So, joining me for this session, we have, on my left is Nicolas Levillain, who is head of data science innovation at EXIN Group. He leads EXIN’s data machine learning and statistical development. He previous run consulting Baring Point’s HyperCube research and development centre, building and deploying big data software. He was also co-founder of effiScience, an early innovator in using algorithms to help businesses increase profit and reduce risk by predicting complex events.
Then to his left is Vincent Branch, is chief executive of XL Catlin Accelerate, which is a XL Catlin’s internal innovation team. He was previously at Vitality Health, where he held the position of chief underwriting officer and chief actuary. He was also prior to that the global chief underwriting officer for personal lines at Zurich Insurance.
And then lastly, we have James Platt, who’s chief operating officer of Aon Risk Solutions. A former partner at the Boston Consulting Group, he joined Aon to lead its carrier consulting business, Aon Inpoint, to drive analytics across the risk solutions business. And then he became COO of Aon Risk Solutions last year and his role includes accountability for the Aon Centre for Innovation and Analytics in Dublin. So, thank you for joining us today.
Just picking up on the artificial intelligence side of things, I mean obviously you mentioned that earlier with some of the work the Accelerate Division is doing, how big is the opportunity to utilise AI in the commercial insurance sphere?
There’s certainly a lot of hype related to that; but we do believe its fundamental to all areas related to our business. But I think we’ve got to be really careful in how we position that and how we think about applying that. So, we very much, believe AI is going to be fundamental, but more importantly, we think it’s going to be much more about augmenting our decision making and supporting our decision making, rather than replacing that. I think that is a really important part. Again, if we think about one of the biggest challenges we all have is the sheer breadth and the size of the data sets that we are dealing with. We know beyond that point where traditional techniques and traditional approaches and our own ability to cope with that, we’re way beyond that.
So, if I think of Aon’s Oxbotica partnership, they are driving these vehicles that we are working with in insuring, they’re producing ten terabytes of data a day, per car. I mean how do we deal with that? How do we consume that? How do we use that? How do we get insight from that? How do we understand what the risks are? Similarly, there are going to be 25 billion connected devices out there and those are already being used in the relevance of insurance. Again, if we think about wearables, if we think about motor telematics. If we think about health devices. So, where do we even store that information and how do we consolidate that information and make sense of that? The only way we are going be able to learn that is by engaging and using that and starting to apply that in very narrow focused ways. To get back to the question is using it in the context of managing big data sets but more importantly, translating that all the way to decision making and insight and augmenting decision making is right at the heart of AI.
What’s Aon's perspective on using AI?
We have AI, we use it; funnily enough, we apply some of most advanced technology or thinking to some of our simplest problems. Trying to match customers turns out to be really hard. So actually, we use AI to match customers. This is irony of it. We don’t use it in other places that much to be totally honest, because if I look at the problems we face, AI can solve a lot of issues, but frankly, I don’t need AI to solve the next problems that I have. So, we have a lot of large datasets, so getting them together and using them. So, it’s one of these things, to me it’s a little bit like a buzzword at the moment. I mean, absolutely, I guess if you’re trying to find really complicated patterns, to work out where the next flu pandemic is, you’re using real AI. Actually, for us, a lot of the challenge is: how do we get together our large data sets, so that we can then provide to the market, say, much better, consistent information around a portfolio of risks?
That is a much more real challenge that we face today to actually create new products and new ways of supporting client risk. These things are great, but there are 25 other problems which I probably wouldn’t solve with AI that sit over here. What is real for us, is technology and the change in technology is so quickly changing what we can do, the problem we’ve got is that we live in a world and look, I’m part of this world, that grew up in a technology set in the 90s. So, my mindset is wired to 90s technology and it takes eight months to change a dropdown box on a screen and it’s really hard to get us out of that mindset. To me, those are the real challenges. Some of the more disruptive technologies. Blockchain is another great technology, we’re interested and we’re looking at it, but I can see 25 other problems that we can probably innovate to solve, which maybe don’t need Blockchain or AI at this point.
