Insurance & Artificial Intelligence
How do AI and internet of things impact the insurance industry?
Host: At the recent SXSW conference, Argo Digital held a panel discussion on the future of big data, artificial intelligence and risk. This interview was conducted on site in Austin, Texas at the Half Step Bar. Senior Vice President, Andy Breen, was on hand to share his thoughts on why now is the time to talk about how data collection, artificial intelligence, and the internet of things impacts risk.
Breen: So, artificial intelligence has been discussed for decades, but there’ve been many winters where it really went out of favor. The big difference now is that we have a volume of data and computing power that we never had before. So we can actually put some of the theory in the practice like we never could in the past. And specifically tied to risk, insurance has always been a data-driven business. And so now, instead of having long-form applications that people have to fill out and use that information, which may or may not be the right information, may or may not be good information, we actually can go and use real information and the data that’s available through public sources, commercial databases on the web, things like that
Host: In today’s connected world we use wearable tech, social media, apps on our smart phone and more to keep track of our daily habits and lives. When asked what the explosion of data collected by everything from mobile to IoT to drones meant for risk assessment, Andy had this to say.
Breen: What it means is that we can actually use different factors in looking at risk. So, say today, we are here at a restaurant, today this restaurant would be asked, “How many employees do you have and what’s your revenue?” I’m not really sure that those are real good indicators of the risk that someone’s gonna slip and fall in the front stoop, but if we have these other devices, we can see if there are spills on the floor, is there something wrong with their roof, or other factors that will actually contribute to a potential downstream claim or other thing where we can actually advise them to improve their risk profile and help them manage that process.
Host: In the past, the biggest challenge for use of big data was the lack of systems and algorithms to translate incoming data into something that can be used to assess true risk. Now there are tools in place which offer real understanding and solutions. Andy goes into detail about how A.I. is used to turn data-knowledge into data-understanding.
Breen: So more data at first was a problem, because the tools that we had just couldn’t flat out handle them, and the approaches that we had just couldn’t handle them. So, with the last five years, we’ve had a real explosion and really powerful tools that were just fundamentally different approaches to doing that. So, what we have now is the ability to actually handle that, and because we’ve changed, you know, what we’re trying to do here. But the thing is, a lot of this is locked away in kind of fairly what we call unstructured data sources, right? So yeah, there’s a database of kind of quantifiable data, but what’s really interesting are the social media and SEC filings and other data sources like that where we can go in and pull out what we call qualitative information and understand that better. And that’s where the artificial intelligence comes in because you’re not just looking for keywords, you’re actually trying to extract meaning from those data sources.
Host: The ability of artificial intelligence to derive meaning form data-analysis is the key to fully utilizing big data. The next step would naturally be for A.I. to fully understand us.
Breen: Yeah. I think that’s really the best application in the near term where we can actually go and start basically… You know, think about how much you talk through your devices, it’s the same fundamental technology, right? So voice interfaces have become very real now. In fact, they just announced the other day that they’ve achieved 96% efficiency and only 4% error rate. That’s actually better than humans as far as understanding language. So, the algorithms now can really actually understand us. We use that same model to actually go and look at textual data sources. In some ways it’s actually easier because we don’t have to deal with noise and other factors, but the big thing is now unlocking the meaning that is in those types of documents, and we’re just at the very early stages of really understanding what we can do with that.
Host: We all know that technology moves quickly, what was cutting-edge 10 years ago has been replaced many times over. We asked Andy to look back over the last 5 years and share how things have changed how it comes to processing and using big data.
Breen: So five years ago, we didn’t have the tools. We didn’t have the process. So think about this. All data kind of structures the processing, you know, five plus years ago was on a fixed known data set. So basically go, get a database, and let’s run some things to do that. So all the tools and all the approaches were done that way, statistical methods were all done that way. We, the tech industry, have this huge explosion and really Googles, the Facebooks, the Twitters, the Amazons in the world were faced with this problem where the old tools and the old approaches just didn’t work. So what they’ve done is they created brand-new tools mostly around something called pipelining where instead of operating on a fixed set of data, you operate on a stream of data and building those tools has completely revolutionized how we approach doing this and made it real.
Host: Big Data is a great tool for companies that have teams to support the increased data-knowledge. Andy offers companies without staffed data-scientists and A.I. PhD’s on how they too can benefit from new technology.
