If I'm a car dealership and I'm a marketer there, I would expect my AI agent to show up for me in the morning and say, Hey, we just, you know, we noticed like three or four things last night. This competitor sold these cars. The price on leads is dropped here. We want to take $4,000 and dedicate it to this or we want to launch these 15 new email campaigns. Do we have your approval? And you can then just say, sure, yeah, yes, launch, don't launch, snooze, delegate, pass it over to my owner to approve it. Whenever it is, the AI will be able to do that.
What's up, everyone? This is car dealership guy. You're listening to the car dealership guy podcast, which is my effort to give you access to the most unbiased and transparent insights into the car market. Let's get into today's episode. Aharon Horowitz is the founder and CEO of FullBath, an automotive tech company that focuses on improving the car buying experience for consumers by helping dealers properly leverage their data.
In this conversation, we spoke about the secret weapon dealers are using a compute with big box stores. The crazy amount of personal data that dealers have about you, how AI and Chad GBT has changed in the car business, and my personal ambitious vision for AI-powered car buying. I love this conversation really energized about the future of car buying. I think you will too.
Here's my conversation with Aharon Horowitz. All views of car dealership guy and guests on this podcast are solely their opinions. None of the views expressed should be treated as financial advice. This podcast is for informational purposes only.
Aharon Horowitz, I want to understand something. How does a small Israeli tech start? I managed to infiltrate the US car market so well. Give me the background here.
That's a great question. We were not originally intending to focus on automotive. We just thought we could build some great technology out of Israel for small meeting businesses more generally. Why would you think that? Why would you think that? Israel is this really. I've been as you can hear from originally from the US. I grew up in Cleveland, Ohio for various family reasons and sort of personal reasons. I moved to Israel after high school study there, but served in the military. Then I came back to college in the US, but I always knew I wanted to go back to live over there.
After college, I did. I was just blown away by the level of technology talent and also just grit and determination. It's just a very start-up country. I guess given the context of where it is and its history and how it's built itself over the last 70 years, it just produces people who don't get stuck on obstacles. We were just really caught up. I was caught up. My co-founders, Yishin Eliav, we're just caught up in the energy of we could build something great with technology.
We were really obsessed with small meeting businesses and the impact that happened. Communities and we thought they were just like a net good of capitalism. We wanted to just build a start-up. That would build great tech that was very automated and give SMBs the power of big companies like Walmart and Amazon and Netflix. We wanted every business to have that power. That was our ideological vision.
We founded the company. It's zero knowledge of the auto industry at that time. Going way back. One of our advisors who had spent a bunch of time in the auto industry, he was like, you have to go to car dealers. They need this stuff. I was like, no way. They're not going to be interested. I don't know anything about car dealers. It's some point he convinced me to fly to Cleveland where I'm from. I just got a bunch of dealers. I went to Colorado, met the LWFOKES quirk in Boston. A bunch of dealers just let me hang out. I was like intern in their BDCs and with their salespeople. I was just obsessed. I was like, this is going to be it. We're going to be called an auto tech company and we just did it. We get a great tech company. We get a great market. A lot of energy, a lot of opportunity. That was the story.
Explain me. What is the problem, fundamentally, that you're solving for dealers and consumers?
请解释一下,你们最基本的解决的问题是什么,为代理商和消费者提供什么帮助?
Listen, dealers have an unbelievable amount of data on their shoppers that can really help them give the shoppers a better experience. Yet all that data is just laying all over the room and disorganized piles. Sometimes a piece of data about someone is in one corner, the rest of the data is in another corner.
What we basically do is put software in for the dealership. We connect to their systems. We get all that data organized. We get a 360 view of every customer settled for the dealership. Then having data is just half the story. You still watch that cartoon GI Joe when I was growing up. It's only half the battle is having the data.
The other half of the battle is actually leveraging the data and using it for something. We then built all these automations to take that data about their customers and just do the marketing work in a way that a machine can do that is way beyond what a person can do. It really just solves for the dealers.
It's like upside opportunity for them. If you think about it, a lot of these dealers we work with, they have 10, 15 years of leads in their CRM. That's very, very powerful, very significant. Tell me about it. I don't know how to access that and use it.
Yeah, it's like, you know, oh, I have 370,000 people in my CRM. How many of those are you actively figuring out how to engage? I don't know, maybe like a thousand, two thousand, ten thousand a month. It's not it's not a data strategy. That's really what we've thought about for dealers.
By the way, I'll go further with that. If you think about it, your average shopper buys like nine cars, I think I don't know, cops just won't put it out. It's like nine cars over the course of the shoppers lifetime. And you add in service revenue and you think about gross profits, not the past few years, but over time, you're probably talking about a $50,000 opportunity for dealership on the shopper. Yeah, exactly.
