This article first appeared in Digital Edge, The Edge Malaysia Weekly on May 22, 2023 - May 28, 2023

Making false or exaggerated motor claims is one of the most common types of automotive insurance fraud in the country, costing insurers RM1 billion in 2021 alone.

It is both costly and time-consuming to assess the authenticity of motor claims. But with the use of artificial intelligence (AI), the menial tasks of analysing large volumes of data and identifying patterns of behaviour that may indicate fraudulent activity are automated, freeing human agents to focus on more complex analysis.

Fraud is the manipulation of data — either to disguise the information to make it realistic or go undetected, or to create new data points that seem plausible. The challenge of many insurers is that the detection of motor insurance fraud has traditionally been a manual process, relying on the expertise of industry professionals and in-house efforts, says Peter Miller, CEO of Fermion Group. “However, with the significant volume of claims processed daily and the fact that fraudsters are getting more creative, this approach can be time-consuming and may not always uncover all fraudulent activities,” he adds.

“Fraudsters often change details and provide false information in order to obtain payment. This makes it difficult for insurers to verify the validity of claims. However, with the rise of AI tools and the ability to collect and store vast amounts of data, insurers are now better equipped to cross-reference claims data and detect inconsistencies that would indicate fraud.”

Fermion, which is part of Singapore Exchange-listed Silverlake Axis Ltd, is an insurance software-as-a-service provider.

In 2019, the Coalition Against Insurance Fraud — a US-based alliance of consumers, insurers, government agencies and legislators — estimated that the global cost of insurance fraud stood at US$80 billion.

The exact cost of insurance fraud is often difficult to estimate because much of it goes unreported, according to the General Insurance Association of Malaysia (PIAM). As fraud detection costs insurers time and money, consequently, the expenses get passed on to drivers through higher premiums.

But AI can flag claims that have a high likelihood of being fraudulent, based on past claims data, medical records and other relevant information to separate authentic claims from suspicious ones.

According to Allianz Malaysia, motor claims fraud in the country cost insurers RM1 billion in 2021. Here is where AI comes into the picture. It can replace the surveyor by visually scanning photos or videos of the scene to assess claims or to more conventionally interrogate vast amounts of motor claims data to detect patterns and relationships that humans are not easily able to identify, explains Miller.

“AI first checks the veracity of a motor insurance claim by cross-referencing the information submitted on the claim, detecting inconsistencies like whether a specific type of vehicle is more frequently involved or if a specific repair shop is making a higher-than-usual number of claims,” he says.

“It then filters claims that require further investigation by flagging those that do not comply with pre-programmed approval decision criteria, such as to compare the date of the accident with the date of the claim and cross-referencing the policyholder’s location at the time of the accident. If a discrepancy is detected, the claim is flagged for further investigation. Essentially, the AI model serves to correct errors by checking every claim against decision rules, thereby reducing the risk of paying out fraudulent claims.”

With AI, insurers can amass vast amounts of data, cross-reference the claims data and identify any inconsistencies that may suggest fraudulent activity is taking place, says Miller. “For instance, big data can be utilised to verify whether a claim had been made on a stolen vehicle by cross-referencing it with the police report and highlighting any disparities.”

By automating manual processes, insurers gain enhanced accuracy and real-time insights, which then improves their ability to detect and prevent fraudulent activities, he says. “This not only safeguards the industry against financial loss but also strengthens trust and credibility with customers.”

However, data is only one part of the equation. “A successful AI fraud solution requires both rich localised data and domain knowledge,” he points out.

Fermion has 22 years of deep domain expertise, specialising in motor claims technology, says Miller. The group has been operating in Singapore for more than 15 years. The company’s data science team works closely with business analysts and insurers’ claims teams to build and maintain fraud intelligence rules and frameworks tailored to motor claims and fraud patterns in Singapore and Southeast Asia.

To combat motor insurance fraud, Fermion’s TrueSight solutions use tools like AI, data science and interactive video to expedite the delivery of insights throughout the claims process.