Breen: So, it is difficult these days to do that because there is some pretty advanced things going on. However, the tools are getting advanced quite rapidly. And here’s the interesting thing, a lot of the best tools and most of the best tools out there are open source or free tools. Right? So the tech community has built them and then for the better of everyone kind of put them out there. So you have things emerging like not knowing how to create a neural network for an artificial intelligence applications, but there’s actually neural network as a service now. Right? There’s things like TensorFlow, which was put into open source by Google, which is incredibly a powerful tool. So your average person still couldn’t go on and use those things, but the bar is being lowered for the cost of them and the expertise needed. So, it will get to a point, probably in the next 5 years, I guess, maybe 10 years where someone can go in and just describe the type of problem that they’re doing and the type of data they wanna look at, and then dynamically an artificial intelligence network will be built.
Host: With so much Big Data easily accessible to the public and private sectors. Andy addresses the question of whether the government can take advantage of data and technology when assessing risk.
Breen: Well, interestingly enough, the government has probably the best data. The government has always been the classic source of collecting data, whether it’s the census or tax filings or all these other things. So, the government is probably sitting on the best set of data that there is. I don’t that they’ve always had the tools to do that. Now, the interesting thing is, how can private industry, academia and others help the government get there, because obviously the government’s dealing with a whole other set of risks, whether it’s catastrophes, whether it’s terrorism or other threats from that, but I think they need to start looking at the techniques. The big thing that’s changed now is how rapidly thing…the environment has changed. Right? Because as technology move things faster, so in the scale that you can think things can happen. So, any counter measure or any way to actually harness and do these things has to be jumped on very rapidly. Unfortunately, government is not known to be super nimble on being able to do that. So I think the government has to find ways to be able to have one of the quick strike forces like they do in the military. I think they have to have the same thing in the cyber area and the data area.
Host: Big Data and the tools to analyze it are advancing at a fast pace. By comparison, many times when we think of the government we could say that they move about in a slower, more measured pace. Like a giant cruise ship which is slow to adjust and unable to make quick course correction.
Breen: Exactly. Here’s an interesting corollary. The military, you know, when you talk about the military that was built to fight the Cold War in the U.S., that was not set up to fight guerrilla wars that was not setup to fight terrorism. The military was not setup to…they were setup to fight tank battles in Asia or nuclear battles. They were not setup to battle guerrillas in urban environments. But that’s why the Special Forces was created, right? So they created the Special Forces to basically go and be able to be highly, highly adaptive, and they affirm all the classic rules of military doctrine to do that. So they’ve had to adapt before. And I think what they realize now is they have to also adapt to new methodologies and how they deal with this.
Host: This leads to the comparison of the special forces being like todays modern business: more like a speed boat which can adjust to changes quickly.
Breen: Absolutely right. Yeah. So you can’t move a whole army across the ocean in 24 hours but you can have a quick strike force that can actually have a real impact. Right? And honestly you see what military is moving to, it’s all Special Forces now.
Host: When asked to speculate on the future of Big Data, A.I., IoT, and risk, Andy shared these insights.
Breen: So, I think on the immediate horizon, as I mentioned, is the ability to basically…for us in the industry, is to basically look at things that we could never look at before. Again, why you would ask a certain business, you know, their revenue, and use that as a proxy for risk? Well, that was probably the only things they can actually give you because that’s one of the few things they know. Now with the ability to use everything from drones to sensors and things like that, I can actually potentially use things that are probably much more indicative of risk, and by the way, lower the friction, and that I don’t have to ask you. So the holy grail that I put out to my team is, I wanna be able to say something like all restaurants in the Northeast Corridor, I wanna understand what are the risks that we like, and be able to go and access them, so I can just walk in and say, “Here you go. I already know who you are and I’m comfortable working with your business because we know about you. You wanna sign on the dotted line.” So that would be actually the ideal case. We’re a little ways away from that, but that’s where we’re trying to head to.
Host: At Argo Digital, we have one mission: Turn the insurance paradigm on it’s head and define the way people assess and transfer risk in the 21st century. Helping small business stay in business.
To learn more about how your business can leverage technology to transfer risk go to argo-digital.com
Argo Digital is a division of specialty insurance provider, Argo Group.