New ammo in the fight against insurance fraud

Exciting analytics

By utilising vast amounts of data from diverse sources including real-time data, insurers can effectively detect and prevent fraud using tools such as predictive analytics and machine learning. 

“AI technology can analyse vast amounts of data, including customer information, claims history, transactional data and network relationships, to detect patterns and anomalies that may indicate fraudulent activities,” said Fabio Sarrico, Insurance Analyst at Celent. “By leveraging predictive analytics, data mining techniques and network analysis, insurers can identify suspicious behaviours, flag potential fraud cases, and take proactive measures to prevent fraudulent activities.” 

Machine learning algorithms further enhance this process by continuously learning from new data, enhancing their ability to detect fraud over time. “Unlike rule-based systems, which lack adaptability by relying upon predetermined parameters to define how information is sorted and may struggle with ambiguities, machine learning can adapt to evolving fraud patterns by learning from new data,” confirmed Gary Sommerford, Director of Larus Consulting. 

Mather seconded this point: “The ongoing development of advanced analytics is creating the capacity to review an ever-increasing number of data sets across thousands of claims and consistently arrive at an accurate real-time conclusion.”

The development of AI-generated photos and images is also a huge risk
to the industry

The human touch

A persistent issue in fraud detection is that of false positives, whereby legitimate claims are incorrectly flagged as fraudulent. “False positives can result in costly, time-consuming and pointless investigations, sometimes causing needless disruption to the business, drawing unwarranted focus and concerns to incorrectly identified risk areas, and potentially having a detrimental effect on customer experience,” warned Sommerford. 

But the use of advanced machine learning models, high-quality data and refined analytics can minimise this problem. This is something that insurers are aware of and striving to mitigate. “To optimally manage the increasing threat of fraud, we need solutions that do more than simply reduce fraud rates,” said Rory Yates, SVP of Corporate Strategy, Global, at EIS. “Although this is undoubtedly the primary role of any fraud prevention system, it needs to happen in conjunction with fewer false declines and a smooth customer experience to be truly effective.” 

Indeed, technology must be utilised in the right manner in order to be truly effective, as Cook emphasised: “So long as technology is deployed sensitively and thoughtfully, it lets claims handling companies scale up the number of cases they can deal with and improves the service levels enjoyed by policyholders.” 

Key to this is the pivotal role of humans and the need to abide by regulatory requirements, as Sarrico underlined: “Human oversight and intervention should be maintained to review and validate AI-generated alerts and decisions, ensuring accountability and mitigating potential biases and data manipulation in generative AI.”

Morton agreed that AI needs to work in tandem with experts: “The technology is there to help the insurance professional make the best decision about what to do with a claim. Is it suspicious? Why is it suspicious? How should we investigate? Can we make the right determination? You never want to be in a position where ‘computer says fraud’.”

Lightning speed

Gen AI is an important tool in insurance fraud, improving fraud detection accuracy and streamlining operations. For example, it can generate synthetic data to train and improve machine learning models, automatically verify the authentication of documents submitted during claims processing, and identify deep fakes that might be used to support fraudulent claims. Crucially, it can do this at incredible speed. 

“The enterprise adoption of generative AI is especially exciting,” enthused Morton. “The accuracy and efficiency this technology can add to the fight against fraud is truly astounding. 

“Take entity resolution, for example. It is not difficult to manipulate information in such a way as to hide the fact that several, if not dozens, of different policyholders are the same person. Multiple permutations of a name, slightly different address formats, multiple email addresses, etc. can make it difficult to know exactly who you are doing business with. Using gen AI to analyse that data, identify common PII [personally identifiable information] and make the connection that identifies these multiple identities as a single entity is already a game changer. To be able to do it in seconds versus hours is where the real power comes into play.” 

Another exciting development is the use of biometric technology, including facial recognition and voice analysis, to verify the identity of policyholders and claimants. “Whilst not completely foolproof, biometric verification is difficult to fake and can enhance fraud protection measures,” said Sommerford.

Combining IT and IQ

To ensure accuracy, efficiency and compliance, guard rails must be implemented around technologies in fraud detection. First of all is the need for strong foundations: “Given there are a myriad of solutions, experiences, and ways to detect fraud, you need technology foundations that make you highly adaptable, allowing you to interoperate with this latest raft of advanced technologies,” highlighted Yates. 

Another requirement is model training and validation. “Ongoing training and education of AI models are essential to adapt to evolving fraud patterns and maintain the effectiveness of fraud detection systems,” stated Sarrico. 

