320: How I'm Using AI Tools In My Business Right Now

Welcome to the Automotive Diagnostic Podcast.

We're going to explore ways to sharpen our diagnostic skills, find learning resources, and hear from

experts in the automotive field.

This show is brought to you by Automotive Seminars.

If you're looking to stay up on the cutting edge diagnostic strategies and techniques,

this is the place.

You've got live four-hour seminars that are put online that you can join in, ask questions,

interact with the instructors.

With some of the best of the best, they've got Keith Perkins and John Thornton putting

on classes like Network Diagnostics, Mechanical Engine Testing, Advanced Electrical Testing,

and GDI Diagnostics, and a ton more.

Once you purchase one of these training seminars, you can access this at any time that you

want through your account at AutomotiveSeminars.com.

You can re-watch the training a year, two years later, which makes this training even

more valuable.

I highly recommend you check this out.

The link is in the show notes.

Hey, what's up, Automotive World?

Welcome to another episode of the Automotive Diagnostic Podcast.

My name is Sean Tipping, and I'll be your host once again for this week's episode.

You got just me, but thank you for joining me on the show, and as always, thank

you for all the feedback and kind words about the show.

I was interacting with people.

I've said this before.

If I don't get back to you right away, I'm not ignoring you, I'm just busy.

It's one of those things where you look at your phone, and maybe you open up a message,

and then you're about to respond, and then you get a phone call, or you get a distraction,

which happens a lot, especially during the week, and then I forget to get back

to it.

Then I see it later.

I never messaged that person back or whatever.

Anyways, let me just say, if I don't get back to you right away, not ignoring, just probably

got distracted or I'm busy, and I try to get back to everybody, because I do appreciate

all the listeners of the show.

It's why I do it still, is try to share some information out there, try to help

everybody out, and move this thing along together.

It's not easy.

We're up against a lot of challenges.

We've got to take whatever help we can get.

It's kind of the premise of today's show that I just wanted to share some thoughts about was

what I'm doing right now to utilize artificial intelligence, AI, specifically large language

models to assist me and help me out in the day to day, what I'm using them for.

I'm planning to actually have a couple other people on here to talk about AI more in

general, maybe some of the concerns or our thoughts in regards to the industry.

So there's a whole conversation to be had there that I definitely think is warranted.

But for today, I'm just going to talk about right now in September of 2025, what am I

using AI for that's helping me out the most, that is moving the needle for me in specific

areas that I found really beneficial, right?

The cost of what it actually costs me for these tools and these services versus what I get

out of it, where am I seeing the best bang for your buck, right?

And I mean, most of these platforms, I don't care which one you use, you're paying

probably about 20 bucks a month for the premium version of it, you can pay more.

And I think there's free versions of a lot of these.

But if you're paying the $20 a month and what you can get out of these tools, if you utilize

them properly, it's insane, the amount, you know, again, dollar to output, it's crazy

small because you can really do a lot.

And part of it's only limited by your imagination and willingness to learn these tools.

But man, if you can learn these tools, you're going to be so much better for it in so many

different ways.

Now, there is a potential downside to this.

And again, that's more of a conversation I'm hoping to have with a couple other people

on here is like, how is it possible to let skill sets atrophy because we're offloading

things to artificial intelligence or a large language model?

Yeah, definitely 100%.

And you should be conscious of that, right?

If I'm not doing this skill anymore, what am I giving up?

What skill set am I potentially atrophy?

Right?

And you might say, well, hey, that's a skill that I think is important to me or important

to my livelihood.

I should really retain that, right?

Like thinking it's a really important thing.

You definitely need to continue to build that skill and work on it and become,

you know, hopefully better over time.

And so you don't want it to just think for you, which is definitely a possibility in a

lot of cases.

And it's not the case I'm going to be making here at all is that you should let it do your

thinking for you.

In fact, I think you should use it to improve things like that and you can if you utilize

these tools correctly.

But anyways, let me share with you some of the things where I feel this has really

benefited me specifically in the automotive side of things in my business in fixing cars.

You can take this out into your normal life and apply it in a bunch of different ways

as well.

Maybe I'll touch on that a little bit, but my main focus is going to be here.

It's the automotive diagnostic podcast, how am I utilizing it day to day in those.

So I've talked about this on the show before.

