20] – Three ways to use AI rent it feed it put to work.
[00:08:17] – AI acts as multiplier not replacement for shop owners.
[00:10:21] – Simple AI tools improve customer communication and service descriptions.
[00:15:02] – Always check AI outputs since mistakes and errors can happen.
[00:19:00] – AI helps create clear customer talk tracks from technician notes.
[00:30:16] – AI quickly analyzes parts data saving hours of manual work.
[00:37:31] – AI summarizes content into audio saving time each day.
[00:45:21] – Use AI internally while maintaining trust with customers.
In every business journey, there are defining moments or challenges that build resilience and milestones that fuel growth. We’d love to hear about yours! What lessons, breakthroughs, or pivotal experiences have shaped your path in the automotive industry?
Share your story with us at info@wearetheinstitute.com, and you might be featured in an upcoming episode.
👉 Unlock the full experience - watch the full webinar on YouTube: https://youtu.be/8_dcnz_4csE
Don’t miss exclusive insights, expert takeaways, and real talk you won’t hear anywhere else. Hit Subscribe, drop a comment, and share it with someone who needs to hear this!
Links & Resources:
- Want to learn more? Click Here
- Want a complimentary business health report? Click Here
- See The Institute's events list: Click Here
- Want access to our online classes? Click Here
________________________________________
Episode Transcript Disclaimer
This transcript was generated using artificial intelligence and may contain errors. If you notice any inaccuracies, please contact us at marketing@wearetheinstitute.com.
Episode Transcript:
Jimmy Lea: Hey, good afternoon, friends. Depending on where, when you are joining us, it could be morning, afternoon, or evening. Good to see you, my friend. I'm glad you're here. Glad we are gonna have this conversation today as we talk about the future of our industry, and how does artificial intelligence really fit into our day-to-day operations?
What does that look like? This is gonna be an interactive conversation. What do I mean by that? No, you're not gonna come on camera. No, we're not gonna unmute your microphone. Go to the comments section. We're live streaming on Facebook and YouTube and StreamYard. Go to that comments section and type in there your questions, comments, or concerns.
In fact, go into that comments section and type in where you're joining us from, the name of your shop. Love to give you a shout-out as we talk about this industry that we love that's doing so well for us. And yeah, drop in your name and where you're joining us from because it's super exciting to be here with you, friends.
It's super exciting. First and foremost is the current coach for our guest, Mr. Wayne Marshall, CEO of GEAR Group Holding, and he is joining us from Iowa. Good to have you with us, Wayne. Thank you for being here, brother. Also Steve from B&C Auto Center in San Jose, California. We've got Peggy from High Street Auto Repair, Jefferson City, Missouri.
Jeff from Miller's Automotive, Orange Park, Florida. Jeff Byrne from German Tech Motorworks, Louisville, Kentucky. And let's see, Fernando, Rohrehard Park Transmission, Northern California. Evans from Evans & Lukes in Columbus, Ohio. Evan, good to see you again, brother. How you doing? Oh, that's awesome. And Justin Pepper, Quality Auto Repair here in Oklahoma City, Oklahoma.
Nice. Glad you guys are here. Thank you for those who are vocal and know where that comments button is. This is gonna be so much fun. We're gonna have such a great conversation here. Joining me today is John from... Oh, one more shout-out. Todd from Atlanta Speedworks in Gainesville, Georgia. John joins us today from, where are you joining us from?
Jonathan Seitzer: Newark, Delaware, Dempsey Service Center.
Jimmy Lea: Dempsey Service Center. And John is a very recent purchaser, a recent joining the ranks of ownership, of shop ownership, and he joins us from computer industry, the computer world. What's your background, John? What qualifies you as a computer surgeon?
Jonathan Seitzer: Prior to my move to the automotive industry, I was the head of product at Moody's Analytics for some of their suite of compliance products for, for their banking and government services.
Prior to that I was at JPMorgan Chase for 10 years in various technology roles. I am not a developer. I live in the product and business analysis world. So for those of you shop owners out there, you should think of me a little bit like a service advisor in my last life. My job was to stand between my customers and my my software developers, help understand what the customer needed it, translated it into something the software developers could build, and then get that information back to the customer when we had a solution for 'em.
Jimmy Lea: I love it. I love it, Jon. This is awesome. Love that you come from the world of computers, and I guess technically we could call you a financial whiz.
Jonathan Seitzer: You c- you can call me all sorts of things.
Jimmy Lea: Yeah. I heard JP Morgan Chase and a lot of financial institutions you were talking about. Congrats on that, that career, that lifespan that you had there in, in that industry.
And oh, my gosh, look, we've got a few more shout-outs. Brandon from Pete's in Topeka, Kansas. Todd Compton's Automotive in Charlotte, North Carolina, and Lance Lupe joining us from... Lance, I, I can never remember where you're at. I think he's in New York. It might be New Jersey. I think it's New Jersey.
Anyways, Lance is here with us as well. Jon so excited with your background. We had a great conversation at MARS in last October. Looking forward to another MARS conference, Marketing for Automotive Repair Shops, coming this October. Our conversation last October, we talked about, you talked about, hey, you know what?
I do a lot with AI. I do a lot with the large language learning, and I really would love to share this with others in our industry and h- how they can use it, and what would make a difference for them. So let's help everybody else catch up to the conversation you and I had, Jonathan. How is it that you're using AI in your day-to-day?
What are you doing?
Jonathan Seitzer: So there's all sorts of different things. I'll actually, I have some demos we'll be showing in just a minute, but how about we head into the the presentation, and I'll walk you through the whole thing.
Jimmy Lea: Yeah. The floor is yours, brother. Hey, so everybody, as you're listening to Jon, you've got questions, go ahead and type them into the box because as Jon is doing this demo on the different AI systems, it takes a minute to process, so that's a good opportunity for us to ask questions.
