March 19, 2026 | Engineer used $25 AI tool to save $4,200 on car; Iran war threatens Japan’s supplies

Are you a dealer creating a workplace culture your employees are proud to be

part of? Applications are now open for the 2026 automotive news best

dealerships to work for program. This isn't just an award. It's a chance to get

real insight into what's working at your dealership and where you can

improve. And we've expanded the categories this year, recognizing everything

from technician experience in leadership development to AI

enablement and employer retention. The registration deadline is April 17th.

Find out more and apply at AutoNews.com Welcome to Daily Drive. For Thursday, March

19th, 2026, I'm Kellan Walker in Las Vegas today on the show. The Iran War

forces Toyota and Nissan to cut production. Uber bets big on Rivians

Robotaxies. And while others ditch planned EV models, Mercedes AMG is

unveiling three new ones. Plus, we're here from a Boston software engineer who

built an AI agent that negotiated a cardio-forming, saving more than $4,000

on a new Hyundai Palisade. What the AI agent did under my command and my

instruction was just email the quote back and forth. Pick whoever I the

lowest one and go to the other person and say they're offering this can you beat

it. Let's run through all the news you need to know to keep up in the auto

industry. Toyota CEO Koji Sato is sounding the alarm about Japan's auto

supply chain. Speaking as chairman of the Japan Automobile Manufactures

Association, Sato said the Iran War is threatening supplies of aluminum and

NAFTA, a key chemical for plastics and rubber. Japan sources about 70% of

both materials from the Middle East. It exports roughly 800,000 vehicles a year

to the region worth about $16 billion. According to the Nike newspaper,

Nissan and Toyota are already trimming production in Japan to account for

possible shipping interruptions. Uber is making a big bet on autonomous vehicles

committing up to $1.25 billion to deploy 10,000 Rivian R2 Robotaxies

exclusively on its app. The ride hailing giant will invest $300 million

up front with additional funding tied to hitting autonomous driving milestones

through 2031. The partnership launches in San Francisco and Miami in 2028 then

expands across 25 cities in the US, Canada and Europe. For Rivian, it's a major

validation for its self-driving ambitions as it prepares to install in-house

autonomy chips in the R2 starting next year. And we've been talking a lot

lately about canceled EVs, but Mercedes AMG is charging forward in a big way.

The brand just unveiled three battery-powered GT models to dealers, a

coupe launching early next year, followed by two crossovers. Mercedes also

revealed a G63 Cabriolay, the first-ever G-Wagon convertible for the US

market, arriving in 2028. It will have a V8 and a powered fabric roof. The

rollout underscores Mercedes' gross strategy with plans to hit 400,000 US

sales annually by decades in. And those are today's headlines, you can find

more details on all those stories at autonews.com. This weekend, the WeatherTech

Sports Car Championship kicks off one of its signature events, the 12 hours of

sea-bring endurance race in Florida. The International Motors Sports Association

oversees the championship, which features brands ranging from Ferrari to

Porsche to Chevrolet to Lexus competing at races across North America. Our own

Jack Wallsworth cut up with IMSA President John Dunin at Daytona International

Speedway, where the season opened with the Rolex 24 and record attendance.

John Dunin, welcome to Daily Drive. Thanks for having me, Jack. It's good to see you

and great to have you back at IMSA WeatherTech Sports Car Championship event,

especially this one. Yeah, I'll find it. Rolex 24. Super cool being here. So, John

we're at the iconic Daytona International Speedway here for the Rolex 24 hours

Daytona. How would you describe the atmosphere this year? There's a lot of

adjectives, electric, emotional, passionate to see what we've seen in terms of

the attendance at the role before the 24 to see all the folks transition right

into the Rolex 24 from the infield to trending ticket sales to be an all-time

record for this event to see the garage area filled with fans. And that's one

of the things we're super proud of is an open, you know, general admission ticket

as a garage pass in a paddock pass. So we're really proud of that. And then the

thing that pushed it over the top for me was the gridwalk yesterday, never in a

million years that I see something like that in all these years in the sport.

