Intelligent Voice: combating fraud with artificial intelligence
More than half a million insurance frauds were detected in the UK in 2017 – more than one every minute. The value of fraudulent motor insurance claims alone was £775 million.
London-based technology company Intelligent Voice is leading a £2 million 2-year project supported by the Industrial Strategy Challenge Fund to trial the use of artificial intelligence to detect fraudulent claims during live claim handling.
This project is funded through the ISCF’s next generation services challenge. The challenge is supporting businesses to develop the next generation of services: accountancy, insurance and legal.
Boosting UK insurance firms’ competitiveness
Intelligent Voice’s technical director Nigel Cannings said:
"The aim of our system is to make UK insurance companies suffer less fraud and be more competitive than any other insurance companies in the world. Insurance fraud increases claims, costs and premiums.
"No one wants to phone an insurance company and be treated with suspicion, only the criminals should be treated as criminals. There’s a hard benefit for customers in reduced cost of premiums and a soft benefit in improved customer service."
Using machine learning to improve detection
Intelligent Voice has pioneered the use of graphical processing units for speech recognition and speech processing in sensitive environments. It counts the US Department of Justice among its customers.
The firm works closely with its partner, Strenuus, a company with specialist skills in identifying fraudulent behaviour. The work involves text transcripts being manually assessed by experts in behaviour that suggest fraud.
The two have already developed a way of automating Strenuus’ expertise.
"We have demonstrated that you could start to send potentially fraudulent calls down the correct pipeline as easily as a well-trained human, but we need to be better than humans. The project will involve the collection of huge amounts of data from the insurance industry and processing it using machine learning. The computer uses a neural network to learn the type of language and behaviour that lies behind fraudulent claims.
"We know that when you use a data-driven approach to assess anything, the results become better than a human. You give your neural network a huge amount of data and a set of categories – for example claims accepted, claims refused, and claims withdrawn – and you say to it: ‘You go and work out what the rules are’.
"However, GDPR [general data protection regulations], the latest data regulations, mean you must also explain why the computer said no. This is not easy to do.
"We are partnering with the University of East London, which has a depth of machine-learning expertise. It is working on how we get the machine to tell us why it reached its decision.
Trialled on live claims
The project will run a trial on live claims handling. The system will run alongside call centre workers dealing with claims and be assessed against them.
"The key to proving the system is to find the calls that the human failed to spot and explaining the reason why."
Targeting global markets
Intelligent Voice won £888,000 from the Industrial Strategy Challenge Fund’s Next Generation Services Challenge to support its work on the project. University of East London won £473,478.
The Next Generation Services Challenge is a £20 million fund supporting new technologies that could transform the UK accountancy, insurance and legal services industries. It is part of the Industrial Strategy’s artificial intelligence and data grand challenge.
Intelligent Voice employs 25 people and expects to add another 5 staff to work on the project. If successful, it is likely to see the company grow significantly in the next 3-5 years.
Our aim is to have a fully working and commercialised system at the end of 2 years. We see the market for this software as a global opportunity. The global fintech market runs into billions. This is a a springboard to taking Intelligent Voice to the next level.
We sell our software into the packaged systems that bigger companies then go out and sell to the finance sector. We want to push this software out through the wider insurance-tech channel and through businesses developing call centre systems, both here in the UK and globally.