EP 355: How AI is Transforming Auto Insurance and Road Safety

This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life.

How has auto insurance seemingly not changed in, like, forever? I mean, I've been driving for almost a quarter of a century and, and it seems like my auto insurance is the exact same as it's always been. It's, it's antiquated, it's slow, and it seems like the last thing that's involved is any type of technology or artificial intelligence. But I think that's changing and I think that this is one of those industries that is actually ripe for disruption in a good way. So we're going to be talking about that today on Everyday AI and how AI is actually transforming the auto industry and road safety. And I'm going to have a guest and I'm excited for today's conversation with the founder and CEO, CEO of roadsend. But before we get into it, I just have to, I just have to start out here. So if you don't know, Microsoft Work Lab is partnering with us for the Everyday AI Podcast. So let me just answer this question. Why should you listen to the Worklab podcast from Microsoft? Well, because it's made for leaders who know they must adapt to stay ahead. Worklab is the place to find real world lessons and actionable insights to prepare you for the next phase of AI at work. That's W O R K L A B. No spaces available wherever you get your podcasts. All right, speaking of podcasts, you're listening to Everyday AI. My name is Jordan. I'm the host and this is your daily livestream podcast and free daily newsletter helping us all learn and leverage generative AI. So, like I said, I'm excited to get into today's topic. Make sure if you haven't already, go to your everyday AI dot com. If you're listening on the podcast, check your show notes. It's all in there, as well as related episodes. Before we get into our topic for today, let's Talk about the AI news for today, September 11th. Yeah, don't worry about that. Live stream graphic. All right, so here's what's going on. A lot. So Mistral has launched Pixtral 12B, a new multimodal AI model for image and text processing. So Mistral, a French AI startup, has made headlines with the release of Pixtral 12B, a new multimodal model that processes both images and texts. This development reflects the growing trend in the AI industry toward multimodal capabilities. Capabilities which could significantly enhance various applications. So Pixar 12B is a 12 billion parameter multimodal model. It is available right now for download on GitHub and Hugging Face and the model is expected to perform tasks such as image captioning and object counting similar to other multimodal multimodal models like Anthropics, Claude family and GPT4. Oh, so currently there's no available web Demos to test Pixtral 12B, but Mistral plans to make it available through its chatbots and API platforms LA Chat and LA Platform in the near future. All right, next yeah, we got some some more celebrity AI news. Taylor Swift has publicly endorsed Kamala Harris amid concerns over AI misinformation Yes, this is an AI story. So Taylor Swift's recent announcement to support a Vice President Kamala Harris in the upcoming presidential election highlights the growing concerns surrounding AI generated misinform. So Swift revealed last night, not even 12 hours ago, her endorsement in an Instagram post stating that AI generated images falsely depicting her support for Donald Trump prompted her to clarify her voting intentions. She expressed fear about the dangers of AI and misinformation, emphasizing the need for transparency and political endorsements. So the incident Swift reference occurred in late August when Donald Trump shared AI generated images of Swift, including one that falsely claimed she wanted people to vote for Trump. All right, last but not least, OpenAI is set to launch its reasoning focused AI model strawberry in two weeks. So according to reporting from the Information, OpenAI is gearing up to release its latest AI model Strawberry, which promises to enhance reasoning capabilities beyond the current offerings. So this development is significant as it indicates a shift toward more advanced AI that can tackle complex problems effectively. So the new model is codenamed Strawberry. Previously it was codenamed Q Star and it will be a part reportedly be a part according to the information of the ChatGPT service, but will differ from existing conversational AI by incorporating a thinking phase before responding, with responses reportedly taking up to 10 to 20 seconds. And unlike current models, Strawberry reportedly will only be able to initially process text and not image responses. So reports also suggest that Strawberry will have the capability to autonomously scan the Internet and conduct in depth research, which could enable it to address more complex real world challenges. So yeah, we don't know if this is officially going to be, you know, called GPT 4.5 or if it's just going to be a new mode in the current 4.0 model. All right, there's going to be a lot more. So if you haven't already, please go to your everydayai.com Sign up for the free daily newsletter. We're going to be recapping all of that AI news and a whole lot more. But today we are here to talk about how AI is transforming the auto insurance and road safety industries. I'm excited for this conversation. We haven't talked about this in 350 plus episodes of everyday AI so that's why I'm excited to welcome on today's guests. Let's go ahead and welcome. There we go. There we go. Rohan Malhotra, the founder and CEO of roadzen. Rohan, thank you so much for joining the Everyday AI show.

