OCBC sees AI payoff in transaction monitoring

OCBC Bank said it would extend its testing with ThetaRay after early results showed that the company's technology was able to reduce the number of alerts that did not require further review.
OCBC Bank said it would extend its testing with ThetaRay after early results showed that the company's technology was able to reduce the number of alerts that did not require further review.PHOTO: ST FILE

OCBC Bank is stepping up its collaboration with an Israeli fintech firm to employ artificial intelligence (AI) and machine learning to combat financial crime.

It said yesterday that it would extend its testing with ThetaRay after early results showed that the company's technology was able to reduce the number of alerts that did not require further review - by 35 per cent.

The technology was also better at categorising flagged transactions by their risk levels, which vastly improved the accuracy rate of identifying suspicious transactions, the bank said.

The test was based on one year's worth of OCBC corporate banking transactions but with names anonymised.

This, in effect, has helped the bank to prioritise the flagged transactions according to risk.

OCBC's head of group legal and regulatory compliance, Ms Loretta Yuen, said the existing transaction monitoring system is a rule-based one, which makes scanning risks very fixed and means they are handled on a "first in, first out" basis.

By embedding the fintech firm's technology into the existing system, around 4,200 alerts have been grouped into 48 unique risk clusters for the compliance team to sieve.

OCBC also noted that its chatbot application called "Emma" has closed more than $70 million in home loans since the start of the year. It took three months for "Emma" to be fully trained to address all possible questions asked by consumers about home and renovation loans, with one of the top three questions being how much a consumer can borrow.

OCBC, which aims to fully implement the technology in the second quarter of next year, said it is the first bank in Singapore to tap AI and machine learning to combat financial crime. It added that ThetaRay's algorithms were effective in weeding out unnecessary risk alerts, and in finding new risks by discovering new transaction patterns, or the "unknown unknowns", as Ms Yuen put it.

The move by OCBC comes as financial crimes have been growing in scale and complexity, Ms Yuen pointed out. She cited a PwC report that estimated global money laundering transactions to be equivalent to 2 to 5 per cent of global GDP, or roughly US$1 trillion (S$1.36 trillion) to US$2 trillion annually.

OCBC also noted that its chatbot application called "Emma" has closed more than $70 million in home loans since the start of the year.

It took three months for "Emma" to be fully trained to address all possible questions asked by consumers about home and renovation loans, with one of the top three questions being how much a consumer can borrow.

"Emma" is trained to identify and associate with all terminology used in the process of applying for or refinancing a home loan. As new or revised regulations come up, "Emma" can be updated to respond to new questions, the bank said.

A version of this article appeared in the print edition of The Straits Times on November 08, 2017, with the headline 'OCBC sees AI payoff in transaction monitoring'. Print Edition | Subscribe