AML regime change – The need of the hour

Published on July 28, 2020 by Jalaj Jonker

What is the current scenario?

“Today we were unlucky, but remember, we only have to be lucky once; you will have to be lucky always.”

This was the chilling statement made by the Irish Republican Army (IRA) soon after their attempted assassination of the British Cabinet at The Grand Hotel, Brighton, UK, nearly 36 years ago.

The situation is similar for anti-money-laundering (AML) professionals and financial institutions that have to be lucky every day at tackling money laundering amid the COVID-19 pandemic. Teams comprising CROs, CCOs, CTOs, MLROs and AML officers at financial institutions need to be especially prepared to deal with the current situation.

Amid the pandemic, with lockdowns and travel restrictions still in place in many parts of the world, it is imperative that the global financial infrastructure is not just robust to support millions of transactions on a daily basis, but also ensures proper financial sanitation. More so, to counter the myriad threats from money laundering, it is vital that the financial infrastructure is supported by platforms and tools so designed that they are technically and conceptually sound, and can also handle exceptions effectively and efficiently. The current period should not be treated as a ceasefire window because money launderers and criminals will not be deterred from trying to exploit any flaws or weakness in the financial system.

The present situation is unique because there has never been a situation in history when the thread of globalisation, supported by a vast financial sector, so tightly knit the world, and yet, a pandemic has affected every possible aspect of human existence.

With the pandemic affecting most of the global population, it is difficult to keep business operations running due to the unprecedented global economic slowdown, the drastically changing transaction behaviour of customers and the diminishing resources pool. From the point of view of preventing financial fraud, the importance of an AML programme cannot be overstated in the current situation. The platform must focus on possible behavioural aspects and should cater to the enhanced risk that institutions are facing currently. To keep an AML programme running smoothly for financial institutions, a series of validations, tests and checks, model enhancements and upgradations, platform maintenance and exception handling need to be conducted regularly. With the resource pool getting impacted in almost all major delivery centres, we need a programme that requires minimal human intervention and is capable of carrying out the minimum requirements to keep the system running. In order to achieve this, we need to opt for process automation. Automation is not restricted to basic repetitive tasks but also covers the more decision making- and exception handling-related tasks, and is more self-reliant. Consequently, the need of the hour is to migrate to a system integrated with machine learning.

Why we need a strong AML programme now more than ever

An AML programme is a necessity for any financial institution. In the current circumstances, a robust and superior AML programme is required, as most of the financial hubs have been impacted by the pandemic and there is a shortage of resources due to the economic slowdown. An AML programme is required to cater to three things: (1) support business as usual, (2) counter the new risk exposure posed by volatile trends in transactions, and (3) ensure scalability in the post-slowdown phase, when businesses scale up.

The current systems, built around a fixed quantitative model, are not very flexible, to say the least, and do not consider behavioural aspects. They already cause a lot of trouble due to the large number of false positives they generate and the rigid model-based approach they adopt. The model is rigid in defining the scope of work and does not provide for exception handling, resulting in a large number of false alerts. We expect a spurt in demand to support businesses to scale up in the post-slowdown phase.

Hence, now is the right time to build an AML programme and to be ready as a first-mover to capture the market with a solution that addresses the enhanced money-laundering risks and caters to daily business requirements.

How to build a stronger AML programme

When quantitative mathematical models were introduced to AML programmes, they empowered the transaction-monitoring platforms to deal with volumes. At this point, we need to focus on the qualitative aspects of a transaction, i.e., the behavioural aspect, or the intent behind the transaction. Current systems need to be supplemented with technology that can act as an anchor for a more robust and sensitive, yet accurate, system. Combining machine learning and natural language processing, an integrated AML framework can expand the scope of our current platforms and take them to much higher levels.

Using the aforementioned approach, we can cover a variety of tasks as we scale up from a single platform to a complete AML framework. Usually, activities such as recalibrating thresholds, segmentation, peer group deviation and understanding customer behaviour trends are tedious and recurrent. Integrating an AML framework with sophisticated technology will result in higher effectiveness and efficiency; this will be explained in future blogs in this series. Tasks such as risk assessment, including possible risk exposure, testing the validity of implemented controls, reporting suspicious activity, ongoing monitoring and testing, identifying new trends, training control and audit can be completely taken over or partially supplemented by technology.

How should we conceptualise the integrated AML framework?

The AML framework with machine learning capability should perform these functions in the following order:

1. Use transactional data to identify the risk exposure

2. Assess which risk areas and red flags are exposed and have no or inferior countermeasures

3. Perform segmentation and form segments

4. Use the information from risk assessment, segmentation and transactional data to identify the possible suspicious activity that may be occurring or could occur

5. Use the above mentioned information to devise models and rules for monitoring transactions

6. Devise a control system that provides feedback to the monitoring system that differentiates between true and false positives, and improves alert quality and control efficiency

7. Use improved efficiencies from the above mentioned processes to lay the groundwork for model validation and prepare for an audit

Since aspects of machine learning rely on historical datasets and transactional behaviour, the factors discussed above need to evolve with the AML frameworks of financial institutions.

Let us consider an example of how an AML framework integrated with machine learning will respond:

• Consider a potential change in transactional behaviour that would have a corresponding change in risk exposure; this change in transactional behaviour would be identified by the control mechanism and notify the compliance manager/ MLRO/CCO/CRO

• Based on this notification, the possible risk exposure will be identified, and to address the residual risk, it will be modified or new controls implemented

• After the risk is identified and proper segmentation carried out, an improved business logic will be implemented

• Along with the feedback on the true and positive alerts, a justification is provided for the false or true positive alerts

If machine learning is integrated with an AML framework as mentioned above, we could expect a robust, self-reliant and efficient system that can work with minimal human intervention and meet regulatory obligations.

How Acuity Knowledge Partners can help

Acutiy can help your AML program to identify, analyse the risk and can help you with developing an integrated AML program that can help you address your requirements. The requirements (not just restricted to) risk assessment, model validation, segmentation, tuning, identification of residual risk and conceptualizing the counter measures.

---

---