
AI Case Study
RBS identifies invoice fraud saving £7 million of losses to customers with artificial intelligence
Invoice redirection takes place when banks are instructed by bogus suppliers to redirect invoice payments to a new bank account. it is worth millions a year to criminals and RBS adopted a new technology with Vocalink to scan transactions of a million small and large business customers using machine learning. The company claims £7M of losses have been prevented.
Industry
Financial Services
Banking
Project Overview
"The bank and Vocalink, a payments business, have created a system that scans transactions by about a million small and large business customers to identify fake invoices and prevent customers from paying fraudsters."
Function
Risk
Audit
Background
"Invoice redirection fraud takes place when criminals contact a business claiming to be from one of their suppliers, saying that they have changed bank and requesting that an invoice is paid into a different account. If the business makes the payment, the money is transferred between accounts by the perpetrators, often going out of the country to make detection harder. Criminals can target individual payments worth hundreds of thousands of pounds, which can be crippling for small and medium-sized businesses.
'Detecting invoice redirection fraud is akin to finding a needle in a haystack, as there are tens of millions of legitimate payments every day.'
UK Finance revealed that there were 43,875 reported cases of authorised push payment scams with a total value of £236 million in 2017. Eighty-eight per cent of this total were retail consumers, losing an average of £2,784, and the remainder were businesses, which lost on average £24,355 per case."
Reported Results
RBS claims to have reduced fraud to its customers by £7m.
Benefits
Technology
"'We apply sophisticated analytical techniques to vast amounts of payments data to build models which identify suspicious activity. Every time a business pays an invoice, a behavioural signature is left behind. By analysing these signatures, and the signatures of historical frauds, we are able to identify and flag suspected incidents of fraud.'"
Data
"Scans transactions by about a million small and large business customers to identify fake invoices and prevent customers from paying fraudsters."