Privacy Policy
more

Norval Scott Tractable Cloud Expo Europe 2020

Norval Scott Tractable Cloud Expo Europe 2020

ANNA FLOCKETT [00:00:09] Hello. We are coming to you live from London Tech Show. I’m joined now by NorvAl Scott from Tractable, Head of Communications. Norval, nice to have you here. Thank you for joining.

NORVAL SCOTT [00:00:19] Thanks for having me.

ANNA FLOCKETT [00:00:20] And so you work for Tractable, a start-up that works in AI for accident and disaster recovery. And could you tell me a little bit more about the start-up and what you guys do?

NORVAL SCOTT [00:00:31] Yeah, sure. So Tractable has been around for about five years. So we use AI. Specifically we use computer vision and we use that to help people around the world recover from accidents and disasters. You think about accidents, disasters. You know, hurricanes, millions and millions of people around the world are affected by thess every year. They cause a huge amount of damage and that creates a huge amount of insurance claims. Recovery process for any disaster always starts with a visual assessment. So you need someone to go and look at the damage, understand what’s happened, and then you can start the recovery process and recover people’s livelihoods. We think you can speed all of that up by using AI.

ANNA FLOCKETT [00:01:05] Amazing. And where did the idea come from?

NORVAL SCOTT [00:01:08] Sure, so about five years ago, there’s a thing called the Stamford Project in California, which is basically to understand can a computer understand what’s in an image as well as a human can. So if if a computer sees a picture of a cat, can it tell that it’s a cat. So about five years ago, computers got to a stage where they could actually recognise what’s in an image as well as a human being. Right. So we are then applying exactly that to damage. And specifically, we’re looking at cars. So we’re getting an AI to understand, if a car’s been damaged from an accident. How much has it been damaged. The extent of what happens next, how much does it cost to fix it and how quickly can we get back to the to the driver.

ANNA FLOCKETT [00:01:47] And so explain to me a little bit about the journey from the idea it started up to where you guys are now in the last five years.

NORVAL SCOTT [00:01:55] So the thing about Tractable I think so a lot of AI companies start as very research focussed idea. That’s not what Tractable was. We wanted to take Tractable and apply that commercially to actually make a difference for big companies and also for people at the end, so people have actually been affected by disasters. So about two years into the Tractables journey, we started working with customers. We started working with the Ageas, which is one of the biggest insurers in the world to actually put this into practise. If you fast forward to where we are today, we now have about fifteen major companies around the world actually all using Tractable to accelerate claims. We’re in 10 countries, 10 different markets. We’ve helped hundreds of thousands of car claims and insurance claims. We’ve actually made a difference to hundreds of thousands of people’s lives by accelerating how quickly they get their cars back or get a decision on what happens after an accident.

ANNA FLOCKETT [00:02:47] And so obviously with disasters, they happen more regularly in some countries. So is this more relevant and a bigger market in certain countries where disasters are more likely to happen?

NORVAL SCOTT [00:02:59] It’s great question. So in terms like the market, the reason why become for auto first is because cars are everywhere. They get damaged all the time. And the thing we need to train an AI is lots and lots of photos. So it’s a perfect first use case for what transport is doing. But down the line, there’s nothing to stop you applying our computer vision technology to anything that really requires visual assessment. So that could be hurricane damage or property damage or, you know, the extent to which people’s roofs have been affected after a big disaster or flooding. So you could basically apply that to anywhere that has had a sizeable problem and speed up the response. The problem with disasters really is one of manpower. So if you imagine that, say, 100000 homes have been affected by a disaster, the problem is you don’t have the assessors to go and see all those houses at the same time. You know, it’s going to be a one by one process, which means you’re gonna have some that’s going to take weeks and months to actually be assessed. With AI, you could do them all very quickly.

ANNA FLOCKETT [00:03:58] Perfect. And what technology is used in your products?

NORVAL SCOTT [00:04:01] So we use computer vision. So that’s the main technology that’s used. So we use that to assess the damage that’s happened to a vehicle. We’ve trained the computer vision on millions and millions and millions of images of car damage. They’re all from opt-in partners. And the AI is now at standard where it’s as it can appraise damage as a human can and understand what’s happened to a car. It’s like an expert has been trained on millions and millions of cases.

ANNA FLOCKETT [00:04:29] Amazing. And how important do you think AI is in our current ecosystem?

