Tomorrow’s technology today

Unlocking AI’s vast potential for real-world impact.

We’re one of the few companies applying AI at global scale. What makes us different?

With rapid advances in computer vision and deep learning, AI research is on the brink of unlocking AI’s tremendous potential. But so far, real-world applications of that potential have been thin on the ground. According to MIT Technology Review, 40% of AI companies don’t actually have operational AI to apply. But Tractable’s history and growth has been built on applied research from the ground up, and our AI team works closely with industry experts in repairs and insurance to forge real solutions to real problems. For us, AI isn’t a dream of the future, but a force to harness right now.

How does it work?

  1. Trained on millions of detail-rich images, our AI can undertake a pixel-by-pixel assessment of a photo of a car or home. The AI can then classify the condition or the amount of damage, and provide a detailed estimate. 

  2. All estimates include a certainty score, depending on part visibility, photo conditions, damage severity and more. The claim can be directed to a human to review, to maintain the highest levels of accuracy. In other cases the estimate can continue through the claims process and speed up payouts.

  3. The transformation doesn’t stop at customer photos. Our AI can also identify parts that can be salvaged from old or junked cars, for example, and help repair shops source the exact parts they need – improving efficiency and recycling across the whole industries.

Request demo

A snapshot of our AI in action

What makes our AI so unique?

Accelerating the future

What sets our AI apart is its real-world robustness, scalability, and flexibility – making it easy to integrate with today’s tech stack – and the inclusion of meta-data that can help us accelerate AI’s industry-transforming potential.

Real-world robustness

Our AI is built for the real world - robust enough to take data that is inevitably noisy and fragmented – and detect subtlety of damage, to a high degree. Every working function of our AI is built to accelerate without sacrificing accuracy, and can be calibrated to align with client systems and maintain a benchmark of consistent output.

AI able to scale

Multiplied by the enormous number of vehicles and properties that get damaged every day across the globe, the efficiency gains are unprecedented. Our automotive AI products, for example, now offer coverage of all major external auto panels and parts by market. As we collaborate with our clients on business needs, our AI benefits from even more real-world data and continuously improves.

Adaptability and open API

Integrating AI technology with existing tech stacks and processes enables systems to ‘speak’ to each other. Through an open API approach, our tools are built to integrate with and work alongside our customers’ existing technology, and to work globally across different geographies and repair standards.

Vision and meta-vision

Visual assessment is our AI’s main focus, but we are extending this to include fine-grained meta-data for key non-visual information. For example, our AI can include metrics like speed at the time of a crash as an extra layer to assess damage, or aerial data to understand the extent of damage to a roof after a storm.

Our collaborative, experimental approach

We work in focused product teams consisting of software engineers, machine learning researchers, data scientists and product managers. These teams work alongside industry experts in, for example insurance or bodyshop repair, harnessing hundreds of years of real-world knowledge into real algorithmic decision-making. Each product we build is the hard work of a diverse team with the combined skills and domain-level knowledge needed to bring AI out of the lab and into real life. Thanks to our fast, iterative approach, and with feedback from our customers and collaborators, we’re able to bring new ideas from concept to production in a matter of months – helping to solve some of AI’s most pressing problems along the way.

Solving AI’s toughest challenges

Proprietary tech you can trust

  • Growing number of patents.
  • Proprietary inbuilt data protection, applied to all imagery, so all information used to model the AI remains anonymous and users are never at risk.

Research publications

Learn about GaLeNet: Multimodal Learning for Disaster Prediction, Management and Relief

We’re looking for you

Apply yourself to problems that matter, in a fast-moving AI company that's changing lives

Explore careers

Further reading