So keep those questions coming in that comments box All right.
Jonathan Seitzer: Awesome. Thank you, Jimmy. So hello everybody. Welcome. As said, my name is John Seitzler. I am the owner of Dempsey Service Center. We have been in business for 40 years here in Newark, Delaware. But as I said earlier prior to that I was...
I've spent the last 15 years on Wall Street in various product and technology roles. And what qualifies me to talk to you all a little bit about AI is actually during my last stop, I was lucky enough in to release two different AI products to the market. One back in 2019 when we weren't really talking about AI all that much, and then again right before I left in 2025 we released our first agentic AI product into the market.
I've got a fair bit of experience with AI knowing what it does, more importantly, what it doesn't do. And I just wanna get that information to you guys here. The fun part about this for me is I'm not here to sell you guys anything. I'm just here to help, and at the end of the thing I'll get you all my email.
So in the event that you have questions, if you need help, if you're thinking about it, you want somebody to bounce an idea off, reach out. I am I am available. I look at AI in three different ways. There are three things you can do with AI as a shop owner. That's you can rent it, you can feed it, and you can put it to work.
Now, in this demo, what we're going to do is I'm gonna... We're gonna talk about the first two. Put it to work is a big conversation. There's a lot of different things you can do, and there's a lot of different hurdles and jumps that you have to make to do it in a way where you can trust it.
So we're gonna split that one out into another webinar in a couple weeks, so I hope to see a lot of you guys back there as we're doing that. But I think this diagram right here kinda, kinda illustrates, the amount of effort that goes into each one of this. Rent it, small and easy.
Feed it, not as big as put it to work, but bigger than rent it. You're still gonna have to do some work. You're going to have to do some learning. And then when it's time to put it to work you have to decide really is the time commitment I'm going to spend making this happen worth what I'm going to get out of it?
And some of you are going to say no, and I want you guys to know that's okay. I know everybody hears everything about AI. It's all over the news. It's everywhere. That do- and there's a fear of we might get left behind. AI is helpful AI is incredibly helpful. It can make you much more efficient.
It can help you unearth things. But I think the big thing I want to get out of this to you is to help understand what the role of the AI in your shop actually could be. AI is not going to replace I don't think very many people here in the automotive industry. The, all of the whys for that is a larger conversation than I've got here.
But w- how you guys should think of AI is not as, "Oh, this is, maybe I can replace one of my vendors at some point. Maybe I can replace, w- my service advisor," something like that. No. AI, or at least the way I use AI, is as a multiplier. AID, AI is PEDs for the shop owner that understands their operations and their data.
It will make you... It will take a good shop owner and make them great. Y- you cannot at any point go I'm not doing, I'm not doing well. Let's chuck some AI at it, and everything's going to be fixed." It just doesn't work that way. So all that said, let's talk about the easiest thing you can do today with AI, and that's rent it.
Now, what do I mean by rent it? Rent it is you all as shop owners pay for a number of services. And right now, because AI is a fantabulous buzzword, and every one of these services that you pay for, especially if they are a publicly traded company, is they're going in front of their investors, and their investors are saying, "What are you doing with AI?"
And all of these companies are trying to figure out where AI fits within the products and services that they sell. So if you use QuickBooks, there's an AI assistant. If you used any of the Google Workspace or Microsoft Office products, there is an AI in there. Your shop management system, more than likely at this point, has an AI in there in some ways, and some of the uses are big, right?
AI and Microsoft Excel as somebody who literally made his career at the start in Microsoft Excel coding stuff because, my old bosses who were around before Excel didn't wanna learn it AI can make, AI can do wild things in Microsoft Excel. Do I use a lot of them? No, 'cause I don't really need them.
You know what I use when I rent AI the most? It's this that you see on my screen. So my shop management system has a a little improve button that I can use when my techs send me- what on the right, which is a very sparsely worded, all caps missing some verbs sometimes write-up about whatever it is that they're working on.
Now, in the past, what we might have done was just copied and pasted that, and that's what the customer got to see, right? J- on their invoice. Now, what we can do is I can hit an Improve button, and it's going to run through that. It's going to try to determine the context, it's going to spit out just a nice paragraph.
Is that the world's biggest time saver? No. Does it lead to a better experience for my customers? Yes. So I wind up using I wind up using that in the rented category, honest to God, more than anything else we're going to talk about today because I have, a fair number of techs, and none of my technicians like using anything other than short paragraphs in all caps locks.
That's rented. Why is rented important? So when we talk about what my past life, right? My past life was f- my whole job was figuring out, "How do I put this into my tools?" So when you're renting your AI, one, you're not, it's not costing you really anything else because you're already, you're getting it as part of the service.
Two, the AI that you are using has been thoroughly vetted, in many cases, by a team of people who are just like I used to be, whose whole job is to figure this stuff out and test it in every way. So the risk of you using it is much, much lower. That's an important thing to understand about AI is AI is not deterministic.
It's probabilistic. And what I mean by deterministic versus probabilistic is AI uses probability to deter- to figure out what the next word it's going to say as it's writing a sentence to you. This, it's all math. Deterministic is literally, I flip a switch and the light goes off. I flip the switch, light goes on.
It's determined. It will never be, it will always be one of those two things. The light goes off or the light goes on, and if it doesn't do one of those things, that means your light bulb's broken. So that's rented. Simplest thing you can do. The return on it isn't as big, but it can make your life easier if you're using things like the QuickBooks Assistant, or it do- it might do something as simple as make the invoices that your customers see a little bit better.