NORVAL SCOTT [00:04:34] I think in the specific area that we’re in, the entire industry sees that AI is coming. It understands that, you know, AI is going to make a difference. What it doesn’t actually fully understand, is that AI is already making a difference to some companies now. It’s actually here. Companies are already changing the way that they actually carry out, you know, important, expensive tasks that really make a difference for their customers now. And that’s really what the what the sector needs to understand. It’s not enough to kind of think I need to get my head around this because it’s coming in six, seven years time. It’s we’re doing it now.

ANNA FLOCKETT [00:05:08] And how far do you think AI has come in the past few years?

NORVAL SCOTT [00:05:12] Well, if you think about our specific use case. So five years ago, this wouldn’t have been possible. It literally was that specific moment where a computer could understand what’s in  an image that it is now. That’s where we are, so today, you know, amazing. Six years ago, completely impossible. And that’s the that’s the scale of the the pace of change in this industry,.

ANNA FLOCKETT [00:05:33] Of course yeah. And what part you’ve touched on this briefly, but what part do you see AI playing in our future. What else can we expect from this amazing technology?

NORVAL SCOTT [00:05:43] Sure. I mean, I think if you if I bring it back to where we could apply it for Tractable, I mean, so we’re doing it for a moment for auto claims. But thers’s so many other aspects of that that you could also apply to, if you think of just for example, when you hire a car, what’s the first thing you do is go around and look, see if there’s any dents. And at the end, when you bring the car back, somebody also looks and see if any dents. That could be done with an AI. My my father’s a truck driver. The first thing he does is when he gets in his truck every day is has someone hit? Is it okay? Again, there’s an AI application for all of those kind of things. You know so you can change anything that really needs the visual assessment. That’s certainly possible to use AI to do that. The only question really is how do you train it quickly enough to get it to that point?

ANNA FLOCKETT [00:06:24] And have you guys faced any major challenges when it’s come to technology development in Tractable?

NORVAL SCOTT [00:06:30] Sure. I mean, it’s not you know, these these things aren’t easy necessarily, you know. So one big tactical challenge for Tractable is really how do you actually feed all the images into the AI in a quick enough way to make us understand. We’ve managed to do a batch kind of processing where you can kind of understand better, understands if there’s a car that’s this kind of make and model that it can do a lot of those kind of photos very quickly. So that speeds up the process. Another very difficult technical challenge is for an AI if you have six photos of the car like front backsides, how to actually make a coherent model of what that car looks like by joining all the photos up. That’s really difficult to do is something we’ve managed to achieve. But it’s those kind of odd technical challenges that like you have to beat them to actually make the system work.

ANNA FLOCKETT [00:07:22] And and I believe you guys have recently been through a funding round. How did that go?

NORVAL SCOTT [00:07:27] So, yeah, it’s very substantial funding rounds. We raised 25 million dollars from Georgian Partners, which is the leading North American VC. The amount we’ve raised makes us the largest insurer tech start up in the UK. I’m sure that won’t be the case for long as many companies in this space. But you know, it’s a nice it’s a nice thing to hang our hat on for now. But I think what it means is that there’s some of the biggest and most educated VCs in AI in the world have seen our solution and have gone, this has a real chance of success and of scaling. And they’ve seen what we’re doing with customers already. As I say, we work with fifteen, fifteen major companies, ten different markets. We’re already making a difference. They’re seeing that kind of traction and they’re saying, okay, we believe this could be the future.

NORVAL SCOTT [00:08:11] And what have you guys got plans for the future? What is next for Tractable?

ANNA FLOCKETT [00:08:15] So obviously we’ve raised money. So we are going to be expanding into new markets. So we’re looking into countries in Europe that we don’t already have a presence. Places like Southeast Asia, Latin America, we have an office of 10 people in Japan. Very unusual for a UK start up with the Japanese appetite for our product has been has been very good. So I’m sure we will be expanding more, more in Asia and also in in North America. We would also say we have the tech lead in in this specific area of AI and applying computer vision to to damage. So we want to hire more people, make sure that we’re maintaining and extending that lead there. And I think in terms of the specific news we’ve got coming up next, we’re about to do a major announcement about how a U.K. insurer actually uses our AI at the very start of the accident process. So when you very first report an accident, so we’ll be announcing that very shortly.

ANNA FLOCKETT [00:09:07] Amazing, we will look out for that. Well, thank you for chatting to us Norval. And thank you for listening.