The July theme feature is about finding a balance when it comes to identifying potential insurance fraud. It’s a difficult subject, with the potential to upset consumers and attract stricter regulation if enough people complain. So we cannot use AI to automate every claim, there has to be a human touch somewhere along the line. Here are some industry comments and thoughts;
TANIUM
Mark Millender, Senior Advisor at Tanium thinks that the greatest benefit of technology is in spotting the problem, via cross-referencing data;
“In today’s digital landscape, identifying and mitigating suspicious behaviour is critical for organisations to protect their assets and maintain operational integrity. Software platforms play a vital role in flagging unusual patterns, malicious activity, and vulnerable endpoints by leveraging real-time data and, in some cases, machine learning algorithms.
“By continuously monitoring network and endpoint activity, software platforms can detect anomalies that deviate from established baselines. Vulnerabilities and suspicious activity, such as an unexpected surge in login attempts from a single IP address can trigger alerts, prompting further investigation. Similarly, the detection of unusual communication patterns between devices or the identification of previously unseen devices connecting to the network can indicate potential security threats.
“Devices associated with suspicious behaviour are flagged through correlation analysis. This involves cross-referencing data from multiple sources to identify commonalities that may signify coordinated activities. For instance, if multiple fraudulent claims originate from the same IP address or phone number, the platform can raise a red flag for further scrutiny.
“Claimants exhibiting atypical behaviour, such as frequent changes in personal information or inconsistent login locations, are also monitored closely. Machine learning models can learn from historical data to identify these irregularities, enhancing the organization’s ability to detect and prevent fraud.
“By harnessing real-time data, the likes of Tanium’s Converged Endpoint Management (XEM) platform offer security teams real-time alerting, allowing organisations to immediately identify and respond to suspicious activities. Harnessing AI to drive autonomous endpoint management, Tanium aims to continually improve response times and mitigation effectiveness. Software platforms also enable remote forensic investigation on suspicious endpoints, allowing enterprises or others managing incident response to take a wide variety of remedial actions, such as imposing network or device quarantines or deploying patches.
“The ability to flag suspicious patterns is crucial for pre-emptive threat mitigation, enabling organisations to stay one step ahead of potential security breaches and maintain a secure digital environment.”
SYNECTICS
Rob Bevington (Head of Data Science) at Synectics, stresses the importance of verifying ID at every step;
In April 2024, the Insurance Fraud Bureau (IFB) reported that fraud resulting from stolen identities has almost doubled in the last 12 months. Our own figures, based on National SIRA analysis, show that ID fraud now accounts for a third of all claims fraud.
All this understandably means spotting ID fraud early in the claims process is an industry priority. Exploring options at FNOL stage is a logical place to start.
It’s a stage that is particularly important to customer experience. A study by EY suggests 87% of customers base their decision to stay with an insurer on their claims handling, making a smooth FNOL stage business critical. But as the stats show, robust ID checks are clearly needed to keep fraudsters at bay. By impersonating a genuine policy holder at FNOL, scammers can (and do!) report non-existent events and losses to fraudulently claim compensation. Leaving the insurer out of pocket, and genuine customers disgruntled.
Digital ID verification is an ideal solution, in terms of both immediacy and accuracy. Take our SynID tool as an example. It uses evidence from public and private fraud databases, and combines this with strength-scored identity and activity history attributes, in line with the UK’s Digital identify and Attributes Trust Framework (DIATF).
In less than 5 seconds, address and date of birth details are checked, ID fraud indicators are flagged, and the data sources are scanned to assess the frequency, recency, and quality of an individual’s digital interactions with organisations. The digital nature of the process is often easier for customers, especially for those who may struggle with traditional ID checks. And for the insurer, it means more genuine claimants can be fast tracked, without increased risk of fraud.
LEXIS NEXIS RISK SOLUTIONS
Xumeu Planells, Data Science Manager, LexisNexis Risk Solutions, offers these insights;
Every insurance provider has a responsibility to investigate fraud in all its guises, not least because when fraud goes undetected, we all pay in increased premiums. However, full automation of fraud detection is unrealistic. It needs the right mix of human skills, data and technology.
Criminals are becoming more technically skilled, generating content that is difficult even for an expert human eye to identify. Deepfakes or shallowfakes over-exaggerate the cost of a claim or simply distort reality. If a human can’t detect fraud, it would be dangerous to rely solely on technology to do so.
Currently, detecting fraud tends to rely on a set of basic rules to look for in a claim, then human skills will investigate those deemed suspicious.
The future of fraud detection will therefore continue to rely on the ‘gut feel’ of an experienced claims professional but there will be more and more tools (especially AI and generative AI) so that those current basic rules are not as restrictive and become more flexible depending on the type of claim. For example, digital forensic tools using AI can be used to identify pixel and image manipulation, even spotting fake images created by generative AI.
More immediately, highly granular claims data for the person and the asset with cross search functionality across Home and Motor has the potential to transform insurance fraud detection. LexisNexis ® Precision Claims, the contributory database solution could help insurance providers predict fraud before it happens. The more the industry shares claims data, the better equipped we will be to develop insurance specific fraud tools in the future.
SAS
We definitely need a balanced approach, a human and tech hybrid model when it comes to insurance fraud. Here are some thoughts from Andrew Pollard, Insurance Specialist at SAS
“In 2024, fraud losses continue to pose a serious threat to the profitability of insurance companies. Insurance giant Aviva recently reported a 39% rise in insurance fraud claims, while the wider insurance sector has seen false applications rise by a fifth (20%) in the last year.
“Tackling insurance fraud presents a clear opportunity for insurers to save money and maintain competitiveness, with many turning to anti-fraud tech powered by artificial intelligence (AI) and cloud analytics to do so. It uses multiple analytic methods and layers of fraud detection to identify new and emerging fraud threats, while helping to uncover previously unknown relationships.
“Nevertheless, for the most successful outcomes, it is important to balance innovation in technology with human oversight. After all, over the last decade, reduced human interaction during the insurance customer lifecycle has led to fraudsters flourishing, as new tactics are made possible by the move to digital channels.
“Technology can flag suspicious activities, while human expertise helps to interpret findings and act on them quickly. A hybrid approach is the best way forward, where experienced investigators can review the insights AI generates from bringing together siloed data systems and sources. Too often, time-consuming and error prone manual data preparation can result in a large number of false positives, which advanced AI analytics techniques help to reduce.”
Be the first to comment