This is a YouTube video titled "Practical Uses of AI for BDC Managers- Session 4" containing a session recorded in August 2025. The content features a presentation or workshop format with multiple speakers, primarily Peter and another speaker, discussing how to use large language models (LLMs) and AI tools like ChatGPT to build custom AI assistants for dealership BDC managers and staff. The session focuses on practical, hands-on guidance for creating custom GPTs tailored to dealership data such as staff directories and employee handbooks to improve daily operations and information access.
Me. So welcome in everyone. We have our fourth and final session that Peter has been working on and sharing with us in terms of how you can use the LLMs and, and these new AI tools to help you in your day to day with the bdc. So what are we going to be talking about today, Peter?
Well, today is, you know, in the, the boot camp world we have a number of developers that have gone through coding boot camps and all of them have some version of what would be called a capstone project. So today we're going to summarize pretty much everything that we've talked about into one final project that should make it easier for our BDC managers, our agents, our dealership managers, etcetera to come up with a pretty neat solution that I only wish was available to me 30 years ago when my, when I started my career. And it's a pretty interesting way of giving your employees the tools they need to be successful in their jobs.
Awesome. So are we going to be walking through slides today or.
This is all hands on. This is all what I would call lab format. So I've done a bunch of practicing, I've built a bunch of these and some of them for my team, some of them for my kids. And then today we will be going through one that I've built for Volley as kind of a test and then we'll actually build one custom GPT live that is based on some mock data for a dealership.
So to say this back in terms that I understand as a non technical junkie, what we're doing today, last time we spoke about keeping some of the scripts and prompts that would help you be able to, you know, rebuild, you know, get chat GPT or something to spit out what you need on a daily basis, but you kind of need to start from scratch each day, so to speak. Today we're going to be talking about building your own friend, if you will, who lives in perpetuity, who will know your dealership, know your bdc, know your policies, et cetera, that you can ask without needing to start over each day.
Is that 100%? Yeah. Okay, 100%.
Well, good. Well, let's, what do you say we get started?
Sure. I'll share my screen here and we'll, we'll get going. So you should be seeing ChatGPT in all of its glory. It says what's on your mind today. So as we talked about in the first presentation, GPT stands for generative pre trained model. And the training data that these companies like Meta with llama, like Claude like OpenAI with chatgpt, what they have done is they have gone through and they've given these, this, this LLM, the large language model source data and in a lot of cases it's the public Internet. So Google built its empire based on understanding what was on every page and ranking them in such a way and returning them to the user so that you get the best search results when you type in what's a, what's a great recipe for chicken tacos for example. Well, when they started working on these LLMs they did effectively the same thing they gave the LLMs or once they had the software developed and in you know, prototype phase etc. They started unleashing these LLMs to learn on the public Internet and maybe some private Internets or private networks etc. But largely the public Internet. And that is what we mean when we talk about a model. A model is simply a predefined set of training of trained data and so it's going to spit out predictable responses etc, well what Chat GPT and some of these others. But today we will be going through it on chat GPT just so that you have a definitive concrete how to out of today's session. What chat GPT has done is given you the ability to effectively add on bolt on functionality to that to those pre trained models. It still uses a lot of the underlying business logic and the algorithms etc. But it allows you to really curtail a custom GPT. Maybe it's a dealership specific based, maybe it's your dealer group based, maybe it's your kids soccer team, maybe it's you have a recipe group or a book club or, or it doesn't matter. All of these are ways that you can effectively bolt on custom functionality and custom data to use the underlying chat GPT framework but based on your own data or a data set that you constrain. And like Jake said when he summarized really instead of for all of the three sessions prior to today, what we've really been talking about is giving you a prompt that you can then take and hand out to your people for them to be able to really confine what they're looking for. But that's kind of a pain. It's not very scalable and ultimately it doesn't allow you to control a lot of the parameters that, that you might want your agents or your team leads or your, you know, maybe it's a case of a service advisor, etc. And what we will be doing today is building a custom GPT where All they have to do is click on a link and and they can get responses that are based on data that you've uploaded and it allows you a much more scalable way of disseminating information in and amongst your team. And at the end I'll go through a few clever ways that I've walked dealers through and I've helped dealers build GPTs when I go out on site visits. A lot of people have been watching this series and I get a lot of questions. If you don't ask questions on the webinar when I come and see you, I can nearly promise that you're going to ask me questions, which is great. I love it, I love to help and that's what we're here for. So today's will be chat GPT but understand that other of the LLMs allow you to do similar type functionality. And so if you, if you look on the side here on the left hand navigation, what you'll see is, is you'll see an expanded menu list. Now I will tell you that fundamentally what we're going to be going through here is available on the paid chat GPT platform. I don't know what functionality is available on non paid in terms of generating custom GPTs, but I'll just walk you through. For me, I pay 20 bucks a month for OpenAI and that gets me all that I need. I use others as well, but that gets me all I need in terms of custom GPTs. So there's this GPTs link and I'm going to walk you through this one that, that I've already done but then we will also go through how to create a new one. So I click on the link, I'm gonna have to get rid of this over here, maybe down here. And it is bringing me to this GPT that I've built using our staff directory which I downloaded from our HR platform. You could create a Google Doc and export it as a CSV. It doesn't really matter but just some way where you can get this data so that you can upload it. And then I've also uploaded a copy of our most recent employee handbook. We just had our annual summer meeting and we went through a new version of the handbook and so I happen to have that. And I figured maybe this would be a fun tool, Jake, for our staff if they had any questions that they could ask our, they could effectively ask to our employee handbook. So I'm going to go here and I'm going to edit this GPT and that's going to bring me into the editor mode, which is really where the power of this lies. And now I got to move this again. We're playing tag here, but I give it. The first thing I can do is I can give it an image. If I wanted to put our logo here or put something else here, maybe it's the Valley. Bdc.com, whatever it might be, I could put the logo here. I give it a name. So in this case, I've just given it a pretty descriptive name and then a description. So I've said this provides a quick and easy way of interacting with the Volley Company handbook. And then for the instructions, this is where you take all of what we've talked about.