That's the LTV that you could go after. Now, most dealers don't really have a way to look at their data and say, oh, wow, here's so much LTV I have left on my career customer base. This shopper is at this part of his or her cycle of purchase. And, you know, and so we think technology can do that automatically. And that's what our tech comes to do.
We basically put in these systems, get all the data into one picture, and then we run digital advertising email on site engagement. But we do it from a place of data and insight and having a thesis about how to get the right experience for that shopper. Now what shoppers should experience is a much more coherent buying process.
Yeah. And that's my question, right? So before we get there, actually, so you mentioned that we have like a cohort of customers from our use cardio ship that I was talking to like the hundreds of thousands of emails. But we attempted to retarget them, you know, years later, to your point, like we saw this opportunity, hey, we have all this first party data. How can we leverage it?
And you know, for the audience, as listening first party, just meeting its data within our own systems that, you know, people have given us at some point of interacting with our dealership historically. Well, we realize is that going back that far, people were like, who are you? Like, what are you doing? Like they forgot that they actually signed up for this, even though they fully legitimately signed up for this.
Have you ran into any of these type of issues where, you know, you're maybe trying to leverage first party data from a dealership, but the data so outdated that maybe you present some challenges with consumers. Like, how's that played out for you?
Yeah, it's an interesting question. I mean, I'd say two things. Number one, part of it depends on how compelling the the first message is or the second or the third, because people don't often see the messages that come in initially.
So if you could generate on those 300,000 people using like a model, like a collaborative filtering model, right? That's how Amazon or Netflix say, hey, people who want this also buy that, right? If you could look at all your inventory right now and say this guy from 10 years ago is most likely to be interested in this particular vehicle I have on the lot now, you could do that and make that the message.
They might find that a much more valuable email that if it was just like, hey, we've got a special going, this is who we are, right? That's just not a relevant email to most people from 10 years ago. Now, the other thing that happens is it also cleans your list. It gives you real list hiji.
So you want people to unsubscribe and you want to make it a really clear unsubscribe. So that when the shop or gets an email based on data, if they're still not excited, then let them unsubscribe. That cleans up your list and let you have a evergreen current list.
And I'll note there that like, we don't see crazy unsubscribe rates when you're getting really targeted one-to-one messaging. We think that people find that to be actually relevant and interesting if it's based on something.
So it explained this to me. What do you mean by like how targeted it is as a consumer right now? Okay. Any consumers listening or even myself, I'm curious.
What would I expect if I was dealing with a dealership that's leveraging this technology? What would that experience look like for me? Full funnel.
如果我与一个利用这项技术的车行打交道,我会期待什么?这对我来说会是什么样的体验?全渠道。
Look, the unfortunate thing is that shoppers are working on so many different platforms that don't talk to each other. So you're going to research on a third party like a a gurus or a cars.com and then you're going to go to a KVB and then you're going to go to a dealer. Most of those do not sink.
And so no matter what shoppers should expect to have a bit of a disjointed experience. But let's say I'm a shopper and I go to a dealership website. Most shoppers go to four or five dealership websites as they research. And I go to a dealership website.
And let's say I browse and search and types of cars. I look at trucks, right? It's certain models. The dealership knows that. That's in their data. And they see that in their systems. Whether they record it in a intelligently or not, that's a question of what tech they're using.
But I should expect, in at least in a dealership that's working with full path, I should expect the next email that goes out not to be about a sedan or a special that has nothing no relevance to me.
我认为至少在一个全面运营的经销商那里,我应该期待下一封发出的电子邮件不是关于轿车或与我无关的特价。
I should expect it to be about probably the vehicle or the class of vehicles I was looking at. And in five years, I should also expect probably to get some information about that as I'm going through that life cycle, a relevant car that fits sort of what I when I wander what's predicted.
I'm going to be interested in based on other people's behavior. So that's like an intelligent system. Moreover, if I fill out my name in one form, I shouldn't have to fill out again in another form, ideally, when I'm, you know, browsing through the side and I should be able to know that the salesperson understands what I've been looking for and when I'm interested in when I walk in the door and I don't have to have a start from zero conversation.
So it's all that layer of like connected retail that we've kind of come to expect that doesn't often exist right now in in auto because the systems are so broken apart from each other. So that that's what we saw. That's the bridge that we're going after.
It is a shopper. It's meant to be a better experience. By the way, you're a little creepy, right? Someone knows something about me and you know, you're seeing that now with the AI stuff, which we'll talk about later. I assume, but you know, there's a lot of every website knows a lot about each shopper.
Yeah. They do with it is a matter of law and it's a matter of how respectful they are of shopper privacy. Yeah.
"是的,他们如何处理这个问题涉及法律和对购物者隐私的尊重程度。"
And look, here's the thing, right? The car business is very fragmented. You have tens of thousands of dealers throughout the country. Yes, there's a trend of consolidation, but we're still a super fragmented industry.
