Analytics & Big Data
Aug 09, 2017

The new AML/KYC monitoring goal: turning 15 days into 15 minutes

The amount of AML- and KYC-related data that needs to be looked at these days has grown at a tremendous rate, reaching enormous volumes. To keep up, we will have to use technologies, digital capabilities, and analytics in a manner that we have never done before.

Why do I say that? Well, to start with, there is so much data that needs to be analyzed, and so much of it needs to be analyzed in near real time. Those driving hard for compliance don't want to lock up some of their best customers for 5-15 days while waiting to see whether they are actually doing some suspicious activity or not. With the amount of false positives that come, you just can't afford to have a whole bunch of your customers unable to transact while you are reviewing their earlier transaction.

In an ideal world (admittedly, we're not there yet) you want to be able to figure out one way or the other whether it's really activity that qualifies as suspicious and get back to the customer in, say, 15 minutes. Another aspect in need of examination: you need a huge number of people to be able to look into and take care of these transactions. Unfortunately, the amount of qualified bodies required aren't there to the degree necessary. The shortfall in adequately trained personnel has led to a work backlog that is getting bigger and bigger, which brings us to still another great challenge that must be faced.

The ability to get your arms around the highly varied types of data that need to be monitored is making it even tougher to rely on humans to do the job. Now, there will clearly be parts of the process that can't be digitized. Take setting up procedures and policies. And there will be an intelligent end of the workflow for which you still won't be able to train machines. But what that's going to do is create another interesting dynamic. On the people side, the kind of resources you'll need will have to be even better trained. That's because they will have to be able to handle the top of the pyramid while much of the lower end work of today is eliminated through advances such as machine learning.

So, how do we make that 15-day-to-15-minute shift? How we rethink our relationship with data, it turns out, will have a lot to do with it. First, we must understand where we are at this very moment: The volume, variety, veracity, and velocity of the data — or four “v," if you will — is so large that, again, you simply cannot rely on human heads or hands to adequately manage the load. Because of this, by necessity, machine learning now allows us to do something very interesting: it enables us to detect patterns in huge amounts of data. The larger the data set the better, in fact.

And so, where not too long ago we might have worried about choking on a backlog of data, those same gobs of information now, strangely, promise to liberate us. Take the potential to be seen in the hidden when it comes to “computer vision," which is able to detect patterns that the human eye cannot detect. This vision gives machines the ability to "recognize" images, that is, to parse them in certain subtle and sophisticated ways. Practically speaking, you could use the tech at one end for image processing and image recognition, and at the other for handwriting recognition.

AML and KYC are both about images that need to be screened, scrutinized, and validated. Computer vision can be a true game changer in this regard. It can make sure that we get the right kind of validation from ancillary documents in a way we aren't able to today. The point? Put AI atop machine learning and computer vision and you quickly find where you can observe, analyze, and deduce from data much faster and easier, and with greater reliability over time. The time is now and the need is real.

About the author

Rohit Tandon

Rohit Tandon

Business Leader, Analytics

Follow Rohit Tandon on LinkedIn