And I think the last piece which I learned late last night was the activity on

the midway for the automakers for Michelin and for the other brands that are

activating there. In one day yesterday, I had an automaker report to me that

they had the most fan engagement, which means you provide your name, address,

and email to ask for more information. They had one day had more engagements

than they had the entire Rolex last year. And for me, that's a big

barometer of attendance one, but the enthusiasm of the fans for the 18 automakers

that are racing here. Right, yeah. I was over there yesterday afternoon, and I was

taken back by hell, large to crosswear. Incredible, yeah. So obviously this is a

very long race. How do you watch it? Do you go to your spots? You try to

say it for the whole thing? What's your, what's your game? Yeah, I mean, of

course, the race day morning and all the pre-races of busy time for us. Once the

race gets going, it sells in a little bit. We have such strong attendance from

our distinct partners and other new potential partners that are looking at

MCSA as a marketing platform or potentially a technology development platform.

So a lot of meetings and then spent overnight. I don't go to sleep. I spend

overnight in race control, especially given the fog and the visibility issues we

had. But I like to have my finger on the pulse of the race itself, but we got an

amazing group of professionals here at MCSA that run race control and pit lane

and marketing and partnership and PR and all those things. So stayed up with

them and we had a lot of collaboration up top and race control about the best

things to do to get the racing back going and we had the fog. So inside some

announcements earlier in the weekend, that Friday, that's the partnership with NASA.

Talk a little bit about that. Yeah, I mean, for me, I remember, you know,

sitting in front of the TV or having the TV rolled into my grade school when

rock and launches were going to happen. And for our sectioning body, MCSA, to have

a partnership with NASA because they believe that our environment creates the

strain and the stress and the endurance of product and thermal properties and

lightweight materials that are exhibited here, sensors that are used real time

here to collect data. And they want to come study this environment. And in the

same manner, they've got some really cool stuff. And they want to give us the

chance to do the same. So that kind of collaboration similar to what we have

with the EPA, the DOE SAE over the years. This one takes it to the next level. In

addition, this space over five decades has been a test bed or a laboratory for

automakers, tire companies, lubricant companies. And we felt like the time is

right, given the technological advancements of not just the GTP cars, but even the

GTP cars to allow and provide a formalized platform for tech companies to come

in here. We've just launched our Lenovo AMD Technical Center, our

engineering trailer. And you go in there and you watch what's happening in

there and the amount of raw data that's coming off of those cars through

probably 150 sensors real time. We have a very data rich environment here that we

can leverage to allow software companies who are on our panel like Microsoft, AWS,

NASA, we already talked about. And there's many more that could use that data to

approve out a new piece of software of their own, develop maybe a new product

that isn't necessarily just for racing. It could be for business and show a client how they

could gain a competitive advantage by this. So for me, it's what IMSA needs to be. It's where IMSA

is headed. And we're thrilled. BDO, an existing partner, is going to help us by being that

digital transformation partner and launching that along the way. So, you know, Michelin

sustainability and racing award gets awarded today. And then who knows down the road if

there's a technology award based on learnings from the race track.

That was John Dunin, president of IMSA speaking with Automotive News reporter Jack Wallsworth.

You can hear the rest of Jack's conversation with John on the upcoming bonus episode of Daily

Drive, available Sunday morning, wherever you get your podcast. Coming up next,

how one software engineer use an AI agent to negotiate with dealerships

and saved more than $4,000. That's next on Daily Drive.

InVidia is betting on cars. And this week, I got a closer look at the company's strategy.

On this week's episode of the Automotive News Shift podcast, I'm joined by Ali Kani,

InVidia's vice president of Automotive. We talk inside of a moving,

InVidia powered level two vehicle about the company's automotive goals and how its chips

are managing the complex tasks required for automated driving. The industry is at an inflection point.

Every car will want to be autonomous. Plus, I'll recap InVidia's GTC in San Jose,

including why CEO Jensen Wong says autonomous driving tech is a solved problem.

It's definitely a solved problem. The rest of this is visionary reclaiming.

I'm Molly Boygun. Join me on Shift, available this Sunday, wherever you get your podcasts.

Welcome back to Daily Drive. I'm Kellan Walker. Imagine buying a car without ever stepping foot

in a dealership showroom and having an AI agent do all the negotiating for you.