And to your point, it's all like one corporation is just connected throughout every single state, every city and all for this experience. So I'm sure everyone is called whether you're a dealer or a consumer or an investor or whoever is listening, I'm sure they've had this type of experience.
And by the way, as you were speaking about the shopper experience, what's the deal with like dealers having like seven calls to actions or call to actions on their website? Like you have to have an opinion on that.
Listen, from a shopper experience, it's very frustrating. Well, there's so much going on. And I think there are examples of sites that are much cleaner.
从购物者的角度来看,这非常令人沮丧。嗯,有太多事情发生了。我认为有一些网站是更加清爽的例子。
But I will say this that unfortunately, the thesis is seeing this from the other side of the table. All vendors have to make decisions on what the what their principles are because when you're on the other side of the table, the balloons and the the funny man, you know, the vacuum that's making them blow up, whatever those are called in front of dealerships that those things. Yes, exactly.
Exactly. They actually work on websites. Is then if you bug someone to give you their information over and over again, they're just they're like, I whenever find me take it and let me move on and just like get top to a human.
So I think the regression to the mean is unfortunately to be extremely aggressive. What many of us and I'm not just speaking for us as full path, but many vendors do is we accept a certain loss and lead conversion for the dealer in order to give a better shopping experience.
And that's a, you know, that's a principle that I think will just increase more and more and more of the time. Now, if you build a really savvy, clean website that really knows how to work with you user experience, then you will get much better connections than driving some much better conversions than driving someone crazy, but it's few and far between. And the reason is is that so many different people control so many different parts of the process.
And by the way, just in general, like going back to your question of like that, list you guys had, you know, a couple hundred thousand that you tried to get people to act act that if you combine systems, if you control more of the funnel, you're going to get better results. Like, for example, we found that dealers that use their CRM. So like they they take their CRM on each is in every day three, four times a day, they upload those to Facebook and to Google for their advertising. They're going to see a much lower CPL cost per lead than dealers that are just using non proprietary first party data.
If we also see, for example, that like if you email CRM customers and combine it with redargeting, then you get a much higher close rate on those emails, right? People see your name on the ads and then they're more likely in their inbox to actually open that email and come through. So it's all about combining systems and making it easy. Basically putting the breadcrumbs in and sprinkling everywhere makes total sense exactly.
I want to dive more into just marketing dealerships like what's effective, least effective, but what's the context on full path? Like what's your dealer account? How many dealers you have signed up? Your growth three, you know, how much money you got into bank account? What's your social security number? And now I'm just kidding. No, social security door. 5, 5, 5, 5, 5, 5, 5, 5. No, our, our, uh, we look, we, we, uh, feel very blessed to now have about a thousand dealers that we work with.
And that's been a, a real journey because, um, in many ways, like, we always think like the dealers are, are, like, product managers and, you know, we, we just try to hear the challenges and problems and build enough stuff that they want to tell their friends about it. And, and, uh, we did recently, expand our sales team, but a lot of our growth over the first few years, which is literally word of mouth and within groups. So we're working with about a thousand dealers, uh, which, you know, which is, I don't know, it's a public information, but now it is, but anyway, it's about a thousand dealers.
And we've grown. Yeah. I did some homework on you guys. I saw you scooped up someone from dealer inspire and, you know, some other great talents. That's true. I get my, uh, my, my, and then you get some great, great folks on the team. And other great people, I'm sure, just someone that, you know, peaked my interest.
Yeah. We were really, really blessed have a great team and, um, folks who were very diverse in, in terms of the perspectives. And then, and then what's your, your growth rate? Yes. We've been growing like, we've actually been doubling the past three years. We've doubled, uh, each year, over 90%, 90, you know, 94%, 95%. So it's been really good. That's been, that's, um, not necessarily a dealer cop, but that's in, in revenues.
And, you know, as we both add dealers and release new products, we add more value to the dealership. They reward us by, you know, paying for that value. So that's the, the basically, because by the way, most of our business model, unlike I think some of the, some of the others in the industry, like a lot of our business model is, uh, is sort of a SaaS fee model. We try to do that wherever possible. We stay away from more from percentage-based stuff.
Explain, explain, explain this to us. Our audience loves to, uh, do you see details? Yeah. Yeah. So like our digit product, other than where we're, some, OEMs force us to, to go with, um, spend tiers, but we prefer it just to be a flat fee, no matter how much you spend. So, you know, we, we have a, a flat fee, we can, you know, if it's a tiny dealership, there might be one fee. If it's a reasonable-size dealership, it up to different fee. And then that's it, you know, spend as much as you want as little as you want.
And, uh, it's a software fee. So we really focus on like a tech fee, as opposed to an ad, did you add type of fee? That's on our digital advertising. And then for our audience activation, which is that email infrastructure, uh, where we can really activate audiences and take, you know, 100,000 people and get everyone at pinpointed targeted message. That's also just a month. It doesn't matter how many emails you said, well, you have, it's just a monthly fee. So we really work on like a flat fee model.
