Auto Recycling World spoke to James Spears, Head of Automotive at Tractable, a tech company developing artificial intelligence for accident & disaster recovery, about their AI system and the benefits it can provide to the auto recycling industry.
An introduction to Tractable – from foundation to growth and the part it plays in assessing the categorisation of vehicles
Tractable was founded in 2014. Its Co-founders Alex Dalyac (CEO) and Razvan Ranca (CTO) were graduate students specialising in computer vision – using AI to understand what’s in an image. In recent years, computer vision has reached the level of human performance, a breakthrough moment in how the technology can be applied commercially.
In the future, Tractable will use it to help the world recover faster from major disasters, like hurricanes. But for now, it is focusing on using AI to accelerate how quickly people can recover their cars after an accident. Applying AI here makes sense: not just because it’s a trillion-dollar industry, but also because the claims process generates a large number of vehicle images – which is exactly what is needed to train an AI to understand damage.
Now, Tractable’s AI has processed hundreds of millions of images. And as a result of that learning, it can now understand the external condition of any vehicle as well as the repair operations that may be required – such as what parts need to be repaired or replaced, as well as any other complex work, such as paint blending in surrounding parts.
Tractable has raised $55 million in funding, and its AI is currently used by over 15 major insurance carriers across 12 countries – processing over $1 billion of claims. Now, it is also investigating other possible use cases – including with bodyshops and auto recyclers.
How does Tractable’s AI help to improve the categorisation of vehicles and speed up the process of salvaging vehicles?
James Spears: Tractable’s AI can make decisions just as well as a human assessor. It looks at damage pixel by pixel and can classify if that panel is damaged or not, or repairable or not. It generates a damage estimate, which is then passed on to the insurance professional to categorise. It’s an enabling tool that helps assessors speed up the visual assessment part of their work – and, of course, the AI doesn’t get tired, unlike a human!
With this ‘sped up’ process, what advantages does the technology bring?
James Spears: Speed, accuracy and service. It reduces the cycle time, performs to a consistent standard every time, and ultimately provides a better service for the customer.
Regarding categorisation (CoP), if a vehicle was to be miscategorised, is Tractable open to receive feedback from its users? Does the AI system learn from the information provided?
James Spears: The AI system is efficient when it has been fed the correct information. Although it can match some of the things that humans do, it cannot do all the things humans do. The best way to understand where to use it is as a tool that helps humans do their tasks better.
That said, the AI does learn – it improves with every single claim that is run through it. Over five years of entering millions and millions of cases into the system, it has reached the level of a human assessor.
Is Tractable aware of the issues regarding the categorisation of salvaged vehicles and vehicles falling through the gaps? If so, how can Tractable help create confidence in their AI solutions?
James Spears: It is up to humans to decide how they handle these decisions, but AI can help. There are limitations with the technology – for example; it’s impossible for an AI now to recognise what material a part is made from, and so it would be wrong to rely on its input there. But when it comes to recognising damage, AI can contribute effectively, relieve workloads and increase productivity. So really, it’s about deploying these tools where they can be most effective and helpful.
Where is the future when it comes to the relationship between AI solutions and vehicle recyclers? How can they work closely together, and how can it be beneficial to all when it comes to those involved in the process of vehicle salvage?
James Spears: There are three major future growth areas for Tractable:
- Triage. If a car is not repairable, our AI can help decide what happens to it next. Should it go to an auction or a direct buyer? Should it be sold for parts or for a rebuild?
- Non-damaged parts. The AI can make an inventory list of the undamaged parts, making the buying journey simpler and more accurate.
- Accelerating inventory management. The AI could match parts that it detects with needs elsewhere, meaning that as soon as – for example – an undamaged panel is located, it could be sold to a potential buyer – even before it’s been removed from a car. And this could happen automatically, enabled by AI.
How soon will you move into this area?
James Spears: We are actually in talks with numerous players in salvage, and we’re looking forward to announcing something soon. Watch this space!