AI, Computer Vision Preventing Injuries, Transforming Insurance

Pete Miller [00:37]:

Thank you for tuning in to the Predict & Prevent podcast! In today’s episode, we’re exploring how AI and computer vision are being used to improve workplace safety and reduce costs for businesses.

Our guest is Josh Butler, the Founder and CEO of Comp Science. With a background in computer science and artificial intelligence, Josh has helped build self-driving car perception analytics and held product management roles at companies like Facebook. But it was a construction accident involving his father-in-law that led Josh to focus on safety. He realized that the same computer vision technology used in self-driving cars could also identify workplace hazards and prevent accidents. CompScience offers what they call “active insurance.” They use AI-generated data to provide risk insights for underwriting and loss prevention. By adopting an MGA, or Managing General Agent, business model, CompScience sells workers’ compensation insurance in a way that businesses are already familiar with. This approach seamlessly integrates their AI technology into the existing value chain. In our discussion, Josh shares his journey and the innovative ways CompScience is making workplaces safer.

To get started, Josh, please give our audience a high-level overview of what CompScience does.

Josh Butler [02:04]:

Sure thing, Pete. So what we call it is active insurance for commercial property casualty insurance. It’s a little bit like what you can get these days in like cyber or property or auto. You know, some of those lines have seen some IOT influence or even satellite imagery starting to disrupt those lines. But what we were able to do is really identify the richest data streams available about the physical world, you know, that really impact people and their place at work. So you know, what we’re doing is we’re starting with, with workers comp, and, soon moving into general liability. And we’re seeing like, there’s, there’s just a huge missed opportunity with these lines to protect people, and, and physical assets as well. Kind of like what cyber and property and auto disruptors are doing. But, you know, by focusing on workers comp and GL, which have been like really traditionally not super innovative, like, you know, pretty commoditized, everything looks kind of the same. We’re able to really see a lot of low hanging fruit here and really bring down our customers’ mod rates and offer some really attractive pricing in some cases.

So, how do we do this? So we had to build a suite of products that turn video and imagery from work sites into risk metrics, we call them causal risk factors, to generate some insights and some simple recommendations and then use that for underwriting and loss control. And I think one thing we’re really proud of is that our customers really appreciate the recommendations we make so much they’re like, we’re seeing 78% adherence rate with the recommendations.

Because they really believe that what we’re telling them is relevant and useful. And then we can actually verify that they made the changes. That’s why we’re so confident in being aggressive in the market here.

I think one more thing about what we do is contrary to some unbundled solutions that do utilize video from like CCTV. And we really designed this as an insurance product. And, and that means like, we really took it from the, from the bottom up. Like what do brokers need in order to sell effectively and what do customers need in order to adopt without a whole lot of pain? So we’ve made this into, something that’s like incredibly easy and quick to implement and, and, and really doesn’t, doesn’t have to change very much from what brokers usually sell as workers comp. So, it’s active insurance for, for commercial property casualty really, really designed from the ground up.

Pete Miller [04:21]:

I like that term active insurance. That makes a lot of sense. What motivated you? That’s a very interesting story. But what motivated you to concentrate on workers’ comp or workplace safety first and sort of the associated AI technologies that you’re using?

Josh Butler [04:40]:

Yeah, I think, you know, contrary to, to, to many people on my team who do come from the insurance industry. I didn’t grow up thinking I was going to be a founder of an insurtech company. I moved to California about 15 years ago. you know, got a job in big tech and I’ve been working in, startups, at, as well as, you know, big public big tech companies.

Most recently I was in self -driving cars. Kind of the thread through my career is I’ve been building AI based products. it’s always been my passion. I always believed this is going to go somewhere even when, you know, before AI was cool, maybe. So, you know, just before like the real, the real inspiration for, for comp sites came from, you know, I’m working away on a self -driving car problem at, an EV company in San Jose.

One day my father-in -law who used to work in construction had an accident on a construction site and to hear him talk about it, he’d actually kind of predicted it was going to happen. He knew that the site wasn’t safe, but you know, like if you’re not the foreman and the foreman says like, don’t worry about it, go back to work, then…

It’s kind of what you do. I mean, he, supported a family for like 30 years. Pretty good, like middle cap, middle -class kind of income. you know, working hard day in and day out and on these, construction sites. And, and one day, he got ejected from a crane and that was the end of his career. And so they’re living through that and seeing the impact, not just on the family and everything. but on like his, physical and mental health, honestly.