We think it's cleanness for us and easiest for the dealer. Uh, they don't have to worry about, you know, incentives being misaligned.
我们认为这对我们来说是干净的,对经销商来说最简单。嗯,他们不必担心激励措施被错误地调整。
Now, Eidook, I obviously love marketing, right? Yeah. I don't think it's a surprise with what I do. Yeah. But I'm curious from your perspective, just I see one business. That's my, that's my sort of my aperture here. You see, you know, 1000, 1000 different dealerships. So what has been the most effective types of marketing that you've seen from the dealership perspective? And also I'm curious what have, what's been the least effective types of marketing?
It's very clear that people still want to know what their payments are. So I'll just say that it's a general point that like anything that has like information about payments, shoppers really feel, you know, feel good about. But what we see to be the most effective is, um, it is really like, again, speaking to the dealer audience here, it's just, it's really first party data, I'm driven marketing.
So if you can activate your first party data, you know, from your CRM or your DMS, what have you that like this person is probably paying this amount of month and you can suggest for them, uh, a lease offer or a relevant vehicle. That's like very much untargeted with what they care about and how they pay. That's what converts and you could do it in almost any medium.
So you could do it in advertising and you'll get a higher click than you will other stuff. You could do it in, uh, in, in email and you'll get it, you'll get, you'll get results. But you're not going to get results off general stuff. Dig a little deeper in that, right? Like how much are you really able to tell about a customer?
Uh, well, I think you're able to tell a lot. I mean, think about all the information that, uh, imagine, uh, we were able to see, um, where they are, we're able to see their name, we're able to see their, uh, their email, we're able to see what pages they've looked at on the website. You know, and so again, no one looks at this, no one's granularly looking at the human, but the system is looking at it.
So for example, if they're looking at the financing page and maybe they're even hovering over certain things, right? Images they're clicking on or specials they're clicking on. Those signals are tracked by uh, the dealership systems and by the vendor system, right? And then you can, if you have the ability, which is one of the things that we've been, um, most of our, our efforts are in this, is to take those signals and then actually construct, dynamically, the marketing message for each person.
Um, now on digital advertising, you can't really get personalized at that level on our end, but definitely with the outbound email, uh, you can, it definitely with when someone comes to the website you care. So if you could actually figure out, hey, a person, this person is in the market for, uh, a minivan and they can pay this amount and therefore these offers are probably relevant to them or these vehicles are relevant and you show those, you will get an interested person to call the dealership and, well, actually stop by.
In that level of targeting, we do every single day, and again, I never like to be like, we're the only ones where I'm not here to sell. I'm just saying, we do that every day. I'm sure others are doing that every day, but it's very, very granular and that's, that's probably ultimately good for the consumer if you're dealing with, um, you know, companies that, that respect privacy and, and do things right.
Uh, and, and I think that that's a, that's a key piece. Yeah. Like I personally, I love targeted ads. I don't think they're creepy. Again, to a certain extent, but I, I like that where it's, you're just saving me time if you do it effectively. So what you're doing is, you know, you've found this like uncapped resource. And you're just, you know, within that dealerships have organically, you know, accrued over the years, and you're helping them monetize that in effective ways by increasing conversion, increasing shoppers, you know, increasing sales.
And I think it makes total sense because you're doing that and making a better customer experience. So yeah. Now, I see this as a, as fighting, they're, we're just giving it, we're arming dealers with the tools that every business uses. And there's no reason why a, an amazing industry that employs a couple of million people that pay as like really good wages that's rooted in communities in a way that most American businesses have long kind of turned their backs on.
There's no reason those folks should suffer from not having the powerful tools that every, you know, every major big box has a big black retailer. And so we are very like ideologically driven. It's like, like, give this industry the same power. And that, that's all, that's the way that we see it.
Before we transition to AI, because I know you're working on some interesting stuff there with Chad GPT. And I've been, I've been reading a lot about the space. I'm, you know, very interested in it.
What are like the core opportunities you're, you're capitalizing on, right? Is it, is it all related to first party data? Or is there, you know, is AI now sort of a tangent from that or explain to me, you know, high level? What are the core opportunities that you're capitalizing on?
Yes, we give dealers like a full operating system further like front of the house marketing and further data. So we, we, we are not like, we're not only first party, we do all the different marketing functions that you'd expect, right? All based on the data layer that we create. So we, we run digital advertising like it's an agency, right? Even though we use, we more automation data. We run email marketing like it's being run by, like, you know, someone sitting in the loyalty department. We do all those functions, but we use technology to do them and we use the data layer that we create. So the opportunity that we're going after is, is what we see as the full digitization of all the dealerships, outbound and customer facing activities.
In terms of AI, we don't see it as a tangent, we see it as a fundamental, you know, phase shift the way everything works. So artificial fucking intelligence. Yeah. Before we dive deep, give me like the five to 10 year vision. If you leverage AI successfully, five to 10 years, talk to me about the dealer experience and the consumer experience of carbine. What does that look like?
