How RegTech is helping to streamline AML
According to McKinsey, approximately $800 billion to $2 trillion is laundered annually through the global banking system. Criminals continue to use more and more sophisticated methods to launder money, putting the reputations of financial services companies at risk. Traditional AML models are no longer fast enough, or robust enough, to combat the more complex measures being utilised to avoid detection. As the number of threats rise, costs are soaring, too.
In order to keep pace, financial institutions are adopting innovative technologies and developing policies that increase the likelihood of detecting illegal transactions.
Regulatory technology, more often referred to as RegTech, is allowing organisations to automate more of the AML process, streamlining the way in which checks are conducted in a faster, more secure, and cost effective way than traditional AML models. This is particularly prominent in transaction monitoring, where these new AML models commonly use artificial intelligence (AI) and data network analysis to significantly improve effectiveness by detecting the riskier, more complex flows that traditional systems have a harder time flagging.
There are many benefits of adopting an AML model with RegTech at its core. By transitioning to AI and data based models, financial institutions can:
Identify more elaborate types of fraud
Data is enriched by using network analysis to connect companies, countries, and accounts to identify those with a higher risk of fraudulent transactions. These types of checks would be very difficult to perform manually, or at scale. Using network analytics, it is possible to examine the connections between entities, looking for similarities in known methods and typical behaviours that indicate a high risk of money laundering.
AI based models are capable of consolidating multiple risk factors at the same time, as well as their dependency on each other. Rather than considering a few factors in a rule, an AI model can assess a far higher number of dimensions on a transaction to extract a risk score.
Improve transaction monitoring
Using a combination of Natural Language Processing (NLP) and external data sources it is possible to add a further layer of screening which looks for transactions with similar characteristics at scale. It does this by analysing transactions performed by customers around the world, alerting AML compliance teams of irregularities and anomalies as well as observable patterns commonly found among money launderers.
AI driven NPL can also detect suspicious interactions between entities by reading, processing and deriving meaning from text, finding underlying patterns in the data, which would be near-impossible for a human to do. As an example, when looking at sentiment analysis as part of the screening process, NPL can analyse huge volumes of unstructured data from emails, news articles and social media posts – and is able to understand not only the content, but the context.
Reduce the number of false positives
By enhancing the precision of the rules to capture across a series of indicators which point to something being higher risk, rather than using a ‘blunt’, static behavioural rule, which captures only one element of the transaction, the number of false positives can be reduced.
The alerts from a data driven system generally are of much better quality – where the industry sees false positive rates of 95-99% for rules based systems. Despite being better at excluding false positives, there is still a lot to improve when AI is being used as the false positive rate is still high, however, it still outperforms manual checks both at detecting noise and signal.
For now at least, detecting false positives will be a hybrid model, but the data driven approach is helping to optimise the processes around these rules.
Reduce operational workload
Manual AML checks are time consuming and resource heavy with employees having to go through transactions rigorously to ensure compliance. While the human element cannot be completely removed from the process for now, RegTech can enhance operational efficiency, freeing up employees to focus on more valuable and complex activities.
Thanks to its ability to learn when exposed to new data without being explicitly programmed, Machine Learning (ML) is able to perform a deeper analysis of patterns that indicate there may be a higher risk of criminal behaviour, and feed this data back into the algorithm, further optimising the process over time and reducing manual intervention.
Speed up customer onboarding
The onboarding process can be sped up significantly using RegTech. Once all documents are submitted, information can be verified automatically, checking that ID requirements are met, while performing sanctions checks against the customer’s name (and people and entities relating to the business), and finally, calculating a risk score.
Low and medium risk clients can be fast-tracked through the approval process, while high-risk clients can be referred for further due diligence checks to be performed.
Be more flexible, tailored to B2B flows rather than retail flows
Rather than a one-size fits all rules based system developed with a retail flow in mind, a data driven model identifies which particular risk factors should be considered.
Transactions can be automatically screened based on an algorithm that looks for signals that indicate a payment is high risk, for example, by flagging a transfer from a specific jurisdiction or industry.
Challenges in RegTech Adoption
There are a number of factors holding back the application of RegTech to AML.
Regulation is just one barrier due to limited comprehension of how some aspects of RegTech (particularly AI and ML) are applied. The inner workings of these models are complex and may not be fully understood by regulators and compliance officers who are unfamiliar with how these models utilise data and algorithms to reach outcomes.
When an algorithm is capable of learning without intervention, it is difficult to define the logic being used, how the system works and how risk is being assessed. Internal investigation teams often want transparency that allows for a full audit trail that provides them with an easily understandable explanation of why something is a hit. There are already methods being developed to tackle this issue, helping to guide the investigator in what to look for if a transaction has been flagged.
When the success of a model is so driven by connecting the dots between data points, the ability to share data is incredibly important. Steps have been taken to implement AML programmes that provide a framework for information to be shared cross border, but regulations vary across regions, resulting in information being “siloed”, preventing effective global risk assessments of customer relationