Automation promises better fraud claim detection, but do we still need humans?

Can a computer program be better at detecting claims fraud than a human? A new tool developed by insurance law firm Horwich Farrelly inevitably throws out this very question.

The firm said it ran 50,000 claims through its ‘car wash’ tool over the last year highlighting 4,500 fraud matches, and even identifying cases where fraud had not been predicted beforehand.

Rick Preston, head of Horwich Farrelly’s Intelligence team told Insurance Times that the firm developed the tool to complement existing manual vehicle data investigations.

“What we wanted to do was make sure we were in a situation where we never missed an opportunity for a ‘quick win’, simply because a claim wasn’t referred for investigation in the first place. It became obvious that it was entirely possible to automate significant parts of the investigations process, those which did not involve a human ‘skill’, to achieve a situation where every single claim is investigated at minimal cost and resource,” Preston said.

Response

Since rollout, the firm has received a positive industry response, with one client asking for its entire book of historic claims to be ‘washed’, Preston said.

Car Wash works particularly well for the credit hire sector as those claims revolve around vehicles/vehicle data. Averaging savings per case can be significantly higher than other sectors, Preston added.

Quizzed as to whether the tool could ever replace human involvement, Preston responded: “We don’t believe that algorithms should fully replace the human element, they should simply aid it.

“Car Wash is very much about automating the processes that should be automated, not the ones that don’t.” He added that instead of fraud investigators wasting valuable time reviewing every claim - and in many cases finding nothing - the system allows them to concentrate on the cases where discrepancies/fraud indicators have already been identified.

Future trends

Such technology is broadly welcomed by major insurers. Nick Kelsall, fraud manager at Allianz Insurance, told Insurance Times that digital and automated detection has to be the future in a world where end-to-end claims processing is also becoming more automated with less human touches and intervention. 

”The sharing of data is therefore key for us as an industry and wider society in order to protect our genuine customers. We need to work together as an industry to get automated detection right and this can only be achieved by working with the Insurance Fraud Bureau (IFB) and contributing to the Insurance Fraud Intelligence Hub (IFiHub),” he said. 

Kelsall believes despite the opportunities presented by automation, robotics and AI, the industry needs to be careful not to lose sight of human expertise.

They still have a significant role to play in fraud detection and more importantly prevention. Simply having an automated fraud ‘flag’ isn’t enough – you then need well trained and dedicated staff to turn that ‘tip off’ into evidence that will convince a Court that the matter is fraudulent – data alone cannot do this. 

Kelsall also pointed out that while being solely reliant on data is good for trends, there is still ”a need for people to establish and identify the fraud of the future as it becomes more elaborate”.

He added: ”It seems to me that the perfect balance is not one or the other, but a mix of digital sophisticated detection supported by cross industry collaboration, combined with supporting our talented fraud experts to cover all bases and ensure that fraudsters do not prevail.”

Catherine Burt, National Head of Counter Fraud at DAC Beachcroft sees automation becoming part of the future of fraud detection.

Whiplash reforms

She told Insurance Times that the forthcoming whiplash reforms are likely to mean that insurers need a speedy, smart and cost-effective way of flagging potentially fraudulent claims to ensure that genuine claims are paid as quickly as possible, while weeding out those claims which need further investigation. ”It will be important for all those involved in fraud detection to have some form of automation going forward”, she added.

Burt stressed however that the human element in fraud claims detection remains important.

”We have always had a human element in our data washing service which has added real value for those who have used it. It is important to verify information and check for false positives, and it remains the case that some large fraud rings are detected as a result of the gut instinct of one handler or analyst who spotted an anomaly and chose to investigate.”