We see, we, from day one, we've used ML and AI in our company. If you think about AI, it's kind of like a, it's like a three, let's put a three phases, something, you know, just kind of anyone listening to this who's in the world of AI will not agree with me, but just for our own sake here. Let's put it into three phases. Okay, there was the base of like no one knows what's going on. It's under the under the surface. There's like researchers and universities doing all sorts of weird things with like AI and language models and neural networks. And you know, it's out there, but none of us know about it.
That's, you know, already dating back to the to the 80s, right? Maybe even before. Then there was the phase where we as consumers started to experience it. So think about like the neural network stuff that Google did, that enabled them to do image matching and to give us better results in Siri and in Google's assistance and Alexa. And then I think past two years with the generative AI really passed three months, six months. It's just been totally transformative in terms of how it's consumer's experience.
So it's been really like three phases. One where it was just like hidden, the second we went into applications that we just used as applications. And now we're we're really seeing its power very much. So we think that the dealers, the auto industry is ultimately going to have an operating system that is fully AI based, meaning where the dealer is basically using the AI to run all of its systems.
So I'll just give you an example. Yeah. And this is stuff that I think is quasi happening. If I'm a car dealership and I'm a marketer there, I would expect my AI agent to show up for me in the morning and say, Hey, we just, you know, we noticed like three or four things last night. This competitor sold these cars. The price on leads is dropped here. We want to take $4,000 and dedicate it to this or we want to launch these 15 new email campaigns. Do we have your your approval?
And you can then just say, sure, yeah, yes, launch, don't launch, snooze, delegate, pass it over to my owner to approve it. Whenever it is, the AI will be able to do that. And that's not just generative AI. It's not just that there's a nice chat interface. It's that all the creative already, at least in our case, is is already being developed in part by AI backed a generator. So you can really get away from having to do the drudge work. It'd be more of an executive function. And I think that's what's going to happen.
AI's are going to run the dealership as a kind of operator while the marketing directors and the owners and the GMs are executives in that role telling them yes, no snooze, delegate, fix, adjusts. That's how it's going to work in five years, less than five years. I think in two years we're going to be there.
Now while that sounds fascinating, it doesn't sound crazy to me, right? It makes total sense. Like yeah, it's, you know, you're getting you're getting rid of all the kind of grunt work in between and all the analysis. It's just, you know, streamlining that. So it makes total sense. What about from a consumer perspective?
I go on a dealer's website, is it like the chat agent before and we can jump into the chat GBT in a second. But like, well, how does that experience change? I don't even think it's the chat agent. I think that's the, these early phases. I think that consumers are right now. It's not.
By the way, I agree with you. I agree with you. I think that's the low hanging fruit here. Oh, it's the lowest hanging fruit. It's just, it's just easy. Like you can get it done. Again, it's technologically right now still not simple as we've discovered, but it's, it's, it's, it's not what's coming.
And I think what's going to come is again, I really ascribe to the agent approach where I think shoppers right now, I mean, to find out what card to buy, to get information, it's really not easy on Google. It's just, there's so much SEO that has been done over the years and there's so much, so many ways in to, um, to just fall down a rabbit hole and not get the information you need.
I think that, um, the ability of AI to synthesize and summarize, right? Again, we all know it can also hallucinate, right? It makes stuff up, but the ability to synthesize and summarize is just going to make it easy for shoppers to, to like, get the answers they want really, really quickly. And then the agents will be able to do searches for cars.
Like you will have agents that will find cars, bring your vehicles, may even negotiate with the dealers agent, ultimately, you know, settle on a price.
I mean, I, I think it's going to go to that point where a lot of the work of the shopping process will be handled and done by the agents. And then you'll just be able to, you know, do the end, uh, as a, as a consumer or do what you want to do. So I think that's, it's just going to get easier to find a research. It's getting it, it's going to be easier to understand what you need. Is a, is a shopper and that it's going to get easier to find the actual vehicles that are available on the lot.
Um, and if you think about where things are going, like, there's no reason why the agents of the dealership don't know what inventory they have. Don't, aren't able to answer a question about a payment and whether it would work for our first and like all the stuff that takes a long time and phone calls and conversations and visits that, that really is accessible to, uh, to these AI agents and we'll be solved by them. Uh, and we think that's going to happen. Not in the next five years. We think that's going to happen much sooner.
What are you building or what did, what have you built for dealers using Chad GPT? Like what's the current tool that you're offering on the market?
你正在使用Chad GPT为经销商建造什么?你已经建造了什么?你目前在市场上提供的工具是什么?
We have so many things we've built internally. We haven't even released. So we're, uh, we're doing a lot of stuff. First of all, for our own stuff, we now summarize massive amounts of internal information at our company for our, uh, team members and managers. So like if, if a, you know, if a call happens in our company with a dealer and there's extra information in CRM or whatnot, it just gets synthesized with next actions with a grade of how much, uh, how many challenges there were that we have to follow up on and it's being used for our own internal business process really effectively now.
