Q&A: AI Drives More Control in Estimating
April 13, 2020—Mitchell recently announced the early availability of its AI estimating solution, Mitchell Intelligent Estimating. The estimating platform is available for early adopters in the U.S., who are ready to integrate artificial intelligence into their collision repair claims workflow.
The intelligent estimating platform leverages artificial intelligence, data and a cloud-based estimating system to turn photos of a damaged car into estimate lines.
Artificial Intelligence (AI) today is similar to what cloud was 10 years ago, says Olivier Baudoux, senior vice president of global product strategy and artificial intelligence for auto physical damage at Mitchell. Baudoux joined Mitchell in 2016 to address disruptions facing the automotive industry such as connected cars, claims automation and platform advancements like the cloud.
"Although it is rapidly changing, I think most companies are still questioning the value AI can bring to their businesses," Baudoux says. "To me it’s not a question of whether AI brings value, but rather how AI can be better applied, and can co-exist with humans to address particular business problems."
ADAPT caught up with Baudoux to find more about how AI in the claims process can help solve pain points for repairers.
In your opinion, what pain points does the Mitchell Intelligent Estimating solution solve?
I think of it as the "Autopilot" feature for Tesla drivers like me. Now, would I trust the machine and fall asleep at the wheel? No way. It will take years before we get there. The Mitchell Intelligent Estimating solution is a paradigm shift in which you no longer have to write an estimate from scratch. Leveraging AI and computer vision in particular, the machine is capable of “scrutinizing” images and auto-populating all or a portion of an estimate.
There are two main benefits to the system. The sooner a repairer learns the claims automation process, the better he or she will learn and differentiate themselves in the market.
And, the system helps drive more control, consistency and influence over the estimating process. The machine will standardize how estimates are being written, like how the optimal parts are selected. Today, the quality of an estimate remains subjective and there are often significant differences between appraisals written by staff appraiser, independent estimators and repair facilities. AI will reduce the delta, and over time optimize the accuracy and cycle time while reducing the subjectivity within estimates.
How in-depth are the estimating lines on the photos?
If damage is not visible, estimate lines will not be written. For example, exterior panel damage is much less complex than internal structural damage.
The lines are a function of the complexity of the vehicle, the accident, and the maturity of the machine. So Mitchell has been very focused on predicting Repair versus Replace, for instance, but we’re rapidly adding more operations and elements such as R&I (Remove & Install), refinishing, overhaul or even being able to identify prior damage.
As we often say, the machine keeps learning. It is a work in progress, and the goal should not be to replace the human but rather to bring value where and when needed while allowing the workforce to focus on optimal use of their time around consumer service and other complex tasks.
At what point in the repair process is it best to use this tool, and why?
It is really for pre-repair and remains focused on initiating the estimate based on photos.
Today, it runs after photos have been captured, and begins to populate the estimate, after which the expert takes over. The expert might be an insurance or independent adjuster or in some workflows could be a repair planner at a collision facility.