How AI helps insurers defend against fraud
Insurers are increasingly relying on artificial intelligence in the fight against fraudsters. There are good reasons for this, because criminals are also becoming increasingly digital.
A rear-end collision between a Ferrari and a Lamborghini at four in the morning on the Rostock industrial estate. Both drivers are 18 years old, the amount of damage is several tens of thousands of euros. In such a case, all alarms go off for insurance employees in the prevention of fraud. Because the time, the place, the age of the people and the amount of damage don’t really match up. However, insurers cannot afford to manually check each individual case; they have to evaluate millions of claims each year. Therefore, insurers get digital help. An artificial intelligence (AI) pre-selects and evaluates if a case could be fraud or not. This offers advantages for honest policyholders and disadvantages for fraudsters.
According to the General Association of the German Insurance Industry (GDV), insurance fraud causes losses of around 5 billion euros per year. Insurers assume that every tenth claim is doubtful. GDV Deputy CEO Anja Käfer-Rohrbach says: “Doubtful does not automatically mean fraud in this case, but rather that the damage contains characteristics that are statistically rare.” she is actually cheating.
AI as filter
Repeat offenders and structured insurance fraud in particular have been putting pressure on insurers for years. This is one of the reasons why more and more insurers are turning to AI algorithms from companies like the Dutch company Friss. “Well more than half of German insurers already use such artificial intelligence algorithms,” estimates Käfer-Rohrbach.
The new technology brings clear benefits to insurers. “The biggest advantage of the new digital assistants is the possibility of being able to examine almost all claims for atypical factors directly when they are reported”, explains the expert. Factors that may be atypical include the duration of the insurance, the amount of damage, and the age of the insured. If the AI detects anomalies, the damage is sent to the fraud department, where specialized personnel verify the cases individually. The use of AI should act as a deterrent to scammers and occasional cheaters. “If the fraudster knows that all cases of damage will be checked, he increases the risk of being discovered and thinks twice before taking the risk,” explains Käfer-Rohrbach.
In addition, the use of AI is also advantageous for honest policyholders: not only that the insurance company has to settle fewer claims, and as a result, contributions do not collectively increase. Claiming a clear case can be regulated in a very short time, because the algorithm rejects clear cases much faster. As a result, policyholders receive their money faster than ever before.
Digital challenges for insurers
Insurers are not the only ones that are further developing their technology infrastructure. Scammers are now working with image processing programs and building damage that never really existed. Also, invoice totals are often supplemented with one or two digits. “Fraud will move more and more into the virtual space,” says Jochen Haug, head of the damage department on the Allianz Versicherung board of directors. “With the increased use of digital media, we assume that by 2030 one in five insurance fraud will take place virtually.”
Therefore, insurers need to stay on the ball and get up to speed in the digital arena. Support comes from the London company Photocert, for example. Its AI technology recognizes edited photos by examining GPS data and timestamps, and checking to see if the image has appeared elsewhere on the Internet. The Ico-Lux company from Jena, on the other hand, has specialized in the field of invoice forgery. AI verifies and detects false or edited invoices.
No matter how advanced AI and deep learning methods may be, insurance fraud prevention will not be fully automated in the future either. “AI by itself cannot detect insurance fraud, but only outliers,” says GDV’s Käfer-Rohrbach. “Ultimately, an insurance company employee will always verify the claim and decide if there is a possible attempted fraud.” However, in the cat-and-mouse game with fraudsters, the use of AI at least puts insurers back on an even keel. equal footing.
Do you already know ours? Newsletter “The Week” ? Every Friday in your mailbox, if you want. Here you can register