First thing that really was like, holy cow, this is an awesome tool that I discovered with

the LLM.

Now, I had done a presentation at Vision in, I think that was 23.

I did that where I did, you know, an explanation of utilizing an automotive and I was kind

of like throwing some ideas out there of like, okay, well, you know, if we attach

a PDF to it, because at that point you needed an extension in order to do that now it's so

easy to attach anything you want to it.

But I was kind of playing around with, well, these are the ways that it could be used.

And the way that I use it most right now, I didn't predict.

And I think that will be true going into the future too is like, we haven't even

really come to understand exactly what it's going to look like as this technology progresses

and like the biggest needle that it's going to move, right?

It's kind of like the internet, like everybody had

predictions and they invested their money in a certain way.

And there was a lot of bust there as far as the websites and the money that people

invested because they didn't exactly know, you know, which companies were going to

come out on top and the usefulness in what way.

But obviously, it was a game changer for a number of things, right?

And AI is, you know, definitely fits that bill as well.

Anyways, the number one way that I found myself using this was doing a talk to text,

right, letting it use the transcribed feature where you just hit the little microphone,

you blab, it turns it into text.

And then you feed it into the prompt that that is your prompt, right?

And then what I'm having it do for me is rewrite

my explanation of what I did on a car specifically for diagnostics,

but I've used it for module programming where I need to make sure that there's

details on a ticket, like, hey, this car ran poorly before and after

the ECM replacement or, hey, there's a broken wire at the DLC

that prevented my tool from powering up.

So I jumped power in order to communicate.

But if the shop tries to talk to this car, they're going to have issues, right?

Sometimes I want detail on the programming side of things, too.

But for diagnostics, I will take my phone, I'll talk into it.

I'll give it the explanation of what I'm doing and I'll give it a ton of detail.

An annoying amount of detail.

And remember that because I'm going to be bringing that up

a few different times throughout, but I will give it all of the context

and the detail and I talk and I can talk way faster than I can type.

Maybe that's not true of everybody out there.

But for me personally, I can get a ton of information out vocally,

much faster than it would be for me to type it out.

And again, all of the detail, all of the context, I let it transcribe,

I put it in and I have it set up in a project, which I'll talk about that as well.

And this is a chat GPT that I use, but I'm sure you could set this up

in Gemini or Grock or Claude or whatever version of a large language model.

You prefer go for it.

But I like the chat GPT setup that we have.

Plus, there is a this is maybe often the weeds, but the business version

of chat GPT, at least according to them and what they say is they contain your data.

They're not training the model on the information that you're giving it

so that everything else out there is using your specific data,

which for what I'm doing, I actually appreciate.

Anyways, often the weeds there.

I have it set up in a project to rewrite the information that I give it

in a formal professional manner with bullet points, with a structure,

like concern, findings, conclusion, that sort of stuff.

Like I've given it all these instructions ahead of time of this is the output

that I want from you and I just talk, I press go, it rewrites it,

I copy it, I paste it into the invoice.

Now you do want to check it for all this stuff.

You want to check it because it makes mistakes and I think it says that right

under the prompt bar, like make sure to check it chat GPT makes mistakes and they

do, it can go off in the weeds too.

But the more you dial it in and you give it the correct criterion,

the correct prompts, the less you have to babysit it, right?

It will actually learn what you're after in a lot of cases and just

become better and better as time goes on.

And I'm sure that's improvements on the back end with what they're doing,

right? It's version four, version five, and there's improvements there,

but it will actually learn what you want out of it the more that you interact

with it as well.

Maybe that's a scary thing to say, but it's what it seems like is that I

have to do less corrections on these things.

So the reason I use it for that is it saves me a ton of time and it's

made the documentation on our invoices for what we're doing 10 times

better than what they were when we were typing them out.

And I've typed out invoices since I was started as a technician.

And I remember my old boss saying, hey, like the best thing you can do is put

a story on there and give as much detail as possible because you'll

either need to refer back to that.

You know, if this car comes back at some point, maybe somebody else

needs to refer back to it.

It's value for the customer so they can see all of the things that we

actually did on the vehicle.

It helps us out in defense in case we're accused of doing or not

doing something because it's all there.

So I would type all this out and it was a big chunk of my day.