AJ Stivenberg is a software engineer with DataDog in Boston. When he and his wife decided

they needed a Hyundai Palisade, he built an AI agent using an open source tool called OpenClaw.

Automotive news retail tech reporter Mark Holmer spoke with AJ about how he did it and what it

means for the future of car buying. Tell me to the point then where you decided to build an AI

agent. What made you get there? And then what did you do to build it in layman's terms?

So I had read about Claudebot which has not been renamed to OpenClaw and I was familiar with

the capabilities and I thought it would be kind of useful. I had actually on my own been trying

to email a couple of dealerships and I just found a lot of dealerships don't want to interact

with email. They really want you to come in. There's a lot of friction there and I generally wanted

to cast a very wide net, but I didn't want to go and find all these people to email myself.

And that was when I was like, okay, this is a great use case to try out OpenClaw on.

And for readers just to be clear, this is an open source tool that you can customize.

Yeah, yeah. The codebase itself is fully open so you can read all of it. You can install it on

any machine you want for the most part. I had it running on an old laptop and what it allows

you to do is you connect it to one of these large language model provider with like an API key.

What is an API key for layman? My understanding is it's basically something to connect to other

software. Yeah, it authenticates you. So the idea is you go to an anthropic or you go to

like an open AI's website, you give them your credit card information and then if you want to

use their platform, not necessarily through the browser, but through like an AI tool like OpenClaw,

they give you a secret. It's like a password. They give you a long string of letters and numbers

that you can kind of paste in. And then it links together. Correct. And that's what you did.

Yep. Did you worth this or customize it for yourself?

Sort of. You have to enable it. And what I mean by that is you need to give it access to the

tools that you would use to take these actions. So for me, that was basically my messaging

application. So like I message on an Apple phone along with like Gmail. And then I also gave it

access to a web browser. So it's able to go to various card dealership websites and find

like the contact forum and ask them about, you know, the Hyundai Palisade that we were looking for.

So when did you create this tool? When were you ready to start using it?

I mean, installing it and getting ready to use it only took a few hours. I've a fair amount of

experience. I've been in the industry for a while. So there it's not an easy thing to just get

set up on your own. And you sort of have to have a little bit of a development background to be

able to get started. But with that, it only took me a couple of hours before I was able to kind of

set it off on its adventure. It was this in the summertime still or the fall? No, this was like

a few weeks ago. Yeah, yeah. So it would have been mid January or late January.

So then once you had it, how did you use it to search for a car?

So I gave it a prompt. There is a couple of Reddit is a website that I'm sure your readers are

familiar with. And there are active subreddits or forums for many vehicles. One of them was the

Hyundai Palisade. So the first thing I asked for was find me what a decent or fair and then a good

and a great price is for this car that I want in my area in Massachusetts. Like what are people

saying? What are people paying? Because the price can vary kind of all over the country. So

I did that. And it was able to go on Reddit and kind of look through the topics in the posts.

People have been sharing their quotes that they have received from dealers. And they're asking

the same question. Anyone would ask was like, am I getting fleeced? Like am I getting ripped off

here? And the AI agent is able to kind of go through and read those pull them together. And then

it kind of kicked out a spreadsheet for me. And that gave me like a target number. That gave me

an idea of what I wanted to work with. Now, so you saying it used Reddit postings? Did it use it?

Yeah. And I told it to you. You told it to. Did you tell it to use any vehicle listing sites?

No. For the most part, I had used the tool called visor.vin. And I had already sort of

scoped out what the prices are. If you haven't used visor.vin, it's just another aggregation site

like the ones you had said earlier. So I had a rough idea of what dealers were listing on their websites.

But there's a difference between what a dealer lists on their website and what people are

paying out the door at the dealership. And that's what I was trying to get at. So, yeah.

So David, did it give you a list? Did you have a top five? Did it narrow to one place?

No, it actually didn't even narrow down by dealerships yet. So that's the second part of the story.

There are some other open source developer made a tool that allows you to like track Hyundai

inventory. This I think is very specific to Hyundai. I don't know if other dealerships have it.

But at this point, my wife had picked out a color combination she liked for interior

and exterior and a trim she wanted. So I had a very specific search criteria.