Things you need to do when you're when you're renting it is, first off- Look through your tools. Who's offering this to you and where are they offering it? Determine the features they're offering, right? A feature in my shop management thing that cleans up things, that's super useful. Maybe they add AI somewhere else and I have to think do I really want AI there?
Do I want something that could make mistakes that I'm not supervising in use? So explore your tools, identify your features, and then start playing with it. Again, these are part of the tools that you are already paying for. You can afford to experiment. You're not-- you don't have to go out and buy a subscription or learn how to use a Claude code or a OpenAI codex or a Google Antigravity.
You have them here and ready for you. They're in the tools as you understand them. It just changes a little bit your process, and you can decide, is this worth changing up my process for?
Things to watch out. The quality of the AI varies by the vendor. Not all AI is created equal. The more powerful the model, typically the more expensive it is to serve. A lot of times what your your vendor that's offering you an AI product isn't using the most powerful model out there.
They might be using something open source. So all things AI, and if you get nothing else from this, get this audit it. Check it for mistakes, especially when you're starting out. I was even prepping for this. I was running a couple demos on my side and it made, a couple of boneheaded decisions and spit out some information that wasn't right.
So you've always got to check it. Then additionally, as you start to implement it, you should have experienced people working at your shop wherever you're using this stuff, looking at it before it goes out. If AI is cleaning up your emails or your or, service descriptions on your invoices, that doesn't absolve your service advisor of looking at the invoice before they fire it off to the customer.
And then also you have to assume anybody who's used ChatGPT I like to say AI has an accent. ChatGPT especially has an accent. If you have somebody that's used AI a lot, you can tell when AI writes something and the way to get around that is to teach AI how to write like you want it to write.
If you just let it go, people are going to be able to tell. I can tell definitely.
All right. Here's where we're gonna have some fun. I have a few different demos we're gonna get to run through for here. And let's first talk about feed it, right? Feed it is what we think about AI how we've been using it a lot these last few years, right? You would log into a browser, there would be your AI chatbot, you could type your question, your comment, your whatever, and it would output some type of result.
As they got more advanced, you started to be able to attach things to it to offer it additional context. And so it went from, "Tell me about the history of the moon landing," to, "I have a spreadsheet, I'm gonna attach the spreadsheet. Tell me about my spreadsheet." So this is the second piece where you can start to get real value out of AI, is you have systems that generate data.
Why not use your-- these chatbots, these AIs, to help you understand your data? There's, me as a person who came from a technology background and moved into automotive with no real experience in automotive at all, outside of being a, an enthusiast my shop management system produces so much more data than, even I could process.
Without this stuff, I'd be hours a week crafting pivot tables and running analysis just to try and understand where we are, where things I can literally do in minutes, if not seconds. And I'll show that to you as we move forward, right? So you can use things like your customer feedback, your service histories, your repair orders, your parts inventories to get real good analysis out of these tools.
So let's go on a ride, folks. I have no, no idea what's about to happen. This should work. So let's start with our first demo, and I think this is the most fun. Like I said, I was prior to this, I worked in technology. I do not have a background in automotives. I am a shop owner that does not know anything about how to fix cars, and I had never written service before I took over this shop just a few months ago, right?
So a couple of weeks ago, when my service advisor wanted to go on vacation with his family for a week and I only have one service advisor guess what? Time to learn how to write some service. And as somebody who doesn't, who, One of my technicians will come to me and say, "This is what I need and this is what's wrong."
I understand it in theory, but less in concept. So I developed a script that I used to help take what the technicians were recommending to me and give me a talk track that I could use to customers. So when I was talking to my customers, they didn't necessarily know how way in over my head I was.
So this i- these are my service writer instructions, not for my actual service writer, but for my AI service writer. This is when you're feeding it or yeah, when you're feeding it, sometimes the thing you want to do is just ask your questions, but sometimes what you want to do is you want to give your AI a role and give it some guardrails to lower your risk that the AI is going to go farther than you want, or worst case, make up something that isn't actually true.
Thing to understand about these things is they want... Want is a bad word. I don't like anthropomor- morphizing machines, but the AI is designed to try and be helpful. It wants to get you an answer to your question, and sometimes when it can't find one, it just makes one up. Or if it can't find a piece of data, it makes one up.
I told you earlier it did something boneheaded. When I was testing this it, I gave it a vehicle, and it decided that the odometer reading on the vehicle was 253,000 miles for a Ford Mustang GT, which would make it the greatest Mustang GT in the world. So in this case, I am giving my AI a set of instructions that it's going to use to help me come up with a talk track for my thing, but here are the rules, right?
So I give it the set of instructions. I give it what's going to happen, "Hey, these are the steps you have to follow." I tell it what a service writer does. I tell it what to consider here for their talk track. And then this is important. When I told you AI has a accent, this is how I scrub the accent and I tell AI to talk like I want it to.
So I've come up with a brand voice and rules for my brand here in, at Dempsey's, and I give that to the AI and I tell it, "You gotta... Here's what your tone is supposed to be. You are not allowed to do this. You are not allowed to use jargon. You are going to present the findings honestly. You're going to avoid certain words.
You're going to recommend certain things if they need to." We do have financing, right? "You're going to tell them about our warranty." Then I tell it how I want it to structure the response, and then here's the u- last thing, I tell it what it is absolutely not allowed to do, what not to do. You can't include pricing.
You don't get to invent a finding that isn't in the RO. You don't get to diagnose anything that the technician didn't say, and you don't get to tell the customer what to do. So now how does this work? This is always available to me, so I'm going to copy it. I'm going to come back to my repair order, and there's a bunch of different ways to do this, but this is the way I like I like to use this, is I u- I pay for Gemini's Gemini's AI.
So Gemini is Google's large language model. And at the tier I use, I get a little thing in my browser that I can do this, and it says, "Hey, how can I help?" And when you do this in the browser, what it's doing is it's sharing my browser with the AI. So now the AI can see basically everything I can see, right?
