VInsurance companies are making enormous financial efforts to meet today's customer demands while operating as cost-effectively as possible. But is this balancing act even possible and what are the real benefits of digital services?
In the age of digitalization, customer requirements have changed fundamentally. Thanks to Amazon and companies like it, customers today expect all services to be directly available and place particular value on speed. Claims departments are also trying to meet the demands of customers and are increasingly relying on modern technologies here.
Services that are perceived as modern by policyholders often involve reporting the claim using digital communication channels and receiving feedback on the course of action from the insurance company in the shortest possible time. Applications based on artificial intelligence (AI) can be helpful.Artificial intelligence is not new
The term "artificial intelligence" is ubiquitous, yet it still sounds elusive to many. AI is only the umbrella term for applications in which machines provide human-like intelligence services such as learning, judgment and problem-solving. The theories behind this are not new, but the technical possibilities for implementing these theories have only been available for a few years. One specific application of artificial intelligence is automated image recognition.
Notification of damage on the basis of managed notification claims report mask
The use of online services in claims is also becoming increasingly popular with policyholders. There are many ways to report damage digitally. For example, the claim can be reported through an app, link or the insurance company's customer portal. In all cases, policyholders access a guided claims screen. In the claim form, policyholders are also asked to upload photos of the damage to the vehicle. A modern damage system can use artificial intelligence to evaluate the photos and determine an amount of damage within a few hours.
But how does that work? An excursion into the workings of artificial intelligence.
AI is fundamentally built on self-learning algorithms. The quality of results improves continuously as new information or data sets are added. Such a process is called machine learning. Machine learning is again based on the functioning of a human brain. Thus, models are built that consist of artificial neural networks with different interconnected nodes and different levels. Within these networks, information is processed by positive or negative weighting and presented as a result. The learning processes of neural networks are called Deep Learning.
Difference to rule-based image processing
Rule-based image processing is based on hard-coded rules, such as the shape, number or position of objects. Image recognition-based AI learns similar to the way a human does, through image examples to identify objects by common features. Thus, a person recognizes an exterior mirror regardless of its location, size, shape, or color. Humans also recognize special exterior mirrors.
An AI system can achieve a recognition rate almost as high as a human through machine learning. For this purpose, it is sufficient to evaluate a sufficient number of photos of exterior mirrors of different types. This allows the AI to detect which objects belong to the "exterior mirror" type.
How does AI help speed up the claims process? Example:
The policyholder hit his garage wall and now has a crack in his rear bumper. The policyholder reports a fully comprehensive claim to his insurance company through an app, link or on the customer portal. Here he has to fill in a guided questionnaire and upload photos of his vehicle.
Now the insurance company's modern claims system, including the linked component of intelligent image recognition, comes into play. The receipt of the information in the claims system leads to a fully automated creation of a claim. The photos are forwarded to the image recognition software. This recognizes the vehicle part which has been damaged and is to be replaced on the photos on the basis of trained experience.
The data extracted from the damage report and the image recognition are then sent to a service provider, where they are compared with a database. This database contains extensive data on spare parts prices (rear bumper of the affected vehicle model), repair times (removal / painting / installation of the bumper) and the hourly billing rates in the policyholder's region.
In this way, a calculation of the expected repair costs can be determined within a very short time and sent to the policyholder.
In conclusion
This example shows that image recognition based on artificial intelligence and trained with sufficient data can be usefully integrated into a modern claims process. Insurance companies can significantly reduce claims processing times and save costs at the same time, and policyholders are pleased to receive fast and binding feedback on the reported claim.
Would you like to learn more about this topic through concrete applications? Then we recommend our free whitepaper "Use of AI in the insurance industry" which is available in German only.
