Machine learning and AI are vital tools in the fight against fraud. They can help spot the regular patterns that indicate potential fraud. Fraudulent claimants know the correlations that algorithms are utilised to identify. They change the content of the "Claim Notification Forms" (CNF) to avoid detection.
So before you look where the grass is always greener in the adjacent and verdant AI technology valley look at an answer where you are right now.
Making inaccessible data accessible for analysis.
This is typically 60% to 90% of the data an insurance company already has but cannot analyse.
- Free text
Take insurance and the free text in the CNF. Most insurance companies cannot analyse this. Fraud inspectors do not have the tools to extract the value there to uncover fraud. So how can you access it?
Imagine you had an enterprise strength "Google Search" tool that accessed the 10% structured and 90% unstructured data
360Retrieve is such a product that let's the fraud team uncover those cases where the insurance company and the motoring public are being taken for a ride (sic). Claimants will include content that frightens an insurance company to make them settle relatively low unit cost claims quickly to avoid litigation e.g.
- include a minor in the claim
- use the full three year claim period to add new passenger claims in the final weeks
You would be surprised to know how many claims are made with minors during school hours. In a location given a generic name e.g. "Hanger Lane".
360TRetrieve gives the fraud team the tool to "follow their nose". Find similar cases and then the common correlations. All with inbuilt analytics.
The reward? Saving tens of millions of pounds a year in just one type of claim- whiplash.
Why aren't more insurance companies taking up this unglamorous but effective tool?
- Cultural resistance
- Data and functional silos
- Fear of showing how computer-based forecasting AND subjective judgement will outclass plain human cognition.
It's already used by some insurance companies, HMRC and Police so why not try it yourself?
Quality data Leonard Austin of Ravelin, a London-based startup which applies AI technology to fraud detection for online payments, says: “Given the choice between better algorithms and more data, I’d always rather have the data because algorithms are commoditised already – there are so many of them to choose from. The better quality data and the more data you have, the more you can predict.”