And now I'm just going to paste in... Nope, definitely not gonna do that one. I am going to paste in my instructions. And then this is also another important one. You typically get options with your AI what kind of model do you want, right? You almost never want to use this. This is basically the free AI. Free AI is bad, and I'll go into this a little later, but free AI is typically the lowest capability. It thinks the least amount, it gives you the fastest response, and it gives you the least accurate response.
There's a time and a place for this but in most cases I use thinking. So I've given it my thing, and we're gonna let it think. And while it does that, Jimmy, do we have any questions?
Jimmy Lea: N- none questions that have come through yet. But I'm absolutely fascinated. When you designed the voice of your service advisor, did you use your LLM to design that language for your service advisor voice?
Jonathan Seitzer: Yes. Yes. Okay. So I created... So I've Russian nesting dolled my my brand voice.
I used an LLM to come up with my initial brand voice document. Yeah. So it, it contains all the rules of the brand, and I ha- I do that by having the AI interview me about what I want and then telling it I want a brand voice document, and it gives me something there, because I'm not, I don't have a marketing background.
I don't know how to create a brand voice. Okay. And then what I do is I share that document again with the AI, and I'm telling it what I'm trying to accomplish. I, my service writer's going on vacation. I don't I need to understand, I don't understand cars super well. I need to be able to give the customer- the information about their repair and their estimate.
Here's, these... This is what I want. Here's my brand voice document. Write me a list of instructions and it- make sure you incorporate the brand voice.
Jimmy Lea: Nice. I love that. So even in your instructions, you were talking about your service advisors leaving. Eh, let's say the service advisor is still there, we've got a brand new service advisor or we want to have a voice that is able to speak to a client or a customer in a, in easy terms and not speak down to them, but speak on a level that they're gonna be able to understand.
That could be part of the instruction for the service advisor voice as well, correct?
Jonathan Seitzer: Yes. Yes, absolutely. And it is part of it is part of mine, right? When we're talking about, warm, straightforward, neighborly, no pressure, you're an expert neighbor who explains things clearly 'cause you genuinely care.
Jimmy Lea: Yeah.
Jonathan Seitzer: Yeah. And that ensures what I get back should be to that level. But also, what did I say earlier? We're not just blindly trusting this.
Jimmy Lea: Yeah.
Jonathan Seitzer: So we're going to look. So here's what we have, right?
Jimmy Lea: And one more question here before you go into this.
Jonathan Seitzer: Sure.
Jimmy Lea: Do you find that as you continue to use this LLM and as you continue to feed it with information, does it improve in its voice and tone and become more refined as to past inputs versus no?
No, sweet.
Jonathan Seitzer: No. This is very important when you're dealing with LLMs. These things don't learn. So the moment I push this button right here and it spouts- Yeah ... out a new one, it's forgotten it's ever told me anything.
Jimmy Lea: Oh.
Jonathan Seitzer: Yes. So there are certain applications now. So ChatGPT, Claude they do retain some memory.
Okay. You can say, "Hey, remember this about me." So Claude's always going to remember that my name's John, that I own a shop, that I'm in New York, that I have a dog named Chrissy, that kind of stuff. But it's not... The AI does not improve itself, right? That's what all, y- that's what all these LLM companies are spending billions and billions of dollars trying to get to.
It's not self-improving. It doesn't remember. If this thing spat out something that was totally awful, what I would probably do is I would figure out w- where and why, and then I would have to come back here and tinker with these instructions.
Jimmy Lea: Change your instructions, yeah. A question's coming from John.
Is there any way for you to share your instructions that you're using with Claude? And I think the answer is yes. However, the deeper answer or should be maybe...
Jonathan Seitzer: Create your brand voice.
Jimmy Lea: Create your brand voice. Have that interview with your ChatGPT or with your Claude and have chat ask you all those questions so that it comes through for you.
Jonathan Seitzer: Yes. And you can do... y- I'm sure a lot of people are saying how do I do that?" You... These things are, it's... Once you get the hang of talking to this- Yeah ... it's wild how fast it unlocks. So how do you do it?
Jimmy Lea: Yeah.
Jonathan Seitzer: You ask, "Interview me. I need a brand voice document for my auto repair shop."
You give it as much context as you can. "Interview me to get the thing." And it will quest- one question at a time, ask you questions.
Jimmy Lea: Yeah. No, that's exactly it. Yeah the answer is yes. However, have your own
Jonathan Seitzer: interview. If y- if I, if you use, if I send you my stuff, you're using my brand voice, which I think I have a great brand voice, but that doesn't mean it's yours.
Jimmy Lea: Yeah, no. And I know, John, I know, John, you're gonna have a great conversation with your ChatGPT and have it interview you what kind of voice you want. And brand voice is the keyword that you wanna use there, John. Yes. So maybe what we c- develop here, Jonathan is some guidelines for them creating their brand voice.
But anyways let's go back to what you're showing here, because- Sure ... this is where it gets exciting in putting this information into the point of sale system.
Jonathan Seitzer: Yep. So here it's welcoming me to Dempsey's, so there's something in there that made... I wrote that told it I'm brand new.
But yep, here we go. Here's our first mistake, right? This is not a 23,000 mile Mustang. You see there's nothing on the here that tells you it's a 23,000 Mustang. And again, we used the big, we used the more complex-
Jimmy Lea: Don't create information ... marketing. You told it not- Yes ... to make stuff up, but it still made it up.
Jonathan Seitzer: Yes. But now it's looked at the, it's looked at the thing, right? It tells you what's completed and approved, so here are all the recommendations it's got, and now it's going to go through each of the recommendations, and it's gonna, So f- the first thing it's gonna do, it's gonna tell me what the fix is, right?
