Skip to content

Banking on Data

Prospect 33 looks at the clash between banking culture and data science in tackling financial crime and offers a solution to resolve this tension.

The investment banking industry faces a challenge – how to build a bridge between the data scientists with the technical knowledge to drive a revolution in Artificial Intelligence (AI) led automation and a culture that is often resistant to change.  

In many investment banks, especially in the corporate function, there is a mismatch.  It consists of two industries coming together and not quite talking.  There is a disconnect between the data scientists and the risk and compliance experts, between those who understand the new technologies needed to analyze vast swathes of data and those who understand how banks work.  It is only by bringing these teams together that banks can introduce the solutions required to tackle these conundrums, especially in challenging data intensive processes like financial crime detection and prevention.

We have two sets of brilliant minds, but because of outdated working culture and siloed work streams the move to automation for many financial institutions has been slow and ineffective.  External insight is needed, taking a holistic view encompassing people, technology, processes, and data, if new automated solutions are to be successfully implemented.  These solutions are urgently required to meet the need for increased compliance, in real time, without relying on ever increasing manpower and with it the risk of human error.     

Combining the strengths

The solution is simple to express but sophisticated to implement.  For banks to succeed in today’s data intensive environment, they need to innovate quickly, despite operating in a technological ecosystem that is often built on legacy infrastructure and processes.  Looking at Know-Your-Client (KYC) controls, it consists of taking a system that only 20 years ago depended on a single check for identity, a passport check and some form-filling, and applying multiple levels of data science to its operation. 

Let’s look at an example.  A few years ago, there was a scandal with a bank transferring $10bn worth of Russian Roubles outside the country and making a 20% loss on every transaction.  The auditors said this should have been flagged as suspicious activity because of the quantities involved (the objective clearly wasn’t to make a profit but to transfer currency outside Russia). An experienced banking professional may have spotted what was going on but only if they were able to interpret the data.   

This case highlights the difficulties banks face in implementing comprehensive Anti Financial Crime (AFC) controls and why the need for automation is so pressing. Complying with a bank’s Anti-Money Laundering (AML) policy, implementing KYC controls to verify the identity, suitability, and risks of dealing with a customer is time consuming.  The data analysis throws out false positives, typically three quarters of all alerts, when a compliant customer is flagged for enhanced checks because their name matches one on the sanctions or Politically Exposed Persons (PEP) list.  The time and money spent investigating false positives, with human intervention by senior analysts, is a severe drain on resources.

False positives can’t be completely prevented, but increased automation, combined with advanced analytics and AI, reduces the level of human intervention required.  AI solutions can spot patterns in data sets in trades that may appear banal, say volumes of mirror trades, that would take teams of people years to uncover.  New technologies, with a system of reliable signals and alerts, enable compliance teams to focus on a smaller number of cases for investigation where there is a greater risk of wrongdoing.

One of the reasons these solutions aren’t being implemented faster is the lack of timely, available, data.  Banks acquire other banks and their data is not interchangeable so information silos arise almost by themselves.  The data in one part of an organisation can’t talk to the rest.  The data professionals come in and try to fix this with little appreciation of the domain and the culture in which they are being asked to operate.  This is an area where consultants can add value, identifying where data blockages lie, by taking an overview of an organization rather than of a single workstream.  

Application of data

Data science has its place, but the analytics and application of data is what really adds value to the banking system and this is where it starts to get complex.  Analyzing data is one thing, but standards don’t stand still and neither do the people trying to subvert and work around these standards.  So banks need a hefty element of AI, systems that think for themselves rather than repeat processes, plus an element of machine learning in which the technology assimilates developments unaided.

Banks also still need that intuitive human element.  It was reported by regulators that the 2011 fraud at UBS could have been prevented by risk and compliance professionals monitoring amber flags in multiple areas, if they could have made meaningful connections between the data.  Back in 2011, this was a manual exercise, but with modern enhanced entity resolution and contextual data analysis these early warnings can be presented to supervisors, risk and compliance managers in order to take meaningful action.

The human factor

The only entities that can offer this level of experience are of course the humans in the loop (HITL).  We’re fallible but we are more likely than a machine to have an instinctive (for which you can read ‘based on experience’) grasp of what is a dangerous false flag and what isn’t.  The difficulty is that humans, as well as being fallible, can be costly.  No business wants to start building headcount unnecessarily, there must be a compromise.

Where to next?

What investment banks need is to find a way to introduce new AI led automation efficiently, understanding its role across the enterprise not just for a single point solution.  Institutions need to use the data scientists’ expertise to funnel a smaller number of potential risks for checking by the banking compliance experts.  This might be achievable from the inside, but we believe it is quicker to partner with an expert consultant.  

Prospect 33 understands not only data science but combines this with human capital management and banking advisory experience to provide a solution that works in the real world.  We not only know how to deploy data efficiently but where to illuminate the areas of a bank that needs it most, showcasing not only how data silos have arisen but also those that can cause the most damage and how to start to make them talk.

Banking remains a venerable industry but ‘venerable’ is only a step away from ‘elderly’.  Applying robotic process automation and then AI to the old paper-based processes is only going to make an out-of-date process go faster.  There is a clear need for fresh eyes, but these eyes need to understand the environment they are entering.  Prospect 33 believes there needs to be a bringing together of the banking industry and the data scientists to build the solutions needed to survive and prosper.

P33 Global Data Lab