Thanks a lot for having me on, Jordan.

All right. Hey, I'm excited for this one. So before we dive in, Rohan, can you tell everyone a little bit about Roadzen and what it is you all do?

Roadzen is using AI to transform the auto insurance industry. Every year, $800 billion is spent on auto insurance premiums. There are 1.5 billion cars on the road. And we are using AI to make underwriting simple, to drive premiums down for everybody, to make claims faster and to make driving safer on the roads. We think those are the three things that matter most to consumers. And we're one of the leading technology companies in this space where NASDAQ listed and we are growing very fast at this stage.

Yeah, and it's always exciting, you know, for me when I have CEOs of large public companies on talking about AI because I think it's people like yourselves, like yourself, Rohan, that are really helping push this conversation forward. So, you know, maybe first let's talk about what's wrong with the auto insurance industry. So I kind of started my show by talking about like, hey, even myself. I've been, you know, driving for almost a quarter century and it seems like auto insurance hasn't really changed that much. Is that the truth is just this, just one of the most antiquated industries there is.

I think insurance in general is a super antiquated industry. If you look at the last 40 or 50 years, insurance works pretty much the same way. You can buy a policy, you buy it through agents or brokers. If you have to file a claim, it'll take weeks for the claim to be processed. Even if you make improvements in your behavior, your rates don't really change. And one of the most pressing problems in the US Today is auto insurance is the single largest contributor to inflation. Auto insurance rates have gone up on average 20% since last year. And the rates have continued going up, especially since the pandemic so, you know, what we're seeing is this is a legacy industry that's not adapting to the changes that are happening in technology. And Roads Inn is one of the companies that's looking to bring better technology to the space.

And I'm interested in that one, Rohan, because, you know, you said since the pandemic and you know, recently auto insurance rates have gone up 20% and I'm scratching my head and I'm like, why? Like, it seems like between work from home, you know, hybrid work scenarios, aren't people driving less? Shouldn't those, you know, rates in theory be going down with fewer cars on the road, you know, presumably less traffic? Why are rates going up when presumably people are driving less?

There are a couple of factors driving this increase. The first one is the overall cost of repairability of cars has gone up. As cars have more electronics and software. When you have a claim, it's just more expensive to repair. Labor costs also have gone up. There are two parts inside a car when you have a claim. There's labor and there's parts. Labor costs have gone up 50, 60% since the pandemic because nobody's training to become an auto mechanic anymore. Right. So the supply is limited. Accident rates have also gone up since the pandemic. So people, the severity of accidents, people are driving faster. They are, there are more unsafe drivers on the road and that is driving rates up for almost everybody. Now what happens generally in the auto insurance world, or insurance in general, is if there is a change happening today, you won't see your rate increase. Today the insurer has to go to the regulator file for new rates and you see the rate increase happen like a year later. So we are seeing kind of this offset play out where inflation started becoming high, supply became limited, more accidents on the road. And that essentially is getting into play today. And you're seeing the rate increases in auto insurance.

You know what, I, I talked to a leader a couple of weeks ago in logistics and I was, I was shocked to find out how old school, right. Some of this logistics companies, how they operate a lot of, you know, still pen and paper. Is that how kind of the auto insurance agents, you know, the auto insurance industry is, you know, not literally saying they're, they're pen and paper, but do they still have a lot of these old school manual processes in place? And, you know, if so, why is this industry maybe so not quick to adapt to technology even from, you know, a decade or two ago?

It's a super antiquated industry. What happens is the insurers come up with the insurance underwriting model, which is really how much premium should someone pay? They do this once every three or four years. And no matter what happens now, they're going to stick to that model. The world moves in real time, but insurers move once in, let's say, three or four years. Similarly, on the claim side, like all consumers care about is if I have a claim, is that going to be a good experience for me? Like, can I get that claim resolved fast? But in the world of insurance, it takes about six weeks to process a claim. Like you, let's say you get into an accident, you have to drive your car into a garage. The insurance company will depute someone to come and look at the vehicle. This may take three to five days. The guy comes in, they do a report on the vehicle, they submit it to the insurer. That takes another three to five days. Then the insurer has to approve the report. So now 15 days have passed and no, nothing, real change has happened. Your car is just standing there, right? Then the repairer begins the repair, the garage begins the repair. It takes, let's say a week or two weeks. So now and then before payment, it takes another two weeks for insurance company to approve and for you to drive the vehicle out of the garage. Now this entire process is just painful and we think there's a better way of doing it because the technology exists today to be able to solve all of these problems using AI, using computer vision, using real time models that can interact with the customer and actually provide decisions on some of this data that we are seeing.