So we're gonna walk through this, then it's gonna give me the customer talk track. These are actual phrases I can use on the phone with the customer. "Our technicians noticed the drive belt showing signs of age," things like that. And then it's going to talk about what our recommendation is, why it's important, and what happens if we wait, right?
Yeah. And for a solid week, and for those of you who don't know if you are s- if you're a single owner shop with a single service writer- Yeah ... the fastest way to make your phone ring is to send your service writer on vacation for a week. We did almost a record amount of cars. And it was John sitting at a desk waiting for Gemini and Claude to spit out the, trying to talk.
And I actually, one of my parts distributors said I closed a lot of sales that week, so it was good. Congratulations. We still managed. All right, so that's that's our first demo here. Let's do- Let's do some analysis, right? Yeah. All so now we're going to go directly into Claude itself.
Give me just a second to pull up my demo file, and we're gonna do a parts... We're gonna do some parts information. You guys aren't... You'll see this change in the Claude screen. I'm dropping two I'm dropping two CSVs. So I'm dropping some parts data from my from my shop management system.
And now- So
Jimmy Lea: this isn't the whole catalog from your parts supplier, this is what you've used- Yes ...
Jonathan Seitzer: in the past. This is parts data for the shop that, have come in, have come out. So basically now I need to tell the AI what I want. Here's why we're doing this. I need to get some information to my accountant.
Can you look at the provided data and let me know what my parts inventory cost was at the end of February, March, and April? Break the cost into parts, tires, fluids, and batteries.
Jimmy Lea: So that was all voice dictated. Yes. You used just the microphone s-... Okay.
Jonathan Seitzer: Yes. So I have a little program on the thing.
I push a button, it records my voice. A lot of computers have this built in. I use a paid one just 'cause it's a little better, and guess what guess what's undergirding it? AI. Yeah.
Jimmy Lea: Yeah.
Jonathan Seitzer: So here we go. So this is what I want. Tell me what my pa- I gotta get my accountant what my on-hand inventory co- or price was.
So we're gonna let it think for a little bit. And yeah, do anybody have any type of questions or anything we can go through as this thinks? 'Cause now we're at the part where this is probably gonna take a little bit of time, 'cause we're asking it to do a lot of different things and generate a lot of different information.
Jimmy Lea: Yeah. So right now the process it's going through is opening each of these files and- ... looking into the dates, the parts, the costs.
Jonathan Seitzer: Yes.
Jimmy Lea: And it's trying to answer the question you've asked. So it's crunching a lot of data, and I'm assuming you've probably got hundreds of lines of data that it's crunching through.
Jonathan Seitzer: Yes. Yeah. This is, every part we've ordered or has gone in or out the door since February 1st.
Jimmy Lea: Wow.
Jonathan Seitzer: So- This
Jimmy Lea: could take a minute.
Jonathan Seitzer: It's... You'd be surprised. I, my guess is probably... So right now it's tr- it's understanding the structure. Yeah, my guess is it's gonna take a minute or two.
Yeah ... and anoth- that's another thing to get, to get used to as you're using these more advanced models. So there are basic models, there are thinking models, and then there's ways to make thinking models think longer. As you're using more and more complexity within your models, as you're turning on more features, if you're on these paid plans, you have usage limits.
And somewhere depending on how much you pay it's, you're gonna hit your usage.
Jimmy Lea: Got a question here from Sierra. She's asking, "Is this dir- linked directly into your shop management system, or do you have it upload all of the documents first, and then it does its searches?"
Jonathan Seitzer: So I uploaded all the document for this demo.
So I have an AI that is linked directly into my shop management system, or parallel linked via a public API. But we'll get more into that in two weeks. This is where the, I'm willing to spend nine hours on a Saturday building a connector into my shop system 'cause I'm a dork.
Jimmy Lea: N- nerd. The preferred term is nerd. You're a nerd- Oh ... and we love nerds. Nerds are awesome. They're amazing. It keeps the world spinning, yes.
Jonathan Seitzer: But yes.
Jimmy Lea: So you love to nerd out on developing... this is your hobby. This is your go-to. This is your fun time. This is relaxing for you.
Jonathan Seitzer: Yes.
Yeah, this is I like building software and I... I like building software and I like building things that, that help me understand kinda what I've jumped into. 'Cause I can't stress this enough, I have not been working on cars for the last 15 years. Yeah. Or been a small business owner, or done anything that I do day to day anymore.
Jimmy Lea: Wow. Wow. I'm glad you have this as a hobby. This is so much fun. This is where AI is going to assist the humans, and AI is gonna assist us to become better. I think it's gonna elevate us as a human race.
Jonathan Seitzer: Yeah. So while that thinks, let's do another one. So we'll jump into Gemini. And so everybody knows, my AIs of choice are Claude and Gemini.
I've used ChatGPT. I have no problems with with it. Actually, ChatGPT's new model, I'm told, is spectacular, which was just released in these last couple of weeks. It's just a matter of, the, there, you can only pay for so many things, and for what I use Claude for- Yeah ... it just makes more sense for me.
But you can do this across just about any one. There's, and, there's a lot of, i- as you get more into this, there's a lot of, "Oh this model's good, and this model's good." The ranking changes week to week, right? Anthropic's on top now. Three months ago everybody was saying Google had ended the debate.
Yeah. Don't, you don't have to do exactly what I'm doing. But here, let's we can... and the beauty of this is we can kick off parallel demos. So allow me to pull my folder here. Let's do something a little easier here.
Jimmy Lea: I thought that said disco.
Jonathan Seitzer: So this, here's what I wanna do.
I've just joined a local chamber of commerce, and they're sending me a welcome packet. Can you read it and give me a list of action items I should do in the next five, 10, and 30 days?
