Latest posts by Maureen Mutua (see all)
- Compliance With A Commitment And Employee Moral Obligation - March 20, 2019
- Biggest KYC Failures In 2018 To Avoid In 2019 - March 15, 2019
- 2019 Trends For Anti-Money Laundering And Terrorist Financing - March 14, 2019
Key fields containing irrelevant or incomplete information – for example: the ‘Source of Wealth’ being described as ‘savings’, ‘personal wealth’ or ‘husband’; the ‘Description of Business’ field containing a lot of ‘buzz phrases’, but no reference to key information such as a customer’s business sector, products or target markets; or the date of incorporation of a company not matching information given about the customer’s business or banking history.
Inconsistency between different parts of the same form – for example: “expected monthly turnover” figures not matching the “expected net annual profit” figure. Stated countries of operation not matching the countries from or to which funds are expected to be received or sent; and declarations by the customer that it conducts no operations involving sanctioned countries, conflicting with stated information on where clients and suppliers are located.
Information which casts doubt upon the credibility of a customer’s stated business model, or in some cases, the customer him/herself – for example: An import/export business stated to be operating from “flexi-desks” or shared spaces inside hire-by-the-hour office facilities, i.e. no apparent owned or leased premises in which to store goods; or a business with an annual turnover of £10M equivalent, but employing only four staff; or stated involvement in a project or market which is known to be heavily regulated, or otherwise to have very high barriers to entry for a business bearing the customer’s profile; or revenue projections of £ X Million from clients in a country where, from publicly available data, no companies exist with balance sheets large enough to make such payments.
At great expense, and in order to comply with the law, financial organizations are employing increasing numbers of KYC specialists to read what the relationship managers or the customers themselves have written on the CDD forms and sort out these errors. People who work in those roles will tell you that a great deal of time is spent dealing with these same types of issues. And this is before you get to the business of curating evidence from multiple sources to verify identity and address and establish UBO – issues on which a lot of industry effort is already being expended.
The negative outcomes are well known to those who manage financial businesses; high rejection rates, lengthy onboarding times, rising compliance staff costs and an inexorable drift of responsibility and accountability for compliance matters away from front line, customer-facing staff, towards second line, back-office risk staff.
What if a machine could do most of this work instead? What if, instead of KYC staff spending their costly time telling RMs that the source of wealth information did not actually reveal the customer’s source of wealth, or that section 4 contradicted section 5. A succession of computer science breakthroughs now brings this ability into reach. The most transformational of these breakthroughs is the impact of deep learning on Natural Language Processing (NLP), which is the art of teaching computers to understand human language. Deep learning allows accurate understanding of human language, taking into account the subtleties of nuance and context.
Augmenting human capability with machine reading is a perfect fit for the document-intensive world of CDD. Detailed investigative work should be the focus of human investigators, aided by AIs’ ability to find patterns and evidence of coordinated attacks across many accounts or applications. The aim is to alert compliance staff to the possibility that either the customer or the RM, or both, might be ‘gaming the system’ by providing details that are replicas of other customers, or which are demonstrably false.
Featured image: Pexels