Yeah, and I'm very excited to jump into that side. Right, the computer vision, the AI side. But hey, quick as a reminder for our audience here, thanks for joining us, Fred and Gordon and Daniel, Michael, Marie, Jay, everyone else, if you're, if you have a question for Rohan, please get it in now. So you know what you were just talking about, there is kind of this, the technology is there, right? Computer vision even is, is nothing new. AI is nothing new. Generative AI now is even nothing new. But how can AI start to modernize this auto insurance industry and, and maybe make it better for everyone? How can that, how does that process maybe potentially play out in the long run? Rohan?

So I think as an insurer, you're looking at four different things. First thing is how do I price a policy? The first and the most important thing is to be able to price a policy with precision. And what happens today is, let's say Jordan and I are the same age group, live in the same zip code, drive the same car and a similar credit score. Insurance rates will be exactly the same. But I could be a 10x worse driver than you, right? Getting into accidents. I'm driving more on highways than in city limits. So how you drive, where you drive, when you drive, do you accelerate, do you corner properly, do you brake properly, do you brake very close to other vehicles? There's real time data that should govern it, but they are not taking into account any of that. What happens with using AI is you can actually lower the rates for good drivers and coach bad drivers to become good drivers. And that is kind of one of the main areas of focus. The second area which is super important is on the claim side. So today, as I just explained, it takes about six weeks to process a claim. However, we believe that a claim can be done in under two minutes. Let's say you have an accident. Immediately, using the data that's coming out of the car, you're able to recognize that there's been an accident. So you don't need to make any phone calls. You just get a push notification on your phone saying, are you doing okay? Do you need medical assistance? If no, do you want to file a claim? You say, yes, we say, okay, why don't you record a 360 degree video of the car? As soon as you start recording the vehicle in in on video, we can recognize this part is damaged, this can be repaired, this needs to be replaced. And as you're walking around the vehicle, we're telling you, okay, it's going to cost twelve hundred dollars to repair all of this. And you can say, okay, I'm gonna take the 1200, get it repaired on my own, or I want to see a list of garages that are near me so I can drive into them or get the car towed. Right now, all of this can be done in under two minutes. Right? So it's a completely different way. And if you ask me, in the next 10 years, almost all claims will be processed in under two to five minutes. And that's the future of auto insurance. And then finally, I think there's the road safety aspect. Technology exists today to prevent accidents before they happen. We can recognize if a driver is falling asleep, is talking on the phone, is distracted, is about to get rear end a vehicle and be able to give them specific alerting well before they are about to take evasive action. So we think there's a tremendous amount of innovation driven by AI that's coming into this massive industry.

You know, you bring up a lot of great points there, Rohan. You know, and being able to take this, you know, older process, make it faster, you know, getting car repairs, waiting for days between each step from the insurance company. So I guess if the technology is there and you know, most new cars, I'm sure have a lot of these, you know, features or capabilities. What's, what's the, I guess what's the roadblock, so to speak, from, from, you know, having this realization, is it there's maybe too many older cars on the road? Is it, you know, maybe drivers unwilling to, I guess, share some of this data from their car that can in theory collect it? What are those big roadblocks or hurdles until we can get to that point where the auto insurance industry is way faster and can help make everyone safer?

I think roadblocks are that AI is actually transforming very quickly. You know, these are large companies, large insurers, large car companies. It just takes a while for, for technology to percolate through the entire ecosystem. There's a billion point five vehicles on the roads today. About 20% of these have connectivity capabilities. And when you have connectivity capabilities or what we call the software defined vehicle, which is really like an iPhone on wheels right now, I can make decisions based on the data coming out of a car very quickly. So we think this is going to be a 10 to 15 year change that has already started and we are seeing acceleration in this change. So the way the Internet transformed, E commerce, identity, entertainment, you're going to see the mobility ecosystem transform and you're going to see tremendous applications built around the car like insurance, what we are doing, entertainment in the car, in car payments, so you never have to take out your credit card when you're inside the car. Identity of the car, cyber security of the car. So we are seeing these changes come through, but it will be 10 to 15 years before they are fully realized in the ecosystem.

