Matching technology to the problem at hand
When Chief Technology Officer Evan Davies joined Solera in 2020, he had already been working with AI for many years. Experience had taught him that no one particular AI technology could solve every business problem, no matter how innovative or disruptive it might be. Evan knew that a combination of multiple technologies sourced in-house and from cloud vendors would be required. He was determined that Solera would effectively apply mature AI technologies to gain and maintain competitive advantages for the business. To his delight, he joined a team that had already figured out the best approach to the problem they had identified—how to use AI and ML to automate Solera’s existing automobile collision claim workflow.
Solera’s product team knew from talking with insurance companies over several years that they needed an automated claims process. Quite a few insurers had tried to use computer vision to automate the collision damage repair process. Marcos Malzone, VP of Product Management at Solera EMEA, explains, “Insurance companies had encountered a number of challenges in trying to commercialize computer vision solutions. They would do their research projects, and could usually build a working solution in-house, but they couldn’t scale. What we learned from this is the importance of building a productized solution in order to avoid failing as an AI project.”
Solera had focused on the most elegant application of AI to the workflow, which was to effectively identify vehicle damage. The initial damage assessment step was transformed into an AI-powered process, and the addition of ML leveraged the company’s huge existing database of claims images and repair information to offer precise method, cost, and time estimates for repairs. Equally important to the success of the solution was the choice not to complicate the process by changing the company’s tried and true backend systems. Davies says, “We wanted to solve a specific problem by applying AI to identify collision damage and then use our backend systems and machine learning to create a plan for how to repair that damage.”
Solera had built a previous version of an automated claims system that showed the promise of what a next-generation solution could be. The team’s original vision combined with the latest cloud and AI technologies would enable Solera to reimagine Qapter using AI and ML.Through extensive research, the Solera team had already advanced development to the point where they had eliminated several less successful approaches. All they needed was the right AI solution coupled with the latest cloud technologies to explore new ideas and upgrade Qapter. This next-generation version would streamline the estimation process for Solera customers and vehicle owners worldwide.
Unlocking possibilities for product development with Google Cloud
The Solera team were already sophisticated cloud technology users when they decided to look for an AI/ML solution that would integrate with a full suite of cutting-edge cloud technologies. Although the company hosts its own data lake in order to maintain contractual agreements with customers all over the world, the accident claim workflow was cloud based. The team knew that choosing the right technology vendor would be key to a successful outcome for the next-generation platform as well as new products in development.
After completing a thorough technology bake-off, Google Cloud’s AI/ML solutions proved to be more sophisticated, robust, and scalable than what other vendors could offer. For Solera, having best-in-class AI technologies that are tightly integrated with the entire Google Cloud portfolio was a decisive factor. These additional capabilities meant that Solera could take advantage of faster processing speed and sophisticated tools that complement its development focus. In short, Google Cloud could provide Solera with everything it needs from a single vendor.
Solera reaped the benefits of a single-source provider by leveraging products across the breadth of Google Cloud. Solera developers were pleased to discover just how quickly Google Cloud technology has evolved and how it offers a highly stable framework for faster, less complex deployment across the value chain. Starting with Cloud Vision for simple image processing, Solera uses the Vision API’s optical character recognition (OCR) to collect license plates and VIN numbers. TensorFlow helps to build custom algorithms and machine learning models for image recognition and extraction of vehicle data, allowing for collection of vehicle make and model, damage information, and parts required. In addition, Cloud GPUs and TPUs allow for accelerated processing of all data models, greatly exceeding the capabilities of traditional central and graphics processing units.