And this was that you already had the model in mind then.

Yeah. At this point, I had picked out, I knew we wanted the Hyundai policy. I knew we wanted

the calligraphy trim. And I knew we wanted it in like a blue exterior with a brown interior.

That was what my wife and I agreed on. So at that point, it was a matter of finding inventory.

And there were a couple of online tools to do that. And the best one that I found, I think I

linked to the blog post. I don't remember it off the top of my head. It simply is like a very simple

Excel spreadsheet style list. And it gives you a couple of filters to pick. And then you're able

to select, you know, the color and the trim and all these things and give you like a zip code radius.

So that's an easy tool for an AI agent to use. So I handed it off to the AI agent to open

claw. And I said, find me all of the the vins than the colors in this color combination within

50 miles of the Boston area. And it went out and just did that and it pulled back a list. And then

I said, okay, now that you have this list, go to those dealerships and contact them and ask them

with the lowest out the door prices. And that's where the magic of the AI agent really took over.

Because for me, that would have been, I would have had to go to each dealer's website, find the

vin, you know, paste the vin in, click contact me, like, you know, ask them about this vehicle,

fill out the form, and then, you know, wait for the email. And the robot is able to, like,

AI agent is able to use the browser, go through and do that all on its own.

And how long did that process take then to narrow, you know, when you said, when you said the

parameters, find me this model with this combination within 50 miles, how long did it take to come

up with something and how many options came up? It took maybe an hour and a half, maybe

somewhere in there. Okay. Not very long. And it had emailed a number of dealerships. I think at first,

there was like five that had the exact color that I wanted. And then one of them, it turned out,

they didn't have it yet. It was in shipping or in transit. And they were expecting it in like a

month, which, you know, kind of was like, okay, nice, but maybe not. And there were a couple of

other dealerships, you know, either didn't respond or weren't interested in negotiating. And,

you know, eventually what happened was there was, there was two left that were interested. And

what the AI agent did under my command and my instruction was just emailed the quote back and forth,

pick whoever had the lowest one and go to the other person and say, they're offering this,

can you beat it? And it basically did that two or three times. And then one person came back and

said, if you commit tonight, I'll give you $4,200 off. And then the other person didn't reply.

So I just said, yes. And that's how you did it. Yep. You know that there's at least one start

about there that is doing this, but there are fees involved. And you did it on your own time. And

I'm guessing for very little of any money invested. Yeah. It cost about 25 bucks in API tokens

when I added it up. So it wasn't that spending. I'm aware that they're brokers that do this.

Then I'm aware that there's, you know, I'm sure that there's there's many software startups

that are trying to do this too. They very much are. And it's fascinating to me that you've even bypass

them and did it on your own. So I mean, that's unusual to say the least. Yeah. It's a rapidly

changing time in software, right? Like by trade, I'm also a software developer. And you know,

the number one news article you see out there now is like, AI is going to eliminate all software

engineering jobs. So there's some consternation in my industry as well about what this kind of

technology means and what it can bring. What's the message here? I mean, in terms of what we

look at, what I mean is our vehicleistions companies or even startups that come up with the

technology to do it. You just did from monthly fees. You know, what's the message? Are these

becoming irrelevant? Or is there room for these companies in the future?

I think there's incredible room for these types of companies. I don't, you know, what I did was

I strung together a number of tools to kind of make this work. It's certainly not like I couldn't

hand this over to my dad today and say go, you know, go buy a car with this tool that that would

not be feasible. It's certainly 100 times easier than it was three years ago. I maybe a thousand

times easier than it was two or three years ago. So with that in mind, you know, it's definitely

more difficult to differentiate. But a business that's starting today that if you are going to

negotiate, you know, car deals, you have a lot of options that I don't have, right? I'm buying one

car. I'm buying, you know, one vehicle, maybe for the next five or six years.

So if you're going to do a startup that does this, I'm sure there's options for you to negotiate

in volume or have some ability to be a repeat business where you're more valuable to a dealership

or a network of dealers than I would be. But you have, and I appreciate that you see room for

these companies, you know, as they adapt to the technology, but you buy past all of them.

I did. Yeah. There's something I've heard that I don't like the term, but vibe coding.