So I, I have a question here that I want to get to from Jay, but before we do, I need to take one quick break and shout out Work Lab here. So real quick, if you don't know, the Worklab podcast for Microsoft is made for leaders who want us to understand how work is changing. Because effective leaders adapt. They stay ahead of trends, they embrace any edge they can get. They also know that AI powered organizations will be better at spotting opportunities, creating new products and business models, and maximizing value for real world lessons and actionable insights. To help you stay ahead, check out the Work Lab podcast. The new season launches September 12th. That's W O R K L A B no spaces available wherever you get your podcast. Yeah, I'm excited about that one. And thank you to our sponsors from Microsoft Work Lab. So, Rohan, like, I want to get to this question here from Jay because I think it's very important. So he's asking, how does roadzen use or manage all of the data that cars capture? Right. Like what we were just talking about. And then also the privacy of that data in. And any consent on the user side. Yeah. How does this work? Or at least how does your company make this part work?

It's very clear that the data belongs to the user. It does not belong to the car company, it does not belong to roadzen. Without the user's consent, we cannot act. So only if you give us the consent to be able to take the data, to be able to help you in making your driving safer, to be able to help you during a claim, could we do that. If there is no consent, there is no data. We cannot use it. And that's very clear now across the world, that the data really belongs to the user. It does not belong to the enterprise.

Yeah. And, you know, you mentioned something, Rohan, just about road safety. Right. So we've talked a lot so far about on the front end, the insurance side, and, you know, some of the antiquated roadblocks that are still in the way. So, I mean, what could this mean in the long run for road safety? Because you mentioned that, you know, since the pandemic, it seems like people's driving habits have changed. There's more accidents, cars are more costly to repair. So how could some of this AI, I guess, on the front end, in theory, help roads be safer in the long run, at least as it comes to, like, working through the insurance agency, the insurance kind of category.

Road safety is fundamentally intrinsic to the world of insurance because if you have, what's the best policy is a policy where you don't have to pay a claim. Right. So the way it works is if you can make the car safer, you can drive down premiums for everybody, and you can actually, like, have better societal outcomes. Because accidents are not just a pain for insurers. There's loss of life, there's loss of other property. There's a lot of societal issues that are linked to accidents, and we believe you can. Today. The technology exists to take human distraction errors to zero. And 70% of all accidents are caused by what is human distraction. You're looking at your phone and you hit someone. You're not focused on the road. And Something happens in front of you. Now here is where computer vision plays a real part. How do we drive? The primary mode of driving is through vision. So as we train cameras to look at the road, to look at the driver, we can actually say, okay, the time to collision is less than a second and the driver's looking down. There's potential for an accident. Why don't we issue an alert? And we think you can take these 70% of accidents caused by human errors down to zero. And we think that will be just a tremendous benefit to all of society. And it's going to be great for insurers because, you know, you can drive down rates for people and still be more profitable. So it's just fantastic for everybody.

So, so how does that work then? Right, so you know, driving down, you know, the 70% of accidents caused by human error down to zero. Right. I know I've been in, you know, I, I still have an old school car, right. It's still running, it's got like a trillion miles on it. But you know, does that just mean everyone needs, you know, cars like, you know, Tesla or cars with, with auto assist? And you know, what happens then if there's still many people on the road who maybe don't want to engage those? Or is it more of just smarter and more proactive kind of alerts in the car? But you know, how does that come into play when there's so many different, you know, makes and models of cars? There's different types of technology in all the different makes. How does that work in the long run to bring that 70% down to zero?

Well, Tesla has eight cameras around the vehicle, right. So it's kind of mapping the entire car using cameras and the road and other people's behavior on the road. And what we are doing is if think of Tesla like Apple, they're building for Tesla technology for Tesla cars. What we are building is technology that can be deployed with any car maker using a simple dual sided camera. So it's a dash cam you can just paste on the windshield and essentially it's looking at the road on one side, looking at the driver. And we've seen tremendous adoption specifically for commercial fleets which are large trucks, etc. And this is less than a thousand dollar device that anybody can put on and they could actually start seeing the benefits, the reduction in accidents and all of this and eventually what's gonna happen as not just the cars become connected, but the entire grid becomes connected. There's something we call a V2X or vehicle to everything Communication vehicles will start to be able to communicate with each other with the grid, with traffic lights, with, you know, recognizing there's an accident like a mile ahead. So let's just take a different route. So the connectivity benefit will not just be introducing collisions, but actually making everything smarter that's involved in the world of mobility.

I've always wondered why cars can't communicate with each other. Right, like, like you bring up a great point there, Rohan. Like it seems, it seems so, you know, basic yet novel at the same time. That's, that, that's an interesting one. A good question here from Cecilia. I'd love to get your thoughts on. So she's asking how will the AI systems be able to differentiate between different driving styles and skills of driving? As an example, some people drive one handed better than others who drive two handed. Some have vision peripherals that are greater than others. Yeah, how, how can AI in the future kind of compensate for so many different, you know, driving styles?

