Tech Superiority

1. Large and unique proprietary data sets

AI systems need data sets to train them. The larger the better. At Tractable, we’ve amassed arguably one of the largest, if not the largest, global data set of damaged car images of any company in our sector, which is a major factor to the technical superiority of our AI system. Put simply, our systems have been trained on 100s of millions of damaged car photos. 

Our massive data set means we can train our AI systems to be more robust, accurate, useful and trusted by our customers and their customers. It’s an immense and exclusive global resource that is continually growing and informing the accuracy and consistency of our damage assessment reports and repair estimates.

It means that when we integrate our tech into customer systems, we have a uniquely strong head start before we calibrate our system even further to customer and country repair standards.

2. Driven by research science 

• We employ 40 researchers with over 400 years of research science expertise and 40 patents between them. 

• Strong academics (Cambs / Oxf. / Torr Vision Group) combined with industry – Amazon [need more]

• As a tech company, we are able to attract and retain the best talent to make sure we compliment our customers needs

• Excel at applying novel academic theory to real world scenarios

3. Thinks like a human expert – human expertise 

While high-quality research is vital, what is critical to the success of our AI system is close collaboration with industry experts. By working with highly-experienced automotive and property experts, we can transfer their years of damage assessment knowledge into our AI system, and use their expertise to continuously train and calibrate our systems. In effect, our AI models are replicas of their decision making process and professional judgement at their very best. So you can be sure of consistently accurate and confident results that aren’t subject to human variables.

System is a reflection of years and years of industry expertise

Expert decision making models

Continuous improvement of our AI models with strong QC process

4. Engineered to perform at scale

If our research teams bring ideas from the lab and prove they can work in the real world, then our engineering teams take that prototypical code / those algorithms and harden them up to function in a production ready environment. Scaling AI systems to continuously deliver results with minimal outages for some of the world’s largest and – rightly – most demanding customers is hard. The systems need to be commercially viable, etc… [not sure this is right but I gave it a go] 

Models = owned by researchers. Engineering is responsible for taking the raw data from the models and turning those into estimates (for example), real world applications. How do we easily connect to insurance companies? Try to be as flexible as possible to ingest the data… systems to absorb any data, JSON, Zip, PDFs, Fax, images – Engineering connections to the model is a strength 

5. Integrated operations

General Omar Bradley famously said “Amateurs talk strategy. Professionals talk logistics”. With that in mind, the glue that holds our Research and Engineering teams together is Operations. It’s a point that is often unacknowledged by AI companies. Our operations experts are critical to labelling data accurately, testing system fallibility with engineering teams, providing continuous damage assessment expertise to help research teams optimise models, QA’ing estimates at scale to ensure accurate and high calibre output no matter what the demand. Without a strong and integrated operations unit, no amount of data or scientific brain power will get the results our customers need. [not sure this is right but I gave it a go]

Backbone of the models – well managed and staffed. SME co-callibration + retro QC 

6. Open ecosystem approach

Increasingly, our customers are choosing to move away from the ‘walled-garden’ that many legacy estimatics platforms – for example – offer. Tying oneself into a restricted ecosystem makes little sense in a future where competitive advantage can be found in being nimble enough to adopt cutting edge technology, often offered through open ecosystems and APIs. We have built our Applied AI Engine with this system in mind. So while every integration needs to be calibrated to specific business standards, we can technically integrate into any system. It means that our customers can remain adaptive and nimble without worrying about painful processes [not sure if this hits the right notes or not… might be overly simplistic and optimistic!.] 

7. Systems that are built to be tested and optimized 

Often testing an AI system will come with unrealistic parameters. Caveats on how many API requests can be made in a given time period, unrealistic timelines for estimates to come back or a certain ‘type’ of image submission. All these are indications of a brittle AI system that can’t stand up to the demands of everyday reality. At Tractable, we invite all potential customers to test out AI systems without the stage management. 

Our confidence comes not only from our unique data and the fact we test our AI models with real customer data before bringing them into production, but also the fact we make regular deep dives into our systems with our own industry experts and partners to unearth any weak points. 

We also run extensive User Acceptance Testing (UAT). We do this working hand-in-hand with our customers to measure the AI performance in real-world situations, and to quickly iterate on any flaws.

By rigorously building, training and testing our applied AI in countless real and challenging conditions, over and over again, it gives us the confidence to invite anyone to test our products using their own images in a live situation.

8. Working towards a trusted AI: robust, credible and fair

Because our applied AI systems are built and tested in live conditions, they are able to perform to human levels of accuracy in terms of visual recognition of damage. 

But for us, it’s just the first step in our journey. 

We’re working towards building a trusted AI that is consistently robust, credible and fair. 

• Robust so that our AI system doesn’t make human mistakes. 

• Credible so that our AI system can present and explain results in a way that human experts would understand them. 

• Fair so that our AI system can produce consistent results across different populations, ensuring there is no bias in the treatment of data. 

We’re imagining a superior AI that’s the basis of a trusted standard for managing transactions. 

A trusted system that can leverage AI’s speed, accuracy, scalability and 24/7 consistency to mediate all kinds of decisions – from repairing a car or selling furniture, to building a house or enabling circular economies.

It’s the future of our journey to build not just a superior AI, but one that can establish new levels of trust – enabling more robust, credible and fair solutions.

Experience applied AI for yourself 

As researchers and engineers, we always say that, when it comes to AI, you can’t know how well it works until you try it for yourself. In the real world. 

It’s why we invite you to a live AI demo. The best way to experience Tractable applied AI in action. 

Book a live demo