By Tim Blackmore
Ever since computers were invented, we spoke their language. To make a computer work we asked them questions in code they understood. So that the computer could get those answers we structured our language in ways that were easy for computers to ‘find’ the information, we call it data.
So far this approach has worked fine, but today criminals are sending billions of Dollars around the world with the core intention of legitimizing their illicitly gained finance by laundering those ill gotten gains. Recently Estonia, announced that Money Laundered through its banks per annum had exceeded its annual GDP. This is astonishing, but how many other Countries have similar problems? The fact that almost one in three of all banks have been fined by their regulator for breaching some form of compliance regulation is perhaps an even more illustrating indication of the breadth and scale of this crime.
Has the compliance industry reached a point where the current systems can no longer cope with the sheer volume of checks made necessary by the 180,000 compliance rules and regulations currently in place. Certainly, most banks can point to the fact that the average compliance department ten years ago, was a desk on the 10th floor and today it fills the entire building.
What has not been discussed is what changes those compliance systems that served the compliance department ten years ago have done to change and enhance their system to match that explosion of need. In truth, most vendors talk about AI as being the answer, but not many actually utilise AI and certainly not many use groundbreaking technology like Natural Language processing.
With NLP, computers are finally learning our language. NLP does not need information to be structured, in fact NLP works better without structured data. This is a key difference, why? Because all other AI needs data to be neatly placed in fields where the model can read the information, the model then needs to be trained how to read these fields and make sense of them. But to be effective it needs lots of this data to be available. Not all banks have this amount of data.
Meanwhile the money launderers are running rings around the banks and causing them to lease another building. So, why do the Banks need so many staff when these AI based systems are supposed to do most of the work? Simple. Because they keep on getting confused with the data and get it wrong. These are called False Positives. True Positives are when it gets it right. The problem being is that hundreds of people are necessary to find the trues positives out of all the false positives. That is why you need a building..
Money launderers change their tactics every day and no matter how well the older systems structure the data, the criminals find ways to outsmart it. The only solution for older AI systems is to broaden the search this is the reason which causes so many false positives.
Why is NLP important in Compliance?
NLP, works better simply because no matter how the data is re-structured, NLP still finds the crime. Or at least, because there is no need to broaden the search, it creates far less false positives.
How does it work? To simplify, it reads the data the same way we would. Human perception is able to dismiss false positives because we make intuitive leaps. Rue de Tehran in Paris is NOT Tehran the City in Iran, Victoria is either a first name or the name of a City or suburb, it is not ‘Victory’ the name of a Sanctioned Vessel, name distance and fuzzy logic would have picked Victoria, wrongly, as a true positive, NLP would have released it as a true negative.
Simply put, this is the key differential, and not only can we prove that this works there are three tier 1 banks in Europe, USA and India benefitting from Pelican’s ability to massively reduce their false positives. We are currently averaging an 80% reduction.
It gets better.
We have also worked very closely with two of those tier 1 banks on reducing the current false positive rate with 3rdparty Sanctions Screening systems. We have put a small overlay on top of their current incumbent systems and produced an 80% reduction. This means that not only has their investment been secured and in fact enhanced, but their staff have also not needed to be retrained, IT has not needed a massive new implementation project during these Covid years and they haven’t needed to take out that new lease on a building. In fact, they are looking to downsize.