One of the core things that AI does incredibly well is to do pattern recognition. So just as your Spotify playlist will get tailored differently than my Spotify playlist, it's the same with driving behavior. I will learn your driving behavior which will be super personalized to how you drive. So you may drive faster than me, but be eventually a safer driver. You may have better vision, you may have better control on your reaction times, but the AI will be able to learn that and make personalized predictions for you versus different predictions for me. And we think that's been, you know, like, look at all the recommendation systems, look at your YouTube feed, your Netflix, your soup. Spotify AI is exceptional at this thing which is called personalization for every single person on the road.

And I was actually thinking this as well. So Denny, thank you for this one. She's saying if there is an accident, who would be the responsible party? Right. And I'm not just asking specifically Rohan about roadzen, but you know, in, in general. Right. Like if everyone is tapping into AI and computer vision to keep them safer on the road and there is an access accident, what if both people are using some kind of AI powered mechanism like who's ultimately at fault then?

That's a great question. We exist in a world where all cars are driven by people and we are going to a world where eventually there'll be some autonomous cars. Right. So we're going to go from what is called personal liability. You are liable to the product itself is liable. So there's product liability or the car company. The AI will be responsible for the driving behavior on the road. But we are not there yet. So we are, we are somewhere here. We gotta get somewhere there and there's gonna be a big journey in the middle.

Yeah, I, I don't know. For me personally, I'm fine with that. Right. I'm, I'd be fine to be hands off and you know, hey, if I get in an accident, it's on, it's on you Car, car company. I wouldn't mind that. Another, another great one here. Our audience is on fire with great questions this morning for you, Rohan. So Monica asking, are you working with commercial and large logistics companies where they have hundreds of drivers on the road driving long hours and then how is the adoption of this technology with those large companies?

This is the fastest growing part of our business is working with large fleets, commercial trucking, large, you know, car leasing companies, etc. Because they have thousands of drivers, they have thousands of vehicles on the road. They're driving long hours. What we found is in a night journey, 25% of drivers will fall asleep at least once. This is just a super dangerous behavior. Right? There's, there's stuff like electronic driver logging, there's stuff like there are different GPS systems in those trucks today. You can replace all of those with this computer vision technology that can recognize, identify and help drivers. We call it Drive Buddy. We think the commercial fleets will be the fastest to adopt because they really care about making sure their fleets are safe, they are not causing damage to other people and stuff like that. And the privacy concerns also lower because you're actually in a job and you know, so the fleets mandate that this technology must be used.

So we've, we've talked about a lot here on the show today, Rohan. I mean we've talked about everything from how antiquated the, the auto industry is, changes that are currently in place, roadblocks keeping us there, the future of autonomous vehicles. Like we've covered so much, but maybe what is your one biggest takeaway as we wrap here about how AI is transforming the auto industry and road safety and what we should all be, you know, as individuals, what we should all be paying attention to.

I think you're going to hear a lot about AI. You're already seeing it. One of the things is that there's AI beyond what is just large language models, LLMs, ChatGPT and all of that. There's AI for computer vision. Computer vision digitizes the real world and allows you to make decisions based on that. 90% of our decisions are based on seeing and we're going to see tremendous amount of innovation and applications built around computer vision and what roadzen is doing is a subset of that in the auto insurance industry. Look forward to many new innovations that outside of the current hype cycle about just LLMs and multimodal models.

Wow. I think this was a very eye opening conversation Rohan, because this is something that in theory impacts us all right. I think auto insurance is one of those things that we really don't think about it until it's slow or it's causing us problems. And I think today that you were really able to help us see past that and to see how AI is impacting us all in this. So Rohan, thank you so much for joining the Everyday AI Show. We really appreciate your time.

Thanks a lot. It was great being here.

All right. And hey, as a reminder everyone, we covered a lot, so make sure if you haven't already, if you're listening on the podcast, we always put that link in the show notes. Go ahead and click that to go to our website. Your everyday AI.com Sign up for our free daily Newsletter. We're going to be recapping today's conversation, bringing you additional complimentary supplementary insights to help you learn even more. And we do that every single day. It is your one place to stay up and stay ahead with everything generative AI. So thank you for tuning in. If this was helpful, please repost this subscribe Leave us a rating on Apple or Spotify, wherever you're listening. And whatever you do, please join us tomorrow and every day for more Everyday AI. Thanks y'.

All.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going for a little more AI magic. Visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.