If you added it up, right, in five, 10 minutes here, five, 10 minutes here.

Again, I'm kind of a slow typer.

I'll just admit that maybe you aren't, but it still takes up time.

Even if you're a fast typer, it's taking up time where this is.

I can blab for 20 seconds.

I can hit go copy paste.

I'm done and it's better than what I would have been able to

type out just by myself.

Now, am I atrophying some typing skills, some writing skills?

Yeah, probably most likely I am.

Do I feel like that is something valuable that I 100% need?

I don't know.

I guess in my opinion, I wasn't that great at it to begin with.

And this gives us a faster and more professional output to the customer

saving us time so that we can move on to the next job.

I guess for me, I want my the time that I spend developing

skill sets to be put in different areas than writing up invoices.

I've made that decision and we do it for our team as well.

And again, the output and the documentation that we have is so much

better than what we used it for before.

Now, that was the first like real big light bulb of like, holy cow.

Yeah, I could be using this every day within the business.

I also transfer that into recording notes for our database, right?

I've mentioned before on here that I record information from cars

that we fix that were maybe a challenge or we learn something new

or we need to capture some piece of data that will help us again in the future.

Now, I've just in the past been using Google Drive and Google Docs

and I'd have a folder for different vehicles.

And when I fix something and I need to write it down, I'd go in

and I'd type out a few notes.

So there'd be, you know, three, four sentences, maybe a couple data values.

Maybe I throw a picture in there, but now I can do exactly what I said

about the invoices, but do it for the database notes.

And I can give it all the nitty gritty details right after I'm done with a car,

by the way, I don't have to wait until the next morning or later that night.

I just do it at the car and then I have that project set up

a little bit differently, different context of like, hey, here's what I'm after.

I want a structured output for the, you know, listing the car,

maybe the engine, what the problem was, what we saw, like measured values,

what a technician working on this car in the future would need to know

and some searchable keywords.

OK. And then I talk into it, I get the output, copy, paste,

and I put it into our Google Drive so that we have a future reference.

And again, was searchable keywords.

So when we go into our drive and we're searching, you know,

a particular code or something like that, it helps us get to the document

that we need to a little bit faster.

So we transferring information that way as well.

Now, what really dialed in the process of those two things for me

was setting up the projects within Chad GPT.

So, and again, other large language models are going to have stuff like this.

Now, you can give your large language model, you know,

overall instructions on how you want it to react or what it want.

You want it to know about you.

There's a setting for that, but you can build a project

which within that project, it has specific instructions.

OK. And this is really important that you give it specific context

for exactly the output that you want when you go into that project.

Right. So I have a project just for creating invoice documentation.

I have a project just for creating database notes.

And those have different instructions.

Right. We've got the engine that is Chad GPT and the large language model.

But it has context of exactly what I want.

And you can attach files to these projects as well.

And for certain things I do, you know, this is the piece of information

that I would like you to reference.

And those are two really important things to understand

about these large language models. OK.

Now, there's probably three really important things to understand

is that giving it the instructions and the context that you want

is really important.

And you can do this as an overarching

umbrella for within that project.

So everything you do, it relates back to those instructions.

Right. But then within the prompt itself, right,

that's the other really important part is how you prompt the large

language models completely changes the output that you're going to get.

And you can give it very vague prompts and you might still get something

that's OK, but it's very inaccurate, or you can give it excruciatingly

exact prompts with all kinds of context.

And I would suggest doing that in most cases and going back

to what I was talking about with the invoice documentation and the database.

Just talk to it again.

If you're a really fast typer, then you can type two.

But I found it that I can just blab.

And again, giving it an annoying amount of detail about what you're after.

You know, say I want it like this, but I don't want this.

And I want you to look and consider this.

Oh, and by the way, also that give it all of that stuff,

which if you were to talk to a person and say all of those things,

they'll be like, hang on, I got to write all this down and consider all this.

But it's able to take that into the process when it gives you an output.

But I found the more context, the more information, the more guidelines

you give it, the better output that you get.

OK, so that's prompting and then using the instructions

and putting some effort into making those instructions as well.

There's some stuff there that you can do to make some really good

instructions for these projects, but then to prompt well,

as in addition to that, you start to get the best outputs possible from this.

Now, the other part of it is information.