Jimmy Lea: That's hilarious.
Jonathan Seitzer: Jimmy, do you- That's awesome ... read PDFs anymore? What are you, Amish?
Jimmy Lea: Evidently.
Jonathan Seitzer: I'm new to all this. I can't be, I can't be bothered to open and read PDFs.
Jimmy Lea: I love it. Okay, go back to the other one. Did Claude finish?
Jonathan Seitzer: Claude did finished. Okay. Don't contain ending inventory. So yeah, this is where it gets funny. So this is where it'll start to quibble with me, right?
On-hand, list of every SKUs. So now it's given me... Here we go. So this is my... We'll ignore February 'cause that was half a month. But yeah, so now I can, to my accountant, I can say this was my parts on-hand cost and please get that to my ba- get that on my balance sheet," right? And, three minutes.
Jimmy Lea: Wow.
Jonathan Seitzer: Yeah.
Jimmy Lea: Yeah, that's great. The other things you would have had to have done is gone into your Excel file and been an Excel wizard.
Which I'm a pretty good Excel wizard. I know a couple of people that are better than me. But yeah, it wouldn't have taken me three minutes. It would've taken me a heck of a lot longer than three minutes.
Or- I click on data.
Jonathan Seitzer: Yeah. Or I would have had to have I would have had to have pulled Excel files for February, for March, for data, and now I just all in one throw it into the thing. I don't have to spend a bunch of time playing around in pivot tables.
Jimmy Lea: Yeah.
Jonathan Seitzer: And now I've got something. Now, obviously, double-check it.
Yeah. I, I would, I've so I, I know this one works 'cause I've done it before. I've double-checked the numbers, so I know when I ask Claude this thing, 99% chance I'm going to get it right.
Jimmy Lea: Nice. Nice.
Jonathan Seitzer: All right. So there's that one. There's this one. And here we go. First five days, right? Oh my God. Things to do.
Here's what to get done in 10 days. Here's 30. Now, in the next webinar, I can show you how to then fire this into your to-do list or project management tool, like a, an Asana or a Trello or something like that. And now it's not just a thing that lives on a web browser on my computer, it's on my phone in my app that I can go, I watched the video, click.
Jimmy Lea: Love it.
Jonathan Seitzer: And...
Jimmy Lea: And you didn't have to read the PDF, Mr. Thomas.
Jonathan Seitzer: Oh, gosh, no. It's, yes. Think of the time, think of the time savings. Now-
Jimmy Lea: Oh, yeah ...
Jonathan Seitzer: w- we joke now because it's, it's one PDF and who can't read one PDF? I'm, maybe I'm unique in this. I get between 15 and 30 newsletters a day- Yeah
across technology, politics, economics, yeah, all the stuff I used to have to pay attention to that I still like paying attention to. I don't have time to read 'em all every day, so I have a tool that collates all of them and then turns them into a 20-minute podcast for me to listen to.
Jimmy Lea: Oh my gosh. So you're, so you've customized 15 newsletters into one single podcast, and you listen to it while you drive home from work.
Jonathan Seitzer: Yeah. So I do it twice a day. So I have a morning one and I have an evening one. So I listen to the evening one after, while I'm making dinner. I listen to the morning one as I'm finishing up my morning paperwork here and walking around and checking on things.
Jimmy Lea: I love it.
You're such a nerd.
Jonathan Seitzer: I know. But it's
Jimmy Lea: all- I'm glad I know you, John ... it's- I'm glad you're part of this industry. I'm glad you're helping share this with the rest of the world. Thank you. And the other John wanna know how it works.
Jonathan Seitzer: It's all about buying back minutes, right? Time- Yeah ... and time is truly the only finite thing that we have.
Jimmy Lea: Yeah.
Jonathan Seitzer: And if it, if 20 newsletter I want to read my 20 newsletters.
Jimmy Lea: Yeah.
Jonathan Seitzer: I don't have the, I don't have two hours to do it every day. So-
Jimmy Lea: Yeah ...
Jonathan Seitzer: but I got 20 minutes to listen.
Jimmy Lea: Yeah. Oh, for sure. For sure, especially while you're multitasking, you're driving, you're building making dinner. Yeah. All right.
So what else are you gonna show us?
Jonathan Seitzer: All right. So this last one is just a variant of the one that we did, but it's a lot more shop oriented. So here I am attaching one, two, three, four, five, six files for the month of March. How do I wanna say this? I need to understand better how my month went.
Please show me a breakdown of my top five services by make, model, and category, as well as the revenue generated for each.
Jimmy Lea: Wait a second, Jonathan. You could also do this for an entire year to discover your most profitable vehicles as well too, right?
Jonathan Seitzer: So that's actually a really good- call out. So technically, yes, but when you're dealing with these AIs, another fun thing to keep in to keep in mind is that the AI can only take so much contact, context.
So what is context? Context is everything that's going into this conversation. So these conversations, like the AI's not really having, we're not having a chat, right? If I go away for four days and then come back to this chat in this window and answer a question, what's going to happen is the entire conversation is going to get sent back into the cloud.
The LLM's going to reread the whole thing. Again, like I said, this has no memory. So every time you're having a long con- a conversation with AI, it's basically sending the whole conversation back and then returning the whole conversation back with a response. So the longer the conversation you have, the more you're filling up this context window of however many million tokens or hundreds of thousands of tokens, and once you hit that window, the AI will start to...
it starts to get weird after a while. It's, it becomes more prone to making making mistakes because it just can't... It doesn't have the capacity to remember everything. So if you're ever on Claude and you're having a long conversation with Claude, and that it's compacting, that's where it's suddenly, it's taking that context, throwing out what it thinks it can, and trying to keep the relevant points to keep the conversation going.