It was really, that’s really what inspired me to think, you know, could I use the same technology that I’m building in self-driving cars, this perception technology that is really start, just starting to understand how the world, the physical world from just 2D video, can we apply that in workplaces to identify potential hazards and then prevent things from happening like what happened to my father-in -law.  When we started this five years ago. It was still relatively early in computer vision. But honestly, there’s been tremendous progress since we started, not just for self -driving cars, but also more recently at OpenAI and then even Google and Microsoft to come out with their own versions, too. So we’re seeing just a tremendous advancement in what’s possible with visual AI. That’s really even accelerating our roadmap beyond what honestly I even believe would be possible by now.

Pete Miller [07:04]:

You mentioned brokers just a minute ago. Can you give us a sense of, or maybe a story about how brokers who use CompScience’s solution have been able to sort of increase the value that they provide to their clients or give them a competitive advantage to win business?

Josh Butler [07:22]:

Yeah, so I can think of one. There’s a principal at a really successful agency in the Midwest, a part of one of the top five agencies out there. He tells all his colleagues about how he won some business away from Marsh last year, which he was really excited about.

This is a big food processor, one of the larger food processors in the world. The subsidiary had about, has about a thousand employees. It was a really good opportunity for him and, he was excited to, to go after it. In fact, the workers comp was the biggest item as is often the case in sort of manufacturing and logistics and, you know, a lot of our core verticals. The way we structure our proposals, we have a lot of analytics upfront that sort of demonstrate how we plan to bring down their mod rates, which can make a huge impact on their overall total cost of risk. And he decided to lead with our product. It’s kind of, the way he puts it is like, when you lead with our product, you get to have prospecting conversations that you wouldn’t otherwise have. We helped them win the business. And we got to work right away and we identified a few like high-hazard activities. there’s ergonomic risks there. There’s, as with most manufacturing and food processing, logistics companies, lots of forklifts. So we identified some activity around those forklifts that was, people were actually being elevated into the air with the forklifts, things like that. We,

provided some insights to the leadership there. They were a little surprised by what they saw and immediately started making changes. And so we’re seeing some great trends there already.

But I think sort of just to bring it back to the sale, it’s really hard to differentiate in workers comp, I think for most, most brokers. So it’s not often the thing that they lead with, but in this case, it’s like having a tech based product with a great track record. I mean, we’re seeing, you know, up to 60 – 70% reductions in injuries for some customers. On average, we’re looking at about a 17%. So, really, I’m really proud of like that, that impact in the first year, in the first few months on average across our customers.

People really look at this as something that’s like a value add. So from here, I guess it looks like we’ll be going after some of the other subsidiaries inside that conglomerate. It’s a huge food processing conglomerate and there’s a lot more. We’ve set our sights high to conquest and conquer the rest of that group over there.

Pete Miller [09:38]:

So you talked a little bit about your technology and, you know, technology is growing rapidly and, you know, we’ve seen some other workplace safety technologies. So I’m curious, how do you stand out from other similar technologies?

Josh Butler [09:54]:

I mean, number one, I think that success is about much more than technology. I think that technology means you get those prospecting conversations that you wouldn’t otherwise have, and you can deliver value that we wouldn’t otherwise be able to deliver. Absolutely. I think the technology speaks for itself. What makes us really unique is that we marry that with an underwriting team that comes from, you know, executive leadership of some of the most respected carriers and brokers in the industry. We really think about, you know, what drives success for our customers, and frankly, that’s our brokers. And we look at what’s important to them. We partner with them to help them grow their books, right?

So what’s a great partner for a broker? It’s insanely fast turnarounds on quotes. It’s simplifying the quote process. So there’s no pre-inspections needed, only in rare cases. We have great paper, A-rated paper. You got a direct line to your underwriter. That’s often not the case with the larger carriers out there. We also don’t use a TPA. Our claims are handled by a really tremendous claims team at Nationwide. and of course there’s great pricing. I mean, we, we can get a lot more aggressive in some cases where we, where we know there are certain types of risks that we can mitigate it really effectively. So I think that really like, that really makes us stand out, in this bundled offering that is like active workers, comp insurance. and of course, like, you know, the technology that’s what helps drive a conversation that helps the broker differentiate themselves.

I think when you compare us to an unbundled video analytics technology out there, there’s just no comparison. They’re in a different market entirely. These unbundled offerings, they take months and months to implement. It’s a SaaS sales cycle. So it usually means like six months to a year to sell through. Where we keep things really simple. We just, you know, we don’t charge any extra for the technology. We just bundle it right in with the workers comp, just like you’re used to selling.

Pete Miller [11:51]:

That’s really interesting. That’s a unique approach. I’ll come back to that in a minute, if I might. So are there any stories you can share about CompScience beyond the broker story you shared that identified a big problem, which, as you mentioned, workplace safety is a big problem, but where you’re really happy about the results in terms of risk reduction?