So what we've released for dealers is, uh, is one thing. Uh, we've released a, uh, chatbot. Okay. What I like to think of is like a shopper assist on the dealership website. Um, that chatbot has all the generative like conversational powers of what you experience when you use Chad GPT 3.5 or 4. But behind the scenes, we've done one thing.
我们为经销商发布的是一件事情,那就是我们发布了一个聊天机器人。这个聊天机器人是一个类似购物助手的存在在经销商网站上。它具有和 Chad GPT 3.5 或 4 一样的生成对话能力。但在幕后,我们只做了一件事情。
Um, and that is, is we've taken the dealers own inventory and data, right? What sits in what we call our CDP layer, right, or customer data platform layer and our car data platform layer, uh, and we've made that accessible and available to the chatbot in very specific ways. So for example, if I'm shopping on the website as a consumer and I, I'm looking at bake corvette, uh, and let's say I come back the next day, the chatbot now knows that I'm interested in Corvettes and it's priming behind the scenes the dealers Corvette inventory at the dealers Corvette specials, right? To be ready to show to me, should I want it?
My own specific dealership data, my own specific dealership specials, and the shopper can now ask a question like, Hey, how does this compare to the, whatever, to a different vehicle? You know, how does, uh, you know, how does the, uh, uh, Corvette compare to something I'm looking at it even a different OEM and and it will have information and answers. But then it might say, Hey, here's the three Corvettes we have or hey, are you interested in knowing what specials were running in it? It can actually answer that with information.
Uh, and that's the information we're providing. It can also know that that shopper, for example, was looking not just at Corvettes, but at another vehicle was also considering a truck. Maybe there's like two things going on there and it knows that and it can then maybe suggest an, uh, an opportunity on, on a truck as well. So that level of interaction is much more natural and fully fledged than basically just what happens today, which is like, Hey, here's some cars, pictures of cars, maybe it's the right cars, not the right cars. Do we get your name and information so someone can call you? Right? This is an actual informative shopper copilot, uh, that will give them a better experience and answer questions.
Okay. So it's not like the traditional chatbots that I've, like I, I always refuse to use those at our dealership because I felt like the experience was just so subpar and I hate them personally myself. But what you're saying is like, look, the biggest difference here is that number one, we've all played with chat GPT and we know that, you know, it's capabilities. But the difference here is that it's, you're saying it's connected to that, it is specific dealerships inventory and specials and so it's able to tailor that specifically for you, your ones and needs as a consumer. And it knows, um, quite a bit of information about what you're more likely to, uh, want because it under, I'm trying to explain in a way that isn't too overly techy. It doesn't really know anything about you. But as you browse, we tell it, I mean, we tell it for this here at Bard or chat GPT, whatever we're using in the background, right? We tell it, hey, here's this person, you know, is, is Billy and Billy is interested in the following three vehicles and also chairs about financing, uh, it cares a lot about financing and specials. Now have a conversation with Billy, go.
Okay. And, and, and it will forget that the next session unless we, we remind it, right? Because it's, it's what they call stateless, right? So we have to, like, these, what's interesting about these GPT bots, they have a lot of, um, they have a lot of complexity and like, attention, like figure it as attention span, right? So, you know, the way they, they make conversations, they have lots of words. And they just, yes, they don't guess. They calculate what's the next best word to use or next, that's even letter to use. And somehow it comes out right. No one really seems to know why. But, you know, the question is how far back can they look? So, uh, a lot of these, you know, these AI models can only look back a certain amount of time. So if the conversation gets too long, they may lose context. You have to give that, you have to give the chatbot in that moment. The context about the person and we do that. And then we tell it what to do.
Uh, and so at any even moment, that's what happens there. Now, there's a lot of, uh, challenges with that on the technology side. Everyone's like, oh, it's just a, it's just some API. Oh, no, no, no, it is not just some API. You have to deal with real, uh, complexities on how you structure the data that you want the, the chatbot to engage with is you have to deal with what did you need to pull out or because in the end of the day, the dealer needs to lead. Uh, you have to deal with the fact that if you're using just the latest release, uh, uh, four, uh, GPT-4, that is an extremely expensive, extremely slow API. Yet if you use the earlier ones like 3.5 turbo, some of these other ones, they're not that, they don't catch maybe multi-threads, right? And so they'll miss things. And so it's very, very complex to actually get a right in tune it to the experience on a, on a dealership website. We've had some great dealer partners that have been pioneers here and in turning it on to their website and it's good that every day are tuning.
Uh, yeah, but that's what we're doing. That's our first foray. Why are we doing it? I don't know. We always wanted to do it. We, we used to use something called Google Dial-Off Flows, which was a Google thing. And we just seemed like better at this stage. So we just were like, let's just do it. Why not?
