“Artificial intelligence is either the best thing or worst thing to happen to humanity.”
With these words, noted physicist Stephen Hawking beautifully captured the divided opinions of humans on artificial intelligence (AI). Proponents see it as a great servant and opponents view it as a terrible master. Who’s right? Both!
Used responsibly, AI and data technologies can enable companies to make their operations more efficient as well as improve the customer experience. In insurance too, AI can be an advantage in several areas, from underwriting to claims handling to customer interaction and even personalization and simplification of products and services.
Within claims, fraud is one of the biggest bugbears of the insurance industry. This is where AI can benefit both insurers and customers by detecting and preventing fraudulent claims.
Imagine this: John* bought a vintage one-bedroom apartment at the heart of a bustling European city. Having achieved his lifelong dream, he furnished it lovingly with a designer sofa, pendant lamps, the works. However, when his mother took ill, John had to spend a lot of time at her home 20 kilometers away, often overnight. So he decided to rent his apartment out on days he was away. The online home-sharing platform was a blessing – easy, fast and equipped with an option to buy insurance for any damages by tenants.
John was lucky with tenants. They gave him no reasons to trigger a claim, always leaving his home in a good condition.
Across the hall, Joe* was also a lessor. An owner of two apartments in the city, Joe rented out one for short stays. On a warm evening, the two neighbors bonded over beer. John told Joe about the home-sharing platform, prompting the latter to sign up. But unlike John, Joe reported several damages by tenants over one year, filing insurance claims.
On the other side sat the insurance company. To offer a customer-friendly service, the insurer had a simple process in the beginning – claims under a certain amount were settled via a straight-through process and the payout was instantaneous.
The process benefited all lessors on the platform. But when Joe reported more incidents than usual, claims handlers visited his apartment to verify the legitimacy of the claims. While some were legit, some others were found to be fraudulent. Joe’s insurance contract was cancelled and he was barred from using the platform. However, he was not the only one to suffer the consequences of his actions.
Such events impact an insurer’s loss ratio and the company raised premiums also for honest customers such as John, who had little to do with Joe beyond a friendly beer.
* Names and events are fictitious. Any resemblance to actual persons, living or dead, or actual events is purely coincidental.
So how can insurers help John keep his premiums low? How can they protect themselves and customers from fraudulent behavior?
At the moment, this is mostly a manual job. Claims handlers find fraudulent cases, and over time and with experience, identify patterns. For example: It might be suspicious if a claim is submitted within the first two weeks of a new policy being issued. The knowledge of claims handlers is then built into IT systems identifying fraud.
While the experience of claims handlers will remain important, AI can help quickly and continuously look for fraudulent patterns in a very detailed manner. It can learn automatically from past fraudulent cases, apply the learnings immediately and get better and better in identifying and preventing fraud with time.
With a mechanism like that, it might be possible to prevent fraud right at the point of underwriting by identifying both inaccurate information and quotes that are at risk of future abuse.
In addition, AI could help cut the lifecycle of a claim by 20 percent by automating some processes. This will allow a claims handler to spend more time in helping the customer cope with an incident rather than on routine administrative tasks.
Leading the AI charge at Allianz is Gemma Garriga and her Global AI and Advanced Business Analytics team.
Their mission: using AI to make processes more efficient and offer more customer-friendly products and services. The team is building the Allianz AI network by connecting data scientists with business experts from across the group. Through collaboration and co-creation, the team is working not only towards finding scalable AI business solutions but also towards changing mindsets and attitudes towards digitalization.
For example, one of the Allianz entities is exploring deep learning technologies for car damage assessment. By analyzing millions of images of mangled cars, AI can link the severity of the damage to repair costs. This could assist in settling claims more accurately as well as assessing risks at the underwriting stage using internal and publicly-available external data.
However, it’s important to remember that humans must control machines and not vice versa. At Allianz, it’s not just data scientists and engineers but also experts in data privacy and regulations who work on AI adoption. Extensive tests are carried out on AI models to ensure they do not contain any biases and irregularities and pose no ethical risks.
The models must be ‘airtight’ before they go live. “Traditionally, the insurance industry has been slow to adopt technology. However, in the interest of customers, it’s no longer an option to ignore AI,” says Gemma Garriga. “What we need to remember and internalize are the ethical implications of AI. And responsible AI is the only way to go.”