Yates affirmed the need for harmony with effective systems: “AI is great at helping us better detect fraud, but we also need it to work directly with systems that manage processes, trigger rules that engage agent activity and further exploratory tasks that shape experience outcomes directly. This intelligent orchestration also spans the entire life cycle, from quote to claims.” 

This underscores the need for human involvement. “You can have the best technology in the market but you still need an experienced investigation team to pick up the referrals, investigate the claims and decipher whether or not they are genuine,” said Cook. “This is why we always say that you must blend IT with IQ and you cannot have one without the other these days.” 

Privacy and security are further key considerations. This involves abiding by data regulations as well as ongoing transparency and monitoring. “Insurers must prioritise data privacy and security, ensuring that customer information and sensitive data are protected throughout the fraud detection process,” said Sarrico. “Data protection regulations vary depending on the market; therefore, insurers must follow the particular regulations accordingly. Additionally, transparency and explainability of AI algorithms are crucial to gain stakeholders’ trust and comply with regulatory requirements. Regular audits and oversight mechanisms should be implemented to monitor the performance and fairness of AI systems.” 

You never want to be in a position where 'computer says fraud'

Cook pointed out that in addition to being compliant, technology must be effective. This is key to customer satisfaction. “It is easy, for example, for a new system to generate so many false positives that it can interrupt a genuine customer’s claims experience,” he 
told ITIJ. “It is absolutely critical that we are not putting genuine customers through an unnecessary fraud process. Carriers must 
be confident in the usefulness of any new technology and not fall into the trap of trying to adopt every tech-led idea before its implementation has been thoroughly validated against intended 
and unintended consequences.”

Time to prioritise

Given the key role of data and data sharing in fighting insurance fraud, there is room for improved collective intelligence. 

“Insurers’ hesitation with sharing data limits the extent to which fraud can be identified at point of sale and point of claim,” Edwards remarked. “An increased appetite from insurers to treat fraud as a priority and agree to mechanisms to support and limit this activity would be well received.” 

Cook noted that there is no centralised database for the travel sector and such a resource would be invaluable. “The insurance industry has a number of excellent databases and organisations that collaboratively and compliantly share intelligence to help combat insurance fraud. These include [in the UK] the Insurance Fraud Register, the Insurance Fraud Bureau and the Insurance Fraud Investigators Group. 

“But the travel sector does not have a centralised means of finding out if a claimant is making multiple claims with different insurers, or of compliantly sharing a complete picture of a policyholder’s travel claims data and history,” he stated. “In one recent case we uncovered a fraudster who had attempted to make £75,000 worth of fraudulent travel insurance claims through 13 policies, taken out with five different insurers. There has been a lot of talk and some exploratory projects to incorporate travel insurance claims data into the Claims and Underwriting Exchange, but nothing has been pushed through to generate a centralised UK database. Admittedly, this would be a complex and lengthy project to complete.”

Two steps ahead

As advancements in AI, machine learning and data analytics continue apace, the role of technology in insurance fraud will continue to evolve. This is a catch-22 situation as these developments provide new opportunities for those looking to commit fraud. “I believe that the role of technology, especially AI, in the insurance fraud space is going to be more and more of a double-edged sword,” said Sarrico. “While technology plays a vital role in both preventing and uncovering insurance fraud, and numerous insurers have effectively implemented digital procedures to combat fraudulent activities, it can also enable fraudsters to create more sophisticated schemes.”

It is absolutely critical that we are not putting genuine customers through an unnecessary fraud process

This is why it is paramount for insurers to remain vigilant. “The bad actors will continue to invent new ways to try and scam insurers. The industry needs to stay a step or two ahead,” advised Morton. 

Cook agreed: “While the market must remain focused on its own collective activities, it cannot afford to lose sight of how criminal trends are developing and where technology is being used to perpetrate fraud.”

The travel sector does not have a centralised means of finding out if a claimant is making multiple claims with different insurers

Indeed, the insurance industry must continue to invest in new technologies and stay ahead of fraudsters who are also becoming increasingly sophisticated. “Whilst technology can be expensive and may be resource intensive to implement, in the long term, outdated systems and accompanying processes are likely to end up costing more to maintain and may not be supportive of software updates, security fixes, changes in regulatory requirements, or indeed changes in the insurance landscape,” stated Sommerford. 

Cook agreed that investing in new technologies and remaining ahead of the curve will produce dividends: “As insurers assess their technology investment strategies, they should consider not only the benefits of being in the leading peloton, but also the dangers of falling behind. Ultimately, being off the pace will prove to be a lot more costly.”