Yeah. Oh, yeah. Is that what this could become? That it would be easy enough to use a tool to do

what you just did, even if you don't know the programming. Yeah. This didn't require very much

actual programming on my part. It required a little bit of research, but I wouldn't say it

required. Like I didn't write any line of code for this. And even if I did, the AI agent itself

can write the code now. And a lot of my day to day now is instructing AI agents to write code

that a year ago or two years ago, I would have had to write by hand. So that is changing really

quickly. But I don't know if I'd say that there's, you know, like no room for innovation here,

because you still, like I said, I didn't use this AI agent at all for any of the vehicle discovery

process or finding what's out there or comparing, you know, two types of vehicles or any of that

kind of stuff. I sort of tried, I don't think there was a ton of value there. I like, I think I'd

asked it to compare the policy to the, to the Toyota Highlander. And these AI models, they don't

have a lot of strong opinions. You can very easily push it one way or the other. And it reflects

back what you would you tell it. So that's why I like hearing from creators. And that's why I like

hearing, you know, like watching a YouTube video from somebody who had the car for three weeks and

drove it around and said, this thing, I don't really like this as much. I think one car dealership

podcast did say they, like they, they kind of criticized the hybrid drivetrain. They're like,

it's a little jumpy, you know, when it switches from electric to gas, it's like a little, you know,

and those are the types of things that the AI model would never say. They're, they're, they're very

agreeable. They're very personable. So I don't think it, it removes that. They're like impressionable

children. It sounds like at least on the stage. And that's changing really quickly. Yeah. And

since, since even this effort, like this, this thing came out that I've worked on, I think both

AI labs have like open AI anthropic are kind of the emergent foundational frontier AI labs today

have released new models. Oh, really?

changes. Yeah. This is changing every month, including like what capabilities are out there.

So if I were trying to work in the space, if I was trying to be an automated, you know,

an automotive sales or an automotive technology, I would just try and keep up with the developments,

don't form too many strong opinions. The AI agents, we're not capable of doing this six months

ago. This would be a laughing matter on social media or whatever they are. So you, and I realize

you're a software programmer and you write code and all of our, all that other stuff. But

do you see a point where with agents being able to code themselves, are we reaching a point

where anybody could do what you did and create an agent on their own to sell to buy a car?

Yeah. I think to some extent we are. Yes. I say that by, I want to like, couch that a little bit.

I do think we're approaching the point where anyone can build an AI agent that would negotiate

for a car better than that person would be able to do. But I also think there is still a lot of room

between that and what you could do with more information information in any kind of negotiation or

any kind of market is the more information you have, the better. And when you start, if you

were going to maybe build an app that did this, you have information about what customers are

looking for and what part of the country, right? What are people interested in what kind of vehicles,

you know, kind of demographic are the people who are looking for these cars in. And you're able to

do that and take information asymmetry and use it to your advantage in a way that someone who's

like writing an AI agent wouldn't be able to do if they were just doing it for themselves.

But it does certainly lower the bar such that if you are relying on the fact that nobody could do

this before and you have the technology, now I don't think that's the case anymore. So you have

to, you have to, you have to be better. That was AJ Stivenberg speaking with automotive news retail

tech reporter Mark Holmer. You can read Mark's full story about AJ's car buying experience at autonews.com.

That's daily drive for today. I'm Kellan Walker. Thanks to our own Jack Wallsworth, Hans Grimo,

and Irvash Kakariya for their reporting for today's podcast. You can get the latest news on

digital retail tools, supply chain disruptions, and everything happening in the auto industry at

autonews.com. Come back tomorrow for a conversation with Micah Morton, service manager at Healthman

Fiat, Alpha Romail, and Maserati of Sugarland about recruiting younger technicians and why dealerships

need to step up their service marketing game. We did direct mail emails, texts, phone calls,

we have a BDC that, you know, it does a really good job in contacting customers and reaching out

to customers. We'd love to hear from you. Let us know what you think of the show and the topics

we covered today. Send us an email at dailydriveatautonews.com or leave us a voice mail at 313-444-2774.

And if you enjoyed the podcast, remember to like, leave a review, and subscribe so you never

miss an episode.