The insurance claim processing world has had enough of AI buzzwords. This is how the leaders do it.
The insurance industry, and auto insurers in particular, have been hearing about how AI will transform their business for the best part of the last decade.
Indeed there are clear benefits for employing image analysis technologies coupled with data science in claim processing. Examples include:
Triaging claims triaging - correctly identifying damage magnitude and vehicle condition, to classify the claim as ‘total loss’, ‘repair’ or ‘quick settlement’.
Appraising vehicles more quickly - saving the travel and human costs of the appraiser by writing the estimate based on photos.
Approving supplements - many cases include these additional items, making the claim more costly for the carrier.
Unfortunately, ‘AI’ has been excessively used in the claim context, causing fatigue and at times reducing the level of trust towards technology companies. The industry has been hearing that a touch-less, end-to-end automated estimate was available here and now - it’s just a question of having a good tech team and the right amount of data.
Reality is more complex than that.
While AI can certainly improve the claim experience, making a decision on repair / replace / total loss is still subject to one’s interest. An insurance adjuster may look at a fender bender and order a simple paint job, while the customer or bodyshop may agree that exactly the same warrants a replacement. The role of AI should therefore focus on enabling tools to make these decisions more consistent and transparent, rather than pretending to remove humans from the equation altogether.
Guiding principles for adopting AI within motor claims
Through our work with leading insurers wishing to achieve efficiency and better customer service, several principles have emerged for a progressive, tech-enabled strategy:
First, good capture of images is the key to the claim process - to truly enable appraisals driven by AI, it is important to get enough data from the vehicle as close as possible to the time of incident. A single image can almost mean anything if it’s taken from the wrong angle, at low resolution or in a dark spot, and even the right analysis of damage doesn’t mean much without knowing about the vehicle itself. This is why we’ve enabled ‘30 second video capture’ that allows the customer to quickly capture dozens of viewpoints and actually put the damage in context. Our algorithms can then analyze these viewpoints like a human inspector would, increasing the confidence in the appraisal and enabling an evidence-based discussion between adjuster, insurer and repairer.
Critically, rather than uploading a full video, it is only uploading the relevant data - ensuring quality and data consumption efficiency
“A single image can mean almost anything if it’s taken from the wrong angle, at low resolution or from a dark spot״
Is this a dent, scratch or just mud? how does this relate to the rest of the vehicle?
“Instead, opt for a multi-angle view putting the damage in context"
Here, AI is scanning every pixel thousands of times as damage is viewed from dozens of angles - exposing paint damage to the front bumper but no structural impact
Second, keep it simple for the user - let’s face it, the last thing a customer wants to do after an accident is serve as our photographer, but their cooperation is key to enable faster claim processing. We have therefore invested heavily in an intuitive flow that can be completed quickly and can help provide quality assurance and some fraud detection.
Third, optimal business value is only achieved when AI is well embedded into operational processes. We hear a few players asking their AI partners to ‘simply’ analyse images uploaded onto their app, and it is entirely possible to do from a technical standpoint. However the true value in creating real estimates is by adopting the AI recommendation as part of an efficient process which takes into account various commercial data points and analysis of the entire vehicle (see first point).
Enabling ‘smart repair’ within 24 hours
In one specific use case, we’re working with a leading auto insurer who is rolling out Ravin’s 30 second scan functionality through their mobile app. Claimants can submit a claim alongside a scan and our AI helps triage the claim into ‘smart repair’ or ‘traditional repair’ categories, based on an assessment of the vehicle’s overall condition and its recent the specific damage. This saves a trip to the body shop for an in-person assessment, fast-tracks the process for claimants whose damage falls into the ‘smart repair’ category and reduces the administrative burden on claims teams.
As described above, this exceptional customer experience is only possible through collaboration between Ravin’s AI, the customers, the insurer and the body shops based on transparent analysis.