Prudential to begin trial to speed up claims processing

Insurer Prudential Singapore is testing a system using machine learning that could assess a claim within in seconds.
Insurer Prudential Singapore is testing a system using machine learning that could assess a claim within in seconds.ST PHOTO: DESMOND WEE

SINGAPORE - Insurer Prudential Singapore is testing a system using machine learning that could assess a claim within in seconds.

The first phase of the trial will begin at towards the end of the month and focus on automating the processing of PRUshield pre-hospitalisation and post-hospitalisation claims from eight major hospitals.

These claims form the bulk of the 14,000 paper bills and receipts that Prudential's claims assessors review every month.

Selected policyholders can upload scans or images of bills and invoices through an online portal. Payable and non-payable items will first be identified and sorted by a text-mining engine.

Claims are then electronically assessed by a machine-learning engine and outcomes like "approve", "partial approve" and "decline" recommended, as well as the payment amount.

Initially, claims assessors will review the machine's recommendations and provide feedback."The system has already been trained and back-tested using claims data from the last two years and has reached a good level of accuracy," Prudential said.

It noted that the claims assessment time could be progressively shortened from seven days to mere seconds by the time the trial ends in the first half of 2018. The full launch will be in the second half of next year.

Fellow insurer NTUC Income said in June that it was also using imaging and machine learning engines to process its hospitalisation claims. It is using platforms from IBM.

Like Prudential, it has to process a significant number of claims a month. There are about 14,000 hard copies of IncomeShield pre-hospitalisation and post-hospitalisation claims.

A platform will similarly analyse data before making claims recommendations and calculating payouts.