Role of AI in ending Insurance fraud
Over the past year the UK has experienced a surge in the value of insurance frauds taken to court (£17 million) than the all the fraud cases from 2014-2017 combined (£11.9 million). UK insurers detect ‘one scam every minute‘ and a total of 562000 frauds in 2017. This increase in Insurance fraud costs the British economy 1.3 billion a year.
The University of East London has teamed up with AI startup “Intelligent Voice” (specialising in speech recognition) in an attempt to tackle this rise in insurance fraud.
The government's “next generation services challenge fund” (a government fund that grants up to £20 million of funding “to establish how new technologies could enable the UK accountancy, insurance and legal services industries to transform how they operate” )has backed the project with a £13 million grant.
One of the biggest issues regarding insurance fraud insurance companies have is the amount of fraudulent calls, which results in legitimate consumers taking longer to be attended and a decrease in overall productivity due to waster labour hours, as well as the cost to the company in legal costs and any successful frauds.
Intelligent voice will tackle this issue with machine learning and speech recognition as well as Strenuus’ behavioural analytics capabilities.The end algorithms will also have an “expandability layer” that , for each case, what decision is made and when to make it.
By using a short-term memory network on Nvidia GPUs which can recognise certain patterns in voice, such as tone or phrases, the system can point out potential fraud and alert the agent to respond appropriately.
The system has been trained using hundreds of thousands of manually labelled insurance related calls for the system to have as high an accuracy as possible.
Strenuus also received £1,361,570 for AI software that detects emotion in voices for the purpose of 'credibility assessment'.This will be tested in an insurance contact centre during live claim handling.
As systems such as this one continue to improve we can expect a fall in the value of successful frauds, and over time a fall in the value of attempted frauds as more and more fraudsters give up against an ever improving system, this will result in lower premiums for everyone else (on average fraud costs add £50 to value of premiums) and higher profits for insurance companies.