So when I say, "Oh, yeah, could I throw this for a whole year?" Yes, but I don't know how well... that just might be too much for the AI. So typically what I do is every month I run one of these.
And at the end of every week I run one of these, and I pull top-line metrics out for the week, and I keep that in a tracker like its own Excel document, and now I can, at the end of the year, I can point the AI at that Excel document and tell it to give me yearly insights without overloading the context and risking you're gonna get some bad information back.
Jimmy Lea: Nice. Nice. Oh, that's awesome. All right.
Jonathan Seitzer: So what- But yeah, look how fast this, look how fast this one came back, right? Here we are, top five service categories, repair versus inspection versus o- over the counter versus our snowplow business. We had 57 Fords, 33 unique, then Chevy, Dodge, Jeep and then we've got...
that's, yeah that was an engine. So here's the actual, this is the fun... here's where you get insights, right? Jeep, revenue we had 10 unique ROs, but look at the revenue, right? So Jeeps are my unicorns. I don't get as many Jeeps as compared to Fords and Chevys, but the revenue, way up there, right?
My- Yeah. So I know per ticket when I see a Jeep come in, oh, there's a chance this is going to be a much higher ticket for my shop, and that way I know. I know with three months of data that Jeeps and Silverados are what keep the lights on here. But as a snowplow seller also, I also know I have a lot of Silverados in this area, so when it's time for me to make my order for what type of snowplow mount kits am I making this year, I'm skewing to what I know I have a lot of, and I know I have a lot of because the da- I've got the data in this thing and it's summarizing it very cleanly for me.
Jimmy Lea: Oh, yeah.
Jonathan Seitzer: So yeah. So that is that is my presentation. Nice. Our little... Or those are our live things,
Jimmy Lea: yeah.
Jonathan Seitzer: What time are... 1:48. All right. I can get through the rest of this pretty quick. All right. So again to summarize our feed it section make sure you're using the right tools that your shop needs.
Start with the big ones, but I can't stress this enough, you're gonna have to pay out of pocket. Start with the $20 models, see where it gets you. Up your spend as you find value.
Jimmy Lea: Yeah.
Jonathan Seitzer: Can't use that one. Don't throw your payroll data into this. The, the, these are going into the cloud. And also make sure that your that what you're getting back is accurate.
You gotta check its work the first couple times, and that's not, I don't think, an unreasonable thing, right? You wouldn't hire a human and just let him go. Same thing here. In two weeks we're going to talk about the next evolution of this. The word you're probably hearing a lot now is agents.
If you think of AI as a giant brain, think of agents as arms and legs. So we'll be talking about that in the next thing. But this is one important thing I wanna give to you guys before and, forgive me if I'm about to be a little vulnerable here.
Jimmy Lea: Okay.
Jonathan Seitzer: On the screen is the email you get from your company when you're laid off.
So this was mine.
Jimmy Lea: Oh.
Jonathan Seitzer: Though, if you wonder why I'm not on Wall Street anymore, this is why. Now, AI did not take my job. I did not get laid off because of AI, and honestly, blessing in disguise 'cause now I have this cool new job. But there are a lot of people, and this is incredibly important I think, this is somebody who was an auto repair customer much longer than he's been an auto repair owner.
If you live in an urban center with a large white collar population, understand that a lot of them are worried that this is coming. So when I'm using AI in my shop, and I am using AI a lot, you guys have all seen this, it's never customer-facing. I'm using it to make me better, but you saw my brand voice, right?
I don't want my customer to ever think that they're dealing with an AI. 'Cause I think they n- you know, I think where I live, where there's a large white collar population, there's a lot of people that are worried about this. So by all means, use AI to increase, to make yourself a Superman inside your shop.
Jimmy Lea: Yeah.
Jonathan Seitzer: But be aware that people that, your customers probably have opinions about this, and use it use it for you. Don't push it to them if you're not sure they're, how they're going to take it. And that's my last thing.
Jimmy Lea: Oh, man. I love it. I love it. A great shout-out here from Wayne. "Great job. AI's not gonna replace people, but love how it multiplies the effectiveness of the staff and the effectiveness of you as an owner and the effectiveness of your service advisor."
It's really gonna help those relationships and those communications to happen at a much, a m- much better level.
Jonathan Seitzer: Yes. Definitely.
Jimmy Lea: Dude, John, this is awesome. Question here coming from... k- question, comment. It's more of a shout-out from Joshua. "There are solutions for connecting directly to your SMS."
This is, in his estimation the best and easiest approach.
Jonathan Seitzer: Y- I think it depends on your SMS. Like in my case to connect to mine, I had to talk to the I had to talk to them. I had to tell them what I was doing. I had to, make sure that they understood I w- I was building something for me, not something I'm trying to take to market.
Yeah ... but y- you're right. Some of these SMSs, I think as we get more into it, they're going to start, it'll move into that rented space, where maybe from inside the SMS you can start to get a lot of this information. But there are ways to connect. Your mileage may vary depending on who your user is.
Mine took a little bit of elbow grease.
Jimmy Lea: Yeah. Yeah. No, that's phenomenal. That's phenomenal. Great information, John. Thank you. This opening up, opens up a whole new world of possibilities for what we should look at as we go down this road with AI. My daughter, she came to me and she says, "Man, I'm really worried about AI.
I think it's gonna replace my job." And I said it's not gonna replace your job, but the person who does understand how to use AI is the person that's gonna replace your job."
Jonathan Seitzer: Yeah. And one of the things I liked to stress back in my old life is right now there are no old graybeards of the AI world, the way there are in just about every industry.
The- Yeah ... those of you, those of us who are using it and figuring it out and charting the course, we're the ones who decide how it winds up going and what it winds up being good at.