嗯,是的,但这就是我们正在做的事情。这是我们的第一次尝试。为什么要这么做?我不知道。我们一直想做这件事。我们过去使用过一个名为 Google Dial-Off Flows 的东西,那是谷歌的一个东西。但是在现阶段,我们似乎更擅长这个。所以我们就决定,为什么不试试呢?
You know, uh, here's my crazy idea, right? So it's actually not that crazy, but, but bear with me. Yeah, go for it. So I spoke with one of the founders of one of the largest DMS companies. And I'm sure you could put two two together who it is, up and coming already put it together. Yeah. And what they told me was that, and I didn't know this, but they told me that when they started, they actually acquired a dealer group in order to test their product, which I loved. But, but it gave me this idea, right?
I was thinking about you. I said, imagine if you launched a dealership or you acquired a dealership, you know, it could be a small dealership. And you just completely like overhauled their website. Like that was your testing routes. I'm sure you have, you know, test demo websites, obviously, but I'm thinking like in five years, you know, whatever, whatever time span, like the way I envision carbying using AI is that I go to a website.
And I don't see anything. Yeah. Maybe there's some in preview of inventory, but it's just a, it's just like a text field. And it's like, what do you want? Yep. Let me serve you. And it's like, hey, I want a 2020, you know, Jeep Grand Cherokee under 40,000 miles. It has to have, you know, this color seats. I needed to deliver to my house within two weeks. I can't pay more than $400 a month. And I'm willing to put this much down. And it's like, data. Data. Data. Okay. Yeah. Use what we have for you.
我什么也没看到。也许在预览库存中有些东西,但那只是一个文本字段。就像是你想要什么一样。让我为您服务。如果您想要一辆2020年的Jeep Grand Cherokee,里程数在40,000英里以下,需要这个颜色的座椅,并且需要在两周内将其送到我家,我每月最多只能支付400美元,并且我愿意支付这么多。数据。数据。数据。好的,让我们使用我们为您准备好的东西。
Well, are you ready to complete the purchase? Just pops up like an apple pay, like that's the type of future that I want to live it. I'll tell you that. I agree with you. I mean, I think it'll be a little bit more suggestions along the way because people do want to look at multiple things, but 100%. We've built something like that internally, like we're playing with that stuff. You know, not to give our road map away too much. You're kissing out our road map away. Give it away. Yeah, so yeah, we're playing with that.
你准备好完成购买了吗?就像 Apple Pay 一样弹出来,那是我想要的未来。我完全同意你的观点。不过,我认为在购物过程中人们仍然希望有更多的建议和选择,但是百分之百的是我们内部已经建立了类似的系统,我们正在尝试和改进。不过,我们不能透露我们的路线图太多,不然会暴露太多细节。是的,我们正在尝试这些方面。
I'll tell you other crazy things, like one of our developers for fun. We collect all the data on every car in the United States every day, right? So we crawl the web, we get every car, we get the price changes. You know, we want to have all that data because we use it for our algorithms. He took, he took an OEM. He took all their vehicles and he put it all in a DB, the types of DBs you need for Chatchy PT and basically let me go as a shopper and just do exactly what you described.
And it was unreal. I was like, hey, I live in near Cleveland, Ohio and I'm looking for this. And here's what I could pay. And it was like, boom, there it is. Here's the dealerships, here's the cars. And he did that in because we have the infrastructure. He did that in basically 48 hours. And I suddenly had every OEM vehicle available. I figured tips and I could just talk to the bot and have it do whatever I wanted. It was a real agent for me. It was an unbelievable experience as a shopper. And it obviously gave us a lot of ideas for how the OEMs can do some interesting things. So that was definitely the way the things are going to go.
And my thing, similar to yours. Do you have some fringe idea, whether it be related to Chatchy PT AI, really the evolution of car buying and the car business using this technology? But do you have some fringe idea? Is something that you think will create billions in value for the industry? I think that there's going to be a lot of opportunity for buying and selling discovery stuff. Like if I'm an inventory person, I'm going to be having an agent crawling in the US car inventory, not just like dealer inventory.
Like everything out there in what you know, Cox calls the car park, right? It's out there in the world looking for what I want. And there could just be like buying it in auto transporting it to me using these AI agents. So I just think the AI agents are going to do a lot of that work. In terms of the consumer side, I mean, you know, as I discussed before, I think a lot of consumer shopping is going to happen through the agents and they'll do all the research and work and bring you cars. But you know that I don't know. I don't know where it's going to go. It's going to be very interesting.
How has that adoption been of your Chatchy PT tool? Foof insane. Insane. We are going kind of slow right now. Though that should change next week, just because we're like really trying to get it right to your same dealers have appetite for this. Oh my God. Yes. Yes. Yes. They're signing up. They're interested. You see it or like, oh, that's a bridge too far for me. Or, hey, wait, let's, let me see how it goes a little longer. But many of them are just like, oh, yeah, I love that shopping experience because you know what, you can get any chatbot on the internet.