“AI assists insurers in evaluating and categorising each claim according to predetermined parameters, by assigning a corresponding risk profile such as green, amber or red,” says Miller.

“Data science aids insurers in cost management by presenting a dashboard displaying current benchmarks for parts pricing and repair logics, thereby enabling insurers to avoid overpricing and detect possible fraudulent activities. Interactive video allows assessors to rapidly evaluate damages, eliminating the necessity of being present on-site for the inspection.”

Etiqa Insurance & Takaful and Allianz General Insurance are among the local insurers that use TrueSight Fraud Intelligence to analyse motor claims submitted via the Fermion Merimen eClaims system, which is designed to provide insurers with instant and real-time detection results.

TrueSight Fraud Intelligence applies AI to screen claims that are processed electronically. The AI is fed with a rich localised data set, gathered from the millions of motor insurance claims made to over 150 insurers in Asia each year. Thanks to the AI-powered solution, local insurers managed to detect more than 7,000 fraud alerts out of the over 100,000 motor claims processed, says Miller.

Fermion processes these claims through its eClaims platform. The platform collects hundreds of structured data points from across Fermion’s entire motor ecosystem, including repair workshops, adjusters, lawyers, drivers and claimants. This large dataset provides greater accuracy when compared with other solutions.

TrueSight Fraud Intelligence augments the eClaims management platform, which performs AI analysis on processed data to suggest where potential fraud risks lie. The result is a visual network diagram that presents the hundreds of data points, aggregated, to help insurance providers quickly identify where fraudulent activity may be occurring.

Claims that are identified as potentially fraudulent are flagged immediately and categorised as suspicious, then referred or assigned to a team member from the insurer’s fraud investigation unit. If the claims are deemed not fraudulent, they undergo standard processing in the eClaims system. However, if fraud is detected, the claims are rejected.

“The benefits of TrueSight Fraud Intelligence have resulted in time savings, which means the turnaround time of claims processing improves: (i) the investment in resources as there are fewer claims to investigate, as the AI identifies potentially fraudulent claims; and (ii) it is less capital-intensive as internal costs are reduced. In implementing TrueSight Fraud Intelligence, Etiqa Insurance & Takaful has managed to repudiate close to RM1 million worth of suspicious claims,” says Miller.

This raises the question: If data and AI are the panacea to combating insurance fraud, why aren’t all insurers using it?

“Some insurers are still hesitant to adopt AI-based approaches, for two primary reasons. The first is the cost involved in implementing these approaches, which often necessitates a substantial investment in technology infrastructure. In many cases, insurers are still utilising outdated and siloed infrastructure, which presents a barrier to the adoption of modern technological solutions,” says Miller.

“The second reason is the lack of in-house expertise required to implement big data and AI-based approaches, which demand technical knowledge in areas such as data analytics, machine learning and computer vision. The quality of the data is a crucial factor in the success of big data and AI-based approaches. If insurers do not have access to high-quality data, their models may not be accurate, resulting in false positives or false negatives.”

Moreover, the business of inflating claims, also known as road touting, is another challenge that stands in the way of detecting auto insurance fraud. It has been difficult for insurers in Malaysia to detect fraud in this area as there is a lack of integration between data sources as well as insurance players working in silos, he reiterates.

A lack of public awareness also contributes to the problem as fraudsters take advantage of victims in shock from accidents. Siloed infrastructure and low data quality can prevent insurers from getting a complete picture of a claim and may lead to missed opportunities for fraud detection.

“However, the increasing digitalisation of motor claims processes and technological advances are changing this scenario, rebuilding trust between insurers and their customers who are often the unwilling victims of insurance fraud,” says Miller.

It is worth noting that not all claims can be processed with automation. In some occasions, when information is flagged as incorrect upon submission, the claim is rejected or sent back to the claimant for more information. “About half of the claims are higher risk claims, which are sent to staff who then interview claimants to assess the more complex claims,” he says.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's AppStore and Androids' Google Play.