Jimmy Lea: Oh, for sure. For sure. John, this is gonna be awesome, man. Any advice going into our next webinar, any advice for what people can do to prepare?
Jonathan Seitzer: I would research things like ChatGPT's Codex or Claude Code and CoWork. Those are the two most accessible kind of agent harnesses out there. What I would advise against is Claude, or not Claude OpenClaude, something like that, Hermes Agent, some of these open source plug your agent in.
Please don't go out and buy like a MacBook Neo or a Mac mini, if God help you, you could even get a Mac mini anymore. These agents are the reason why you can't buy Mac minis right now or a Studio Ultra. The next piece is, so I would say this, if you know what the terminal is on your computer the next webs- the next website or the next web- webinar's gonna be very useful.
If if you've never coded anything or you're not super, I'll do what I can to show you guys where it is and how it goes. But the next stuff is all nice to have fun extras, but don't feel like the next one you gotta, it's all stuff you gotta do. You're not missing out if you don't do this.
And it comes with work. It like, how important is your Saturday? Me? Well- Not super much, but...
Jimmy Lea: but this is your hobby. This is what you're doing. Exactly. This is your relaxing enjoyment time.
Jonathan Seitzer: Yes.
Jimmy Lea: So what I'm hearing you say is let's get into a ChatGPT and have a conversation. Yes. And for the other Johns of the world, have a conversation, have chat interview you about what you want your brand voice to be-
Jonathan Seitzer: Yes
Jimmy Lea: so that you can create a prompt for other LLMs to use as your brand voice as you're talking to them
Jonathan Seitzer: Yep.
Jimmy Lea: So beyond them creating a brand voice and getting ready for two weeks down the road, open ChatGPT, have a conversation. Open Claude, if that's the one you wanna use. Open Gemini, if that's the one you wanna use, and have a conversation preparing for what we're gonna do in two weeks.
Is that... Are we gonna go through setting up an AI agent here in two weeks?
Jonathan Seitzer: I, I can... So cl- that's where the Claude code. So Claude Claude Cowork is like a out of the box agent, right? And you just kinda have to point it at a spot on your computer to go. Like a full setup. Now I've got full dork, integrated agent on a server somewhere else that I talk to in Slack.
Y- we're not gonna do that. But- Oh, s- Yeah ...
Jimmy Lea: you are such a nerd. This is awesome. Oh,
Jonathan Seitzer: yeah. Yeah. Oh, I see a comment from Jeff. You heard at a conference Claude is better. I said this earlier. It's, it bears repeating. Claude is winning now. Gemini was winning a month and a half ago. ChatGPT will have its moment again in the sun.
Use the one you get the most value from and the one that you're comfortable paying whatever the price is to use it. My preference is Claude. I like the answers it outputs the most. Some people really like the way ChatGPT sounds or comes back to them. They like the quality of the answers.
It really is a pers- preference. There, Claude, ChatGPT, Gemini whatever you pick, you're not you're not losing out too much I don't think. I think you're gonna get roughly, especially if what you're getting it, looking for is data back, you're going to get roughly the same quality of answers if you're using the paid versions.
As long as you're using the paid version you stick.
Jimmy Lea: Yeah ...
Jonathan Seitzer: peggy.
Jimmy Lea: Yeah. Last question right here with Peggy. Where can-
Jonathan Seitzer: YouTube University ...
Jimmy Lea: AI answers for simple tasks like emails and calendars, et cetera. She didn't even know where to start, so this is for the total novice.
Where can she start?
Jonathan Seitzer: So there's a great channel on YouTube. Search a person named Elliot Prince. He's he's a British guy. He does a lot of stuff with Claude Cowork. He's got a bunch of videos of, like, where to start, here's what it does, here's what you can do. And also he makes his prompts and his lessons available publicly.
Beware when you're on YouTube, a lot of these YouTube channels are really designed to get you to sign up for their paywalled course. That hasn't been the case I've seen with that one, and that one was really helpful as I was transitioning from Gemini more into Claude.
Jimmy Lea: Yeah. And he's even taking the stance now that he's been with Claude for so long, he's now looking at ChatGPT and saying, "Oh, my gosh it's improved so much."
I've gotta dig back into this to dig more into it. Yeah. So he's, even he's going through those gyrations of-
Jonathan Seitzer: Yeah ...
Jimmy Lea: they're constantly improving.
Jonathan Seitzer: Yep.
Jimmy Lea: Oh, that's awesome. Yep. That's awesome. For those of you who are listening thank you for joining us. John, thank you for joining. Thank you for sharing your nerd wisdom.
We, we need people like you in the world, and in fact, we need all sorts of people. It's great that we're not all rubber stamp identical of each other. We are all different in this world, which makes us great. So thank you, John. I really appreciate it.
Jonathan Seitzer: Thank you, all. It's been a pleasure.
Jimmy Lea: Yeah. And for the rest of you who are listening, we we at the Institute, we are a coaching training company.
We're a coaching training business. We're here to help you take those next steps, like we did with John when he bought a shop and didn't know what to do. We were able to step in. He hired us as his coach and his mentor. We even started coaching him prior to him buying the shop. So that's how valuable the, and important a coach can be to you and to your business to help take it to those next levels.
So if you found this information valuable, if you found it interesting, understand this is the tip of the iceberg. There's so much more that we can do and that we can do together. Reach out. We'd love to have a conversation with you and to talk about your shop and your situation, 'cause yours is gonna be different than John's.
Let's have a conversation to talk about what you can do to build the best business you can possibly build. My name is Jimmie Lee. I'm with the Institute for Automotive Business Excellence. So excited to be here with you today, and look forward to having our next conversation. Talk to you soon.
No comments yet. Be the first to say something!