Bitesize InsurTech: Tractable

November 5, 2016 Greg Brown

Imagine you’ve just had a car accident. You log onto your insurer’s app, take a few photos of the damage and press send. Twenty minutes later a recovery vehicle arrives, you are taken home and your car is taken to the bodyshop. Your claim is approved and the bodyshop gets paid. The whole claims process is done without human intervention and (almost) instantly.

This is the kind of proposition that Tractable wants to enable.

Tractable is a startup that uses AI and machine learning to extract meaning and insight from photographs in claims process. Their ambition is to do this in fields where highly trained human experts are currently required, for example claims estimation, medical image or crime scene evidence analysis. They have chosen claims as their starting point given the size of the market and available training data. (See our recent introduction to AI for an explanation of training data.)

Tractable are able to validate repair quotes by comparing photographic evidence with the quote. Any overestimates are flagged for review e.g. unnecessary part replacement. The system is live in production with an insurer and as of August this year they are correctly flagging up to 15% of estimates. Tractable state this demonstrates better than human performance levels. In addition to reducing the cost of the claims this automation also reduces the time to return an estimate, thus cutting claim handling costs and improving end customer experience.

CEO and Co-founder Alex Dalyac is confident: “In the next 12 months we will be able do to full repair estimation without the need for human intervention. This could have a major impact on the COR of insurers”.

Tractable are soon to announce a large deal with an insurer. Watch the Oxbow Partners twitter feed for more news.