Josh Butler [12:07]:

Totally. One comes to mind, they’re one of the top three global logistics companies. One of those three letter companies. We’ve been working with them for two or three years. We’ve been rolling out across the Southeast and the mid Atlantic. And in one of the sites we noticed, we measured that…for forklifts, were traveling at a much higher speed than they should be, with elevated loads. Among safety professionals, it’s pretty well known that forklifts are some of the most dangerous equipment that people operate in and among. I mean, something like 11, one in 11 forklift operators are injured on the job every year. So we saw that these things were moving significantly faster than they should with their loads elevated. It’s a flip over issue as well as a potential line of sight issue. They might not see people in front of them.

So we showed these insights, we showed the numbers to the team there and the initial response was, it’s just not possible. Like this, this is impossible. We’ve done our work here. We calibrated the rate limiters on these forklifts. So that you can’t operate them at a high speed with elevated loads. And they were very convinced, honestly, we would have been very convinced if we didn’t have the data in front of us. So we showed them the data and they’re like, well, is there any way this could possibly happen? So they went back and they looked at the configuration of these forklift rate limiters and they realized that they hadn’t been calibrated properly.

So the results were actually pretty incredible once they recalibrated these things. We actually saw not just a 64 % reduction in injury rates at this site, we actually saw an 88% reduction in property damage, which is not something we generally focus on, but often there’s the side effect that property damage goes down. Apparently they’d been spending a lot of money fixing the racks because the forklifts were bumping into the racks as people were zooming in and out. But now with the rate limiters properly calibrated, all of that was completely eliminated. And then not only that, we went to the other sites that were active and we saw the same thing. So they’ve actually been recalibrating their forklift rate limiters across the board. And that really never would have happened had we not had the hard data to show them. So really, really excited about some of the results we saw there.

Pete Miller [14:23]:

That’s really interesting. Honestly, I hadn’t thought about the property damage, but that makes sense. Forklifts are running into things, right?

Josh Butler [14:30]:

Yeah. Have you ever seen a forklift without a dent in it? I’ve never seen one. We do actually insure a forklift distributor, probably the leading forklift distributor in the Southeast. So as you might have, their clients are getting access to a free trial of our software. So forklifts are a great example of a piece of heavy machinery where we can really have a big impact.

Pete Miller [14:53]:

Interesting. You mentioned that right now, workers competent. I think you said commercial general liability. Are there, do you have other plans to expand CompScience?

Josh Butler [15:03]:

the most natural extension for us is really to move into construction. And we’ve already done some work in construction. In fact, we have a relationship with one of the largest construction associations out there where we’ve been really kind of co-developing a solution that adapts our vision technology into the field to address field work. You know, there are a few extra challenges when you talk about field work. It’s less structured. It’s often less supervised. You might have loan workers even. But also you have more challenges around data and power. We’re not using existing CCTV video feeds. in some cases we use a cell phone and others we use a temporary mounted camera.

But there’s what’s special about construction as well as it’s extremely high-hazard activity. It’s like the leading driver of fatalities of any industry in the United States. It’s a very large industry. In fact, that’s why it’s a big focus for the National Safety Council. We’ve partnered with the National Safety Council and some people working there on their quest to drive down severe injuries and fatalities.

And to do that, we’ve built this app that is actually helping educate and drive engagement around a new model for understanding the precursors to severe injuries and fatalities, the high energy hazards, the high hazard activity, if you will, that can lead to death or severe injury.

We have some customers that are in the industry, but we haven’t rolled it out broadly yet. We are actually opening up a wait list for anyone who’s interested in learning more about this. They can find that about on our website.

We’re always developing new products and we’re expanding into new verticals and new lines. We’re planning to open up a wait list for early access to construction, workers comp in the next few weeks. Excited to talk to experts in the field that, you know, work in construction risks. Also, there’s a natural extension into general liability with retail. There’s just as much risk to customers inside retail and restaurant as there is to the employees.

Pete Miller [17:21]:

Let me just circle back because you had a very unique approach with an MGA and you mentioned that, could you just expand on that? Like what caused you to do that? Was it the long sale cycle that you’re experiencing or was it, you know, resistance or because there’s a different way of purchasing or can you just talk a little bit more about that? Because that’s pretty unique, you know, Josh, in terms of what I’ve seen.

Josh Butler [17:47]:

Well, it’s interesting you noticed that. I’d say I learned a lot on how to build products that people can sell. Early in my career here in California, working at an early startup, founders that I worked with were excellent product people. And they taught me, if you want to build a product that’s successful, don’t try to sell someone something they don’t already know how to buy.