Yeah. Nicolas, what’s your perspective on using AI?
Basically, we have two ways to use AI. For example, by using image and so on and so on will help us in managing the claims and also, beating fraud, for example. Because by comparing tons of data, and you have a lot of images about car damaged on the web and so on and you can learn a lot from it and then recognise when somebody is saying, “This is what happened.” Then you can compare that with the AI very rapidly to do some of the things that should be similar and they are not. So, you can better understand which type of claims you have and if it’s a fraud or if it’s not. So, these are the types of technology that we have. But we see also and the way we want and we are using AI, is to go to something more personalised and in context of the needs of insurance of our customers. Because if we want to go in the future into a product, which is in insurance, which is really contextualised, so you get that insurance when you need it and we have a product, which is exactly matching your needs, your way of life, your way of taking a risk in your life, to get to that point, you need to be able to gather a lot of information and clearly understand who is the customer that we have in front of us. This is what AI is going to help us to do by really being able to understand, with all the information we can gather, who you are and what’s your need and how much you are able to pay for that, which is another story. This is exactly where we want to go with AI and the technology is there to do it.
So tailoring products and risk selection and pricing?
And how far are you on the journey to achieving that?
I can’t say but we are very close to that; we are very close to being able to understand in your everyday life when do you need a product of insurance, when are you more likely to buy it and how much you are willing to spend on this one and what is the exact product that you need. Just to take a kind of example, you don’t want to get an insurance for practicing sport and so on and so on, when you don’t practice sports more than two times a month. So, you don’t want to pay for the time where you’re not practicing. Exactly the way that you don’t want to pay for a car insurance if you’re not using your car, unless on weekends. So, this is the type of things that with the AI, we are going to learn from the people, and try to be there when they need us. That’s it.
So, how do you gather that data?
That’s another good story and that’s the other way round. I mean, just think about who we are as an insurer. We are providing the service through a couple of touch points with our end customers. Whether it’s individual people, whether it’s companies. And you usually see your insurer three times a year, when you have to negotiate a new insurance and you have to negotiate the price. If you have a claim, you’re unlucky, you have a claim. So, that’s when you have a touch. And a year after, when you have to renegotiate the arrangement. You only have three touch points.
So, what we think is that we should have much more touch points; so, we think that an insurer should be there with you every day and we are not just there to discuss about price or to discuss about a claim, but we are there to tell you, “Hey, be careful, you are facing some risks. Be careful, I advise you to do this or not to do that.” And by finding ways to have those touch points, using technology, this is how we are also going to learn who you are and how you make decisions. So, by having those much more regular contacts, we are going to learn much more of you and we will be able to provide you with the right insurance, at the right moment. This is where we think we have to go and how we are going to gather the data.
And to build on that example and to give a real life… So, my previous company, Discovery; their focus was around the use of wearables. But their emphasis was around, how do you help your customers be healthier? Rather than the focus wasn’t around providing insurance and it being a transaction, it was around: how do you focus on those outcomes of health? So, for example, we’d be having touch points with customers around about three to four times a week, giving them little nudges around helping them understand your own health, so there’s an awareness piece. There’s then an incentive’s piece over: how do you encourage people to take those first steps? And then there’s a reinforcement piece about: how do you give little nudges along the way to keep encouraging those right behaviours?
What’s really surprising is the nudges don’t have to be particularly big to get quite a significant life in behaviour. But again, it’s got to be done in the best interests of the customer and it’s got to be something that the customer wants to do and doing it in a very transparent way. I think what’s really interesting and part of my role within XL Catlin, is how do you translate those sorts of ideas, in either personal lines or other sectors, and how do you translate that into a commercial and specialty environment. Again, if you think of the world of IT and sensors and industrial sensors, there is really strong read across to those types of environments.