This episode is brought to you by L one

automotive training and Keith Perkins.

If you're looking for education on module

programming, J 25, 34, eProm work, key and immobilizer,

electrical diagnostics or drivability diagnostics.

Keith has a website, L one training dot com, that's got over 60 hours

of training videos on all those subjects and more.

When I first started out doing mobile, I utilized Keith's videos

on module programming and J 25, 34 in order to get my head wrapped

around what I would need for the tooling, the computers,

the software setups, what kind of obstacles I would be up against

when I'm out there programming modules on cars.

And it was a huge benefit to me.

And I continue to use the training videos that he has on his website.

So I strongly recommend checking out L one training dot com.

The link is in the show notes.

And that's where these things really shine.

And where I feel like the average user of a large language model

doesn't understand the power of these things right now is what

information they have reference to.

And this was way back when I did my presentation ad vision and I

did a class down in Orlando on it is the power of these things.

Right.

The engine behind it is wild, right?

The GPUs and the transformers and I'm, you know, I'm weak

on understanding exactly how that all that stuff works.

I think people that build these things have also admitted

like they don't really know everything about exactly how it works.

But man, it works really well.

Like that's so powerful.

But in order for that to really benefit us and mean something to us,

it needs the correct information.

And, you know, I've seen people send me stuff or post stuff about like,

look at this stupid answer this thing gave, like, oh, we're fine.

This can't figure out anything.

And in most of the cases, it's because of the information

that you've either given it, not given it, or it has access to

that really changes its ability to give you the proper answer

or what you're looking for, the output that you're expecting or wanting.

Now, when it comes to, you know, experts in a particular field or subject,

humans are still probably the best in real specific areas.

But again, it just comes down to, well, what information does it have access to?

So you want to consider that.

Now, again, within these projects that you build, you can attach certain files.

OK, you can attach a 300 page PDF and say, hey, I'd like you to reference

this when you're working in this particular project.

Right. This is the user's manual for Otis.

This is the user's manual for QuickBooks.

This is, you know, the key reference I would like you to use for this

project for these outputs, and you can build as many of these projects

as you want, at least I haven't found a limit yet, for different tasks.

And so when you have a specific thing that you want to get done,

you go into that project and it's already set up to cater that output

to what you're looking for in those instructions.

And that changes how you utilize this just absolutely in leaps and bounds

instead of just hopping on there and giving it a prompt for something.

The other thing here is, too, you can prompt with more than just text, right?

You can give it URLs, like links to things.

You can put files in there, whether that be an audio file, a spreadsheet,

even scan tool or scope captures, right?

If you save them in the right format, put them into a CSV or a dot mat file.

You can have an analyzed pico scope data and it's pretty wild.

What it can do without a whole lot of background.

I posted something in a group just recently where you put it.

I put in a relative compression and vacuum waveform.

And I didn't really give it a whole lot to go with.

I just said, hey, Channel A is starter amperage.

Channel B is a pulse sensor in the intake.

I'm cranking the engine.

Let me know what you think about the mechanical health of the engine.

And it figured out the RPM during cranking.

It figured out that it was a six cylinder engine and it gave a really good breakdown.

And then it asked me if I wanted to do four other things with it.

And it was really impressive.

And you have to play around with that specific application of the pico files.

But you can get stuff in there, right?

Now that's different than just, you know, just putting a picture of a

waveform into a prompt, which you can do.

You can do images as well.

And I think the ability of it to assess images has improved.

But it probably is a lot weaker than something actual that is data based,

whether that be a spreadsheet or, you know, something like this, where it has

actual, you know, data points they can assess.

But so that's that's the thing is like, OK, well, how do I get

it the information that I wanted to assess?

What form can I put it into that it works with, right?

And here's the other thing.

And this is what I kind of want to circle back to is.

If you work with it, you can ask questions of like, hey, this is my goal.

I want to be have you help me with this.

How do you think you'd best be able to do that?

Right. What do you need from me in order to get X output?

And then it can kind of teach you how to best interact with it to get the best outcome.

But that's not the only thing that you can learn from using these tools.

And this is where, you know, the hey, we're just going to offload

all of our thinking to this and that's definitely possible.

But you can flip that on its head and you can actually learn a ridiculous amount.

Now, OK, you don't have to be careful because where is it pulling its

information from? Sure.

