Eloy’s accident manager – FNOL in under 3 minutes

There is a lot of buzz around the use of artificial intelligence and machine learning to rapidly identify damage to cars in order to help insurers process claims. Applications include image recognition technology and onboard-sensors that can detect if a car has been in an accident.

These innovations can add value, but they also bring a number of use cases that are sub-optimal, such as false positives for accidents. We also believe that although helpful in some situations instant answers are not always necessarily the right answer, particularly when the human thought process may be slower.

Woman Standing By Crashed Car

Eloy’s accident manager helps you generate a first notice of loss statement

This focus on instant answers, fuelled by the power of automated data processing, isn’t unique to the automotive and insurance industries. Humans can get hooked on instant gratification but evidence is starting to suggest it isn’t healthy.

Finance has its own examples too: home buyers don’t need instant mortgage approval if it will take 3, 6 or 12 months to find the right house. Mortgage sellers want the feature to acquire customers quickly but there is no guarantee that customers won’t look for quotes elsewhere. 

The focus on the automated processing of data misses out on a large number of important use cases because product innovators avoid dealing with complex human decision making. This human-in-the-loop challenge is one that is growing across many sectors. When we look at autonomous cars, how will the automated processes react when humans make unpredictable decisions on the road? How does an autonomous car decide where to park whilst a human decides which venue they want to visit first?

Understanding car accidents has  similar challenges. Knowing an accident has happened is important but the severity, the condition of the casualties, and the road layout are key pieces of information that can be subjective because they need human input.

So Eloy’s focus is on building quick and simple human-controlled processes within an app that can notify the relevant services and provide critical human generated data. From here, automated processing can then do its magic.

Generating car accident reports

Inside the Eloy app we have an accident manager tool – we call it our Big Red Button. Within 3 minutes, we want you to be able to:

  • If necessary, call emergency services with your location easily displayed (I am in danger)
  • Alert your insurer that you have been in an accident (I think I will need to make a claim and send roadside recovery)
  • Alert roadside recovery you need assistance (I may need a tow)
  • Tell your insurer that you need a replacement car or a taxi home (I need transport)
  • Collect at-the-scene photos and witness statements with GPS and time stamps (we can’t have cameras and sensors everywhere we all carry phones)
  • Share this report with others instantaneously so they can edit (multiple human input to help with more accurate recall)
  • Generate official first notice of loss statements for insurers (human-in-the-loop data sent for automated processing)
Loss Adjuster

Providing a seamless FNOL service is key to delivering a positive claims process to customers. 

We see this as having significantly more benefit than rapid quotes to fix cars – you don’t need that information as quickly as you think. What you want quickly after an accident is:

  • Emergency services if you need them
  • Roadside recovery to fix your car if you need it
  • A replacement car or a taxi so you can get home

Taking this a little further, the cost of fixing your car will also fall if good quality information is provided to the insurer as quickly as possible. Any AI/ML that estimates cost first misses the vital point: whoever fixes the car may have a different price structure. And this is only fair – we live in a world of supply and demand and variance in prices and car mechanics are no different.

The result is that the time delay between the accident and the accident being reported should be an input into the AI/ML model. The human provided data, including photos and the health of the casualties, is just as important to the incident report as what an automated sensor can provide.

Faster first notice of loss

After launching the app, users access the Insurance section, and then tap our Big Red Button, which starts the accident manager process. Pressing this button immediately signals to Eloy that the driver is in trouble due to in-app event tracking.

Our back-end event tracking is now trying to determine what automated processing should be triggered but insufficient human inputs have been provided. Further information is required to confirm what is required, ie:

  • Send emergency services
  • Send tow truck and replacement car or taxi
  • Send roadside help (breakdown recovery service) for assessment or fix (such as a new wheel)
  • No action required yet

We rule out the first option of emergency services by asking the user. In some cases it is required by law that you inform the police when you have been in an accident such as if the road is blocked, a crime has been committed, or there are casualties. Other times it will be subjective depending on, for example, the severity of the accident, or how safe those involved feel.

After ruling out emergency services, we need to understand the car damage, the scene, and the location. Being quick here is very important. If a driver can see that they will be able to drive away, a quick collection of data and exchange of content and insurance details is sufficient.

To determine whether a car will need towing or roadside help will be sufficient, photos and human information can help an expert make a decision or feed into an automated process. Every minute the user takes to send this information across will be an extra minute they will need to wait by the roadside.

We have got the collection of this data down to 3 minutes and users can edit the information after the event or share it. For example, you can see in the video below that users have the option to email the report to themselves. In the future we plan to offer users the ability to send the report directly to their insurer which means that the FNOL will be even faster.

More persistence

As the Eloy app has multiple users that can edit a car’s details and supply information to the platform, we needed to cater for when some users are not using the app and when internet connectivity is not available. 

This is referred to as database persistence. Our case was more complex as we want to store key information locally on the device (such as Accident Manager) and upload only when users are on Wi-Fi (accident photos can be large). This needed some tweaking to a standard Firebase solution. 

Our Big Red Button can gather accident data in less than 3 minutes

We see 3 immediate benefits:

  • We can provide information that enables the correct roadside response within 3 minutes. This is less time waiting by the roadside for the drivers and passengers.
  • We can help insurers get to scenes quicker and take the vehicles into their claims network. This lowers the cost of claims for the insurance industry.
  • The accident scene information can be sent to the National Incident Liaison Officer (NILO) and estimates of the expected traffic and road diversions can be enhanced.

More stories

Road traffic as a software to help with coordination

30 Reasons For Implementing EV Battery Swapping

Eloy And The World’s Oldest Connected Car Event

Ordnance Survey Map & Hack: Day 2

Ordnance Survey Map & Hack: Day 1

What can we learn from the panic at the petrol pump