Straight from the pages of a sci-fi novel, an artificial intelligence system that can learn to recognise fraudsters has been developed. Clive Nicholls explains

To download the Power Point presentation for this article please
click here (1.03MB)

How much does it cost for a good fraud investigator? You know the ones who have the knowledge, experience and intuition to spot a potentially fraudulent claim.

The answer is tens of thousands of pounds per year. Now imagine being able to do the same job automaticaly using a sophisticated software programme. Well that's where NCorp's Ijen comes in.

Ijen uses advanced pattern recognition technology to spot potentially fraudulent insurance claims. It acts as a virtual fraud detection expert which can sit at the heart of a claims processing system or 'on the shoulder' of a claims handler, alerting him to suspect activity in real-time and aiding decision-making.

The software mirrors the understanding and intuition of human anti-fraud experts and is intended to improve the efficiency and effectiveness of current fraud detection processes.

It works by learning the behavioural patterns of fraudsters (a bit like Amazon learning how to predict what your favourite books and music are). Using this knowledge, Ijen can then automatically analyse claims data to spot similar patterns, which may indicate an attempted fraud or identify a potential fraudster. This information can then be passed on to anti-fraud investigation teams.

Ijen is at its best when implemented as a first-line filtering mechanism that enables insurers to identify and 'quarantine' those claims showing the highest risk of fraud.

Insurers can then test these high-risk claims further by applying additional techniques such as conversation management. These techniques confirm the validity of the claim or encourage claimants to voluntarily withdraw their (fraudulent) claims. This means resources can be more efficiently deployed.

In trials, Ijen increases fraud detections by over 50% when used to filter and select claims for further scrutiny.

The filtered claims that do not exhibit factors that may indicate possible fraud can be fast-tracked for payment. This not only improves customer service and enhances loyalty through prompt settlement, but it also reduces insurer administration costs.

Ijen is able to continually learn and adapt as dishonest claimants change their behaviour and methods, enabling it to maintain a high degree of accuracy and performance. The ability to learn makes the solution fundamentally different, and superior, to other less-sophisticated technologies, which use predefined rules or work from a static 'snapshot' of fraud patterns.

It is often the case that these snapshots are updated infrequently, leaving a massive window of opportunity for fraudsters to change their methods in order to beat the system.

Ijen can be rapidly implemented with low ongoing maintenance costs. It never gets tired, always applies itself consistently and continuously learns. We reckon it's the perfect virtual fraud detection expert.

  • Clive Nicholls is chief executive of Aquillo which works on implementaion with the system developer Ncorp