But you should be careful of that for everything, right?

Podcasts and books and YouTube and I mean, where do we learn stuff from?

How often is the information that people are going to be basing

their content on incorrect or missing key information?

But we're taking that as, hey, this is the gospel, right?

We're this is how it is.

But they're missing information.

And it's the same thing for a large language model.

So, you know, take things with a grain of salt.

But but if you can give it specific context,

maybe you give it specific information, maybe you give it a PDF

on a user's manual of a scan tool you want to learn and say, hey,

here's my goal.

I want to be very proficient in this particular scan tool application.

Here's the information.

Let's build a plan that you can help educate me on how to use this.

And then you can have it quiz you and say, hey, OK, you know,

we'll ask you a question, you answer the question back and then,

you know, assess me and help me so that I, you know,

make sure that I understand this correctly before we move on.

And you can do this in a voice mode to there is a voice mode

where it just goes back and forth with you.

You don't have to type anything.

You can do this while you're driving around in your car.

You can be learning things if that's a way that you want to spend your time.

But for me, it is I want to increase my knowledge in specific areas.

And this is a tool to do that, that, you know,

very powerful on the technical side of things,

as far as processing information and creating questions about data

to assess whether I understand that or not.

And then here's the other thing, too.

Is it's never going to get tired or bored or annoyed of my questions.

I can just keep going with it and going with it, right?

And I mean, going down that road, you know,

there's a real possibility that AI replaces a lot of education in the world

that's looking, you know, farther out into the future.

But man, you can learn so much interacting with these tools.

And that's the way that I'm trying to utilize it in that regard,

if not just replacing the functions that I would normally do,

but amplifying the stuff that I already do or want to do more of

or want to learn more of.

And I can tell you, I have learned a ton about so many different subjects

using this well beyond automotive as well.

Things that just were kind of maybe I considered like out of reach

or I don't even know where I would begin to learn about this particular subject.

But you can ask and it's actually pretty impressive what it can come up with.

Now, like open AI is trained chat GPT on the internet, essentially,

and a lot of the other large language models have as well.

So again, yeah, there's BS information out there, but you can fact check it.

You can ask it, like, OK, well, how accurate is this information?

You know, is there a potential bias here as far as what I'm getting?

Again, but you're going to find that with human information as well, right?

So no matter what you're doing here, you have to be careful with that.

But it's really impressive what it can do with the information that has access to.

And again, you give it the context that this is what I'm after.

This is my goal here, the criteria I want you to work with.

It's crazy what it can come up with.

The last thing I mentioned on here that I found has been useful

and I'm still developing is just a sounding board during a challenging problem

like diagnostics, like programming a control module on the board level.

It's actually it's pretty impressive.

If you if you snap a picture of a circuit board where you can, you know,

see the chips and the legs and the detail, it's really impressive

what it can come up with.

Now, it's not always accurate, definitely not.

But it can get you moving along in the right direction

and it can find chip families and it can find programmers

that work with particular chips and you can get off in the weeds for sure.

But you probably are going to get off in the weeds all by your own,

too, is at least a guide, but you can use it as a sounding board

as you're going through, like, hey, does this make sense?

What I'm doing or seeing right now and you can tell the car, like,

hey, I'm working on this particular car.

I have this particular problem.

Here's what I've tried here.

Here's what I'm seeing.

Can you help me, you know, brainstorm some ideas

because I'm at an impasse right here.

And sometimes it'll pull really useful information from the Internet

that just, you know, silver bullet.

That's not what I'm after here.

I'm after of like a back and forth of like, OK, all right, try that,

try that, try that.

Oh, OK, that's that's something I hadn't thought of yet or just, you know,

it didn't use in my process.

Maybe I skipped over. OK, that's a really good idea.

And if you're able to couple that with some of your own information

that you've gathered together over a period of time,

it becomes even more powerful.

And that's what I'm working on right now is linking the database

that we've created through S T mobile to one of these large language models.

And there's companies out there that have done, you know, something like this.

We just talked to Brendan a little while ago and they have their

the tech Tina that is a language model that has information,

you know, based off of their forum that you can interact with

that it has reference to when answering questions.

We're working on our own, you know, internal database there

so that we can have that set up as well.

