By using this web site you accept our use of cookies.

More information about cookies

Computer Vision for Auto-Insurance

Faster processing of claims; reduced cost & time of investigation

One of the popular applications of 'Computer Vision' has been in enabling the Autonomous Cars. The ability to process a video stream and continually extract information & context from it, is the cornerstone of self-driving. However, there is another under-appreciated application of Computer Vision (CV) in Automobile sector.

'Claims Processing' of Auto insurances is a worthy candidate to be disrupted by Computer Vision Algorithms. In this post, let's understand a few pain-points in claims processing and explore how AI/ML can solve them.

Claims Handling – Current Scenario

A claims process is initiated when the claimant submits a First Notification of Loss (FNOL) application. The insurer assigns each FNOL to a 'claims adjuster' who will inspect the vehicle, investigate the damage and causes leading to it. Based on the investigation, he will determine 'a fair amount' for settlement. Clearly, there is a significant investment of time, effort and resources in evaluating each claim.

For an insurer, this workflow must be repeated a hundred times, every week, and round the year. There is a clear case for automating the preliminary evaluation of FNOLs.

Motor Insurance Claims - USA (for 2017)

  • 1.1% of people with liability insurance had a bodily injury liability claim 
  • 4 % of those with liability insurance had a property damage liability claim
  • 6.2% of collision insurance policyholders had a claim, while 2.9% of people with comprehensive coverage had a claim 
  • The average auto liability claim - property damage was $3,638; bodily injury was $15,270
  • The average collision claim was $3,425
  • The average comprehensive claim was $1,817


The opportunity

Here is what makes the things interesting. In our conversations with multiple Auto-Insurers, we learnt that over 60% of the claims eventually get categorized ‘low-value’ and/or ‘low-complexity’. The claims involving damaged mirrors, bumpers, grills, fenders, hood, headlights, tail-lights, body-dents and scratches fall under ‘low-value’ category. The ‘low-complexity’ cases refer to applications where investigation is straightforward, but the payout may involve a higher cost-value. An example could be ‘cracked windshield’.  

In both cases, an in-person investigation is expensive and may not even be necessary. Oftentimes, it can introduce further delay (due to unavailability of adjusters) to the workflow causing customer displeasure.

The New-Normal (optimal solution)

Imagine an AR-powered Mobile App that will help the claimant record and upload a brief video of his vehicle, highlighting the damaged areas. It will factor-in the adequate lighting, focus, and ensure capturing the vehicle’s license plate and VIN number.

Once the video is uploaded, the computer vision algorithm will process it to first analyze the make and model of the car and then determine the external/visible damage to the car. 

Today the Computer Vision algorithms are powerful enough to,

  • Identify the external damages to the car
  • Assess the severity of the damage (classify as minor/moderate/severe)
  • Make a Replace-or-Repair decision   

Based on the assessment by the system, 

  • In case of minor to moderate damage, the system recommends the parts that needs replacement and even compute the cost of spares and labor.
  • In case of severe damage, the algorithm will update the assessment details to the FNOL application and forward the case to an appropriate ‘claims-adjuster’. 

A Win-Win Scenario

The opportunity for insurers is obvious,

  • Computer Vision algorithms are today, mature enough to function as the first line of ‘Claims Processing’. They can classify and pre-evaluate the FNOLs efficiently. They are fast, available all the time and are scalable. Their reliability is improving by the day.
  • Computer Vision driven automation can help ‘claim-adjusters’ focus their valuable time & attention on cases with complex scenarios and internal damages.
  • The system can facilitate the ordering of parts and schedule the visit to the garage. This will reduce frictions in the logistics and inventory management. 

This approach is a win-win for both customers & insurers. Customers get ‘quick response and resolution’ while insurers ‘improve CSAT rating’ and ‘reduce the cost of operations’ significantly. Isn’t that cool? 

Recent Trends

Did you know? A recent study by J.D Power found that 42% of the claimants (of Auto Insurance) use mobile apps to submit photos & videos of their damaged vehicle. Furthermore, when insurers use the photos and videos submitted by claimants for the evaluation, the overall customer satisfaction went up significantly. Clearly, the claimants believe in sharing the damage details of their vehicle first-hand.  

Now, that’s good right?  

Would you like to share some interesting perspectives on the topic?
Or perhaps, explore this solution? Let us know.


AUTHOR: Jayanth Jagadeesh
He is a part of EVRY India and has years of experience in crafting technology (Software & IoT) solutions for Automobile sector.