Hans Brown, BNY Mellon's global head of innovation, tells Finextra that the bank's operations team has applied AI to the relatively mundane task of reading e-mails and auto-forwarding them to the relative department.
Brown explains how BNY is using natural language processing (NLP) to read the e-mail, determine what the enquiry is about and which is the best team or department to forward it to, cutting down on "unnecessary administrative overheads".
He states that BNY's operations teams may receive around one million e-mails a year, each requiring someone to read them before deciding where it needs to be forwarded on to.
"That's not solving the enquiry - it's merely figuring out where in the organisation it needs to go so that it can be solved," Brown says.
"We're looking at how we can apply machine learning to determine what the e-mail is about - settlement, corporate action, tax, a query on a query etc - and then get it through to the right team.
"Even if it only takes a human a second to read an e-mail and forward it to the right place, you're still saving hundreds of hours a year given the sheer number of them that can be processed with this engine."
Brown estimates that the initial engine BNY have generated is able to perform this with a 90% accuracy rate for the most common queries. These account for around half of all the e-mails received.
While in some areas of business, 90% would be far from satisfactory - fraud detection for example - Brown believes in many of the mundane areas of operations, this is an acceptable level.
"Going forward, what we're going to do is get the less common queries up to that same level. We believe this generates better value than trying to improve the success rate for common ones to 99%."