So like all of the data that we've captured over the last five years or so

that I've been recording stuff can then be accessed by a particular model.

Now, you do want to consider the like, hey, does that mean everybody

that has chat GBT can access this?

Again, according to what they have, the business model does not.

So you do want to consider like, you know, what am I feeding out there?

I think having information that is not publicly accessed

is actually going to be really beneficial in the near term

because these language models are trained on the open internet in a lot of cases

and stuff behind a paywall is still that behind a paywall.

And so you can use that as an internal database, build your own model.

It's a lot of work that goes into that, but you can still retain that for yourself

or for your company, which gives you, you know, maybe a little bit of an advantage

in a particular aspect or, you know, this is going way off,

but it's a potential where you could sell that bulk of data

that you've collected off of real world experiences at some point or another.

You decide like, OK, I'd like to, you know, cash in on this at some point

or another. I'm done using it. It's a lot of data.

You can't get this unless you've had these real world experiences

and, you know, potentially down the road, you know, some company is willing to buy that.

I don't know. That's that might be just speculating.

But we have worked really hard over the last few years

to capture as much data as possible.

We continue to do it every day.

And I think that's a pretty important part of what we're doing specifically

because we run into new problems almost every day and continuing to build

that database really helps us in the future as OK, maybe we see that same problem again.

Maybe we see something similar.

But now, again, with technology that's out there, we could have a large

language model help us interact with that bulk of data to help us during the day to day

with that diagnostic assistant, if you will.

One of the things I'd suggest everybody try

if you haven't is the deep research mode.

So when you go to put the prompt in, you can look, you can pull down

and you can see deep research.

Now, what this does is it takes probably about five to 10 minutes,

depending on exactly what you're after.

You're giving it the context.

And a lot of times it will even ask for more details before it goes off.

And then you set it off to do this research, if you will.

And it does really in depth search based off of the question

or the outcome that you're looking for.

And it will come up with a really detailed report.

And again, the context that you give it, the details that you give it,

the specifics for your outcome all change here.

But it's much, much more detailed and in depth than what you would just get

from the normal prompt.

You'll get links to certain web pages.

And it will find things that the normal search, if you will,

just doesn't come up with or is less detailed about.

But you do have to wait.

And depending on your pay plan, you only get so many of those a month.

But what I'll do here is if I'm dealing with a weird problem,

something that's kicking my butt, something I haven't dealt with before.

I don't really seem to have any more direction with my normal paths.

I'll set this up and say, Hey, here's what I'm dealing with.

Can you help me find X, Y, Z?

Here's all the context.

I'll hit it and usually we'll come back with a few more questions of specifics.

OK, do you want this and how about this?

And do you want me to blank?

Yes. No. Here's what I want.

Come back in 10 minutes and you get a really good detailed report.

Now, sometimes there are things out there that there's not that much

info out on the internet or whatever resource it has.

So you can't do much with it.

But sometimes it's like, oh, wow, this is way more information

or for me to find all of this information through like old school

Google searching would have taken me hours, hours, if not days

to get all this, you know, organized and put together in this detailed report.

But it's been really good to get me going in the right direction

on a lot of stuff.

You know, we're up against it, especially doing what we're doing here

in the mobile world, working on all makes and models,

taking on the weirdest problems and have an assistant like this

and a resource like this is so huge.

So that's kind of where I'm going to wrap this one up today.

Again, I'm hoping to have some people on to have

another conversation about this, you know, AI in the automotive space

in general, and maybe some of the downsides or things

that you really should consider as you use this.

But I also want to share like, hey, I'm using this stuff every day.

And it's been it's been a game changer for me in a number of ways.

And I'm still learning how to do it.

And I think that's really important is that you learn how to utilize these

and then you can, you know, decide whether it's worthwhile or if it makes

sense for you. But those that learn this stuff in the short term,

at least, are going to be very successful compared to those that don't.

That's my opinion. So anyways, hopefully you found that interesting.

If you've got some interesting use cases for AI or large language models,

let me know, share it, put it in the Facebook group.

Love to hear about it.

I'll share what I've got there, too.

But other than that, let's all get out there.

Start fixing the world one kind of time.

320: How I'm Using AI Tools In My Business Right Now

About this episode

Request an Explanation

Want to learn more?

More from Automotive Diagnostic Podcast