Nicolas Michellod is a senior analyst at Celent with over fifteen years of experience in the financial industry. His research focuses on insurance markets and trends, including predictive analytics and data technology providers. He presented this story at the FRAUDtalks 2018 conference in Amsterdam.
Fraudulent claims are not a new problem to the insurance industry, but recent technological developments do offer new solutions to this problem. One of the most radical innovations is the use of artificial intelligence (AI). A recent survey among insurance companies revealed that over 75% of them are investing in AI, compared to only one-third the year before. Two-thirds of the surveyed companies expect the highest impact of AI to be in Claims.
AI opens up exciting possibilities to enhance the process for handling claims and increase fraud detection. But, as with every new technology, the organization has to create optimal conditions. The following four factors contribute to a favorable organizational climate for successfully implementing AI technology:
Data – integrate!
Insurance companies have access to enormous amounts of data from a wide variety of sources. From online sources such as social media, where people share their personal information on an unprecedented scale, to the internal data that insurance companies generate themselves. This way they are learning from past cases that have already been identified as cases of fraud. Technology can help insurers to apply this knowledge to future claims cases and to automate and optimize their claims process.
However, just because an insurance company has gathered a lot of data, it does not mean the data is easily available. Insurance companies often still struggle with a ‘silo mentality’ where different organizational units of a company don’t share their information with each other. Integrating the data is a necessary condition to access the valuable information the data provides and effectively use it to get an overview of the customers.
Fraud detection techniques
To extract valuable information from raw data, certain techniques need to be applied. The following methods can aid in detecting possible fraudulent cases:
- Pattern identification – For example, a medical provider who overcharges certain treatments based on false diagnoses. In a model for pattern identification, this tendency for overcharging would show up as a red flag.
- Anomaly detection – When certain events significantly deviate from the majority of the data, you have an anomaly, which might indicate something suspicious.
- Text mining – A technique where patterns are derived from large amounts of text data. This can be used to flag certain words that are known as good indicators for suspicious activity.
- Social network analysis – Social media can help identify the close social connections of claimants. Let’s say that a customer files an insurance claim following a car accident. Social network analysis reveals that the third party involved is in fact a close friend of the claimant. This might be something the insurance company wants to further investigate.
- Image screening – Increasingly, insurance companies have been using image screening to go over the photographic evidence provided by claimants. If claimants upload images of a car accident, you want to make sure that the image does indeed show the location where the accident took place, or the car that the claimant says has been involved in the accident.
- Voice analysis – Another useful technique that’s being used increasingly by insurance companies. For example, they can collect the data that’s generated in their call centers and apply voice analysis to recorded conversations and search for certain emotions in the voice that might be a possible indicator to detect fraud.
- Machine learning – All of this processes can be modelled. Insurance companies are expanding their use of machine learning so that the machine makes the decisions that humans would normally make.
Skilled employees need to be able to apply these data analysis techniques. This is where insurance companies struggle, because especially on the Western European market there is a shortage of these particular skills. Data scientists and computer scientists are highly valuable and therefore highly in demand. Insurance companies face tough competition from technology firms and need to find ways to attract skilled data scientists.
It’s the people, stupid!
Even with all the data, techniques for analysis, and skilled employees, insurance companies still don’t have everything they need to implement AI in their organization. The main issue is to have the right organizational culture that facilitates the implementation of AI in the work process. Not just CEOs and managers, but all employees must be open to change and accept incorporating AI in their daily work. To convince them, it is important to show the staff real business cases of the value AI generates.
The key question for insurance companies is how to find the optimal way to introduce AI in their organization. It might be beneficial to create specific departments for AI implementation and let them consider how AI will affect the claims fraud detection department.
Same old fraud, new weapons and battle ground
Insurance companies need to apply technological developments in order to improve the way they work. Technology is developing at a breakneck pace and can help the insurance industry to work faster and more accurately. AI will undoubtedly have a huge impact on every aspect of insurance, and especially in claims fraud detection there are great strides to be made. While the enemy of insurance fraud is not new, the weapons and battle ground are continuously changing. Insurance companies must keep up with these developments and keep AI at the top of their agenda.