Right. And I think that’s something that a lot of the other safety technology startups out there are kind of falling prey to is there, they’re saying, I have this new technology. It’s great. It’ll make a huge impact. And then trying to figure out who can buy this. And unfortunately, there’s, you know, there’s like the safety manager, who doesn’t have a ton of, a ton of allocated budget for these kinds of new, new technologies, innovative new technologies. That said, like every, every large industrial or every mid market company out there with a significant risks in their operations, they spend a lot of money on workers comp.

And they do rely on recommendations and sort of audits and inspections from their carrier to get ideas for what basically they should be focusing on. So, I looked at that and I said, you know what, like, why don’t we just sell them what they already know how to buy? And you know what, there’s, there’s also this incredible distribution channel, brokers who really like generally do know their customers quite well and are looking to really make a unique impact in some way. So figuring out how do we fit ourselves into that value chain effectively.

There’s not a ton of margin there, frankly, for an MGA. You gotta squeeze the product into a pretty narrow rung there in the value stack. So it forced us to rethink from the ground up, like how do we build this product? How do we deliver it? How do we make it so that it actually, you know, that we can pay for this and without, you know, having to raise the prices on customers and it forced us to be really efficient.

So now we figured out how to run the the computer vision and the analytics and deliver great service for the same price as what it would cost a typical carrier just to run loss control, essentially. And I think the indications are that it makes a lot of sense. It’s kind of a no brainer honestly at this point. When a broker brings our product to a customer, it’s a great A-rated paper, great claim service, can’t deny that, but with so much more, so much more value than you get otherwise. And that kind of makes it a no-brainer for the customer.

Pete Miller [20:35]:

I think that’s a really smart idea. I can tell you, like, as I interview people and talk to people, particularly insurtechs, there’s sort of a Rubicon if they try to change a workflow, change a company’s workflow. Or, as you’re saying, change the way they buy things. That’s sort of a dividing line, right? And the more you try to do that, I think, you know, that’s a really good idea.

Just a couple other things. You mentioned that the primary input is high quality video. Is that correct?

Josh Butler [21:12]:

Yeah, so the core technology is essentially running video analytics on top of video from CCTV camera systems. There’s no need to install new cameras, which I think is one thing that’s quite unique about us. No need to install new cameras. We don’t have to get access to your network. We don’t have to go through all of those kinds of hoops to start to deliver value here. We’re not dogmatic about it. You know, we also built an app that can just take a photo, or we can generate insights just from the photo. And you know what? We have partners that look at telematics data, and we can process their data to understand likelihood of injury for a driver, for instance, on the road. So we’re not dogmatic about vision being the only useful data set. We just believe it’s the richest that you could possibly ever have. And frankly, up until three or four years ago, it was way too expensive to really leverage it, if you could at all. The amount of data that you have to transfer when you’re looking at video is enormous. I mean, talking about what, what about a gigabyte an hour? If we are very efficient about how we get it.

By now we’ve gathered over a hundred billion images. That’s millions of hours of data. The cost is significant. But one thing we knew is that the cost of processing that data and storing it would come down pretty precipitously. One fun fact I just heard is that OpenAI’s costs are expected to drop by 99% over the next two years. That’s the rate of change that we’re seeing in this industry. I think it’s probably not going to be that drastic in the video processing that we do. But one thing’s for sure, processes are getting faster and faster and more efficient. And then our software is also getting more efficient as well. And with that, we’re able to process more and more data, get a richer and richer picture of our clients’ operations, and deliver more insights.

Pete Miller [23:21]:

Josh, it’s great to hear you’re getting adoption for your approach because it’s the way we’re going, right? Predict and prevent. Makes a lot of sense if we can just avoid the loss.

Josh Butler [23:30]:

There’s, I know some investors that agree with you. yeah, it’s, it’s, it’s, it’s really obvious opportunity. The biggest opportunity is certainly on the loss side. and everyone wants to apply AI to claims instead and like, you know, process and everything. Or, okay, sure. You’ll maybe, you probably can make like, you know, three or four points of improvement in your combined ratio, but like, what if you can drop losses by 30%? I mean, why would you look anywhere else?

Pete Miller [23:56]:

Well, and think about the story you tell your customer, hey, guess what? I got a tool. We’re going to avoid it in the first place. You’re not going to lose time. You’re not going to lose production. You know, you’re not going to get sued. All that headache can go away.

Josh Butler [24:09]:

It’s absolutely. And I think a lot of them really do appreciate the improved visibility. Like they got a million things going on. If they have hard data, they can have confidence in that. At least they can have confidence that like they have transparency into their operations. So yeah, I think I really appreciate that

Pete Miller [24:31]:

Well, Josh, thank you for your time. I really appreciate it.

Josh Butler [24:33]:

I really appreciate you guys inviting me on. I mean, it’s a really cool podcast you’ve got going and it’s just so in line with our vision for creating active insurance.