Caspian were recently invited to present at an industry roundtable, organised by UK Finance, on establishing the ultimate beneficial ownership (UBO) of corporate customers and counterparties, and the challenges of scale in particular.
What challenges does the financial services sector face?
The problem of determining UBO is becoming larger and more complex under the pressure from many industry drivers including:
- Money laundering networks get more complex and sophisticated, so understanding the threat means analysing the risk from entire networks rather than individual entities. Graham Barrow has written a great case study illustrating this point
- Financial institutions need to continually monitor risk, throughout the customer lifecycle, in response to changing patterns of customer behaviour and inline with normal business activity, and not just during onboarding or periodic review.
- Many institutions want to take a more ‘holistic’ approach to mitigating Machine Learning risk, breaking down silo’s between business units, products, and internal sources of data, so that they can get the complete picture of the activity of customers through the whole business
- At the same time UK, EU, and international statutory regulations are raising the bar in terms of thoroughness and sophistication of the investigations and checks required by financial institutions.
What are the opportunities to improve our approach to UBO?
At the same time as facing challenges, we must also recognise the great opportunities to improve how we conduct UBO as an industry, particularly through increases in the quantity and quality of data available. More national company registries are moving online. Initiatives, such as the Fifth Money Laundering Directive, will increase account data sharing between institutions. And Companies House are consulting on ambitious plans to increase the transparency of UK corporate entities through better verification, data sharing, and anomaly reporting.
We are therefore faced with bigger and more complex problems, that have to be solved faster and more thoroughly, using more and more sources of data. That will require fundamental changes in how we conduct UBO checks in all parts of AML: onboarding, KYC, counterparty and enhanced due diligence, account and periodic review, and investigating transaction alerts. Of course automation will be part of this but, more fundamentally, we need solutions that are scalable. Traditional AML processes, both human and automated, are non-scalable: they get more cumbersome and expensive as the problems get larger. Scalable solutions, on the other hand, aren’t simply ones that are able to cope with larger problems, but rather they get better as they grow.
Scaleable technology solutions
Google Search is a great example of a scalable solution. In the earliest days of the web, search engines would painstakingly maintain hand-curated registries of useful information: of course these solutions couldn’t cope with the explosion in the size of the web in the mid-90s. Search engines that used rule-based artificial intelligence to curate and rank pages automatically did better at coping with the sheer volume, but couldn’t cope with the increasing diversity and sophistication of information available. Then Google’s PageRank algorithm revolutionised web search by providing the first truly scalable solution for dealing with the size and sophistication of the web. It used links between pages as clues to their quality, rather than relying on a fixed rulebase, and it used feedback from users to score the quality of its results. It turned the increasing volume of information and human usage into an advantage rather than a problem. The result was a virtuous feedback loop in which the quality of the search results inexorably increased as the web, and Google’s usage, grew. The more it was used, the better it got; and the more useful the results then the more it is used.
Financial investigation technology
Caspian are building truly scalable solutions to the problems of AML and UBO, based on our Financial Investigation Platform (FIP). The key to FIP is to make better use of human expertise, to create virutous feedback loops, inexorably raising quality as the sophistication of the problems, and range of data sources, grows. We start by working closely with human experts, in-house analysts, and independent subject matter experts, translating their understanding of the risks that lie behind the data into software agents that can deal with volume and velocity of cases now required. In many cases these can automatically mitigate risks, providing decisions at better-than-human levels of accuracy. But, just as important as the decision itself, FIP can produce human-readable explanations of those decisions.
The power of explainability
Explanations are vital to building scalable solutions, to closing that virtuous circle from volume to quality, in three ways:
- First, those explanations are of sufficient quality to satisfy in-house analysts, risk stewards, and external regulatory bodies. There is little point generating automated decisions if a human analyst is required to re-do the work required to generate the explanation of the reasoning that lay behind the decision.
- Second, explanations guide human analysts’ investigations in those cases where an automated process can’t make a decision accurately, pointing out what information is missing or what the next steps are in getting a resolution.
- Third, explanations are crucial for getting feedback and guidance on decisions from QA processes and top authorities. New threats and new data sources means that the AML processes need to be continually learning and improving; and if those in-house authorities can’t see how decisions are made, then it makes it harder for them to raise quality in the future.
The first AML systems built on these principles are being used by a Tier 1 global bank, reducing overheads and increasing quality. And those principles have been worked into the FIP platform to provide targeted solutions. Our Transaction Investigation solution is focused on transaction alerts, whilst Entity Investigator uses public data sources to assess the risk represented by corporate entities with approximately 95% of the accuracy of industry-leading experts.
For more information on Caspian solutions, contact us here.