Be it a human backed one, be it a, to do stupid stuff. If you want it to, I mean, you could do whatever you want. You can make them do stupid stuff. You can take a screenshot of that and be like, oh, look how stupid that chatbot is. But you know what, the average customer of which we see tens of thousands of customer conversations on a regular basis, you know, for order or whatever, they're just like they're asking normal questions. They're getting normal answers from our chatbot.
You know, and here we don't have tens of thousands yet, but we're, you can imagine what we're doing in the background, which we're taking chats that we have from our pre-GPT era. And we're running them through to see what happens in order to model how it works. So anyways, that's like one of the techniques that we've been using to get this right. And it's just really powerful to see that for like regular shoppers, it's doing a really good job, it's just a much more consumer-friendly experience. And it's still getting leads for the dealers.
Meaning dealer conversations are high, excuse me, dealer chatbot to shopper conversations are high and conversions to leads are doing really well. So we have right now about under 100 dealers using it, but that should go up really quickly in the next few weeks. I mean, it's our, we've been limiting supply just to get it right.
Yeah, I think that's fair. Look, I think I've also, you mentioned just a doctor has been high, makes sense. I've, I've had conversations with dealers and, you know, like their managers and stores and consumers beforehand are using chatGPT now for like, consumers are using it as an example for, you know, scripts for emailing the dealers. Oh, that's LULIT. Yeah, dealers are using it also for various things, maybe responses or whatnot, but I haven't seen, you know, some technology like this, just get adopted.
Buy, you know, everyday people so well and so quickly that it's crazy to think what it's going to be like to buy a car, sell a car within five years. Just that entire experience, it sounds to me like it's going to be a very streamlined business. I don't necessarily think that dealers will know more about you than they do today. And I could be wrong. That's my personal opinion. But I just think that that entire, all that friction in the middle will be eliminated. Yeah.
It will create a much better car buying experience. And frankly, I think it's going to level the playing field a bit between, you know, the Carvana car maxes and smaller dealers because if you have this technology layer that's commoditized and anyone can license it and you're paying on a, you know, per usage basis, I don't see why car max or whoever can have this light years better technology, you know, unless they significantly build on top of that. But it just feels like in a way, companies are like full path, like what you're building on in others, sort of democratizing that for every day dealer, which is, it sounds to me like you could even be a tailwind for the small business in the industry, especially at a time when there's a lot of headwinds and we're seeing all this consolidation happen. Yeah.
Now, I think what's going to happen is that either will be less noise, as you said, just less noise in the research, less noise in the back and forth. And in the end, what will happen is that I think the agents, the shopper agents will basically just tell a person, hey, those show up here at this time and you're going to test drive the two vehicles you want to test drive. And then it'll tell them the next to go show it's this other dealership nearby. You'll test drive those two and then you will tell it, hey, I want this.
It will go negotiate that price with the dealer agent incredible with the oversight of the people at every point. And then you'll have a car. Now, I think that would be who the OEMs in general to think about, like, just kind of simplifying the features of their cars and making it like easier to understand the packages. And that that'll help the whole process ultimately be simpler and easier. And we have a way to go on that. It's still pretty complicated to understand what you're getting in a car you're buying.
As opposed to say, like a Tesla where it's a little bit, it was designed in a year of more simple packages. I think the mainstream OEMs are still going in that direction and that, by the way, will make things easier for everybody in the process. Our own. This was very insightful. Appreciate you coming on. You have the audience wants to learn more about full baths yourself. Where can they go to learn more? Yes.
Go to fullbath.com and check us out. And you can definitely, if you are a dealer and you're interested in getting on our wait list for the GBT, we'd love to get it in front of you. So there's a link there on our website. But if they mentioned car dealership guy, we're going to let them cut the line, right? We'll pop it up. Absolutely.
I love it. I love it. Hey, thank you. This was awesome. I had a good time. This was awesome. And you know, it's going to be interesting. I'll ping you. First of all, I'm sure we'll be in touch very frequently over the coming months and weeks.
But more importantly, in the future, I'm going to be very curious to see when we look back at this episode, you know, a year, two years. And we say, wow, like did that come, you know, did that manifest, you know, did this actually happen, did that not happen? What does this look like? I think that's going to be very interesting to see because things are just happening so, so quickly right now. So, I think it's going to be an exciting time.
Yeah, we see ourselves as like one of the RD labs for the industry. And so we want to push the envelope, but I agree with you. We got to look back and see what actually happens. So I look forward to that. I really appreciate you coming home. Take care. Appreciate you.
Alright, hope you enjoyed that episode. Please give the podcast a rating, consider subscribing to the show and check the show notes for links to what we talked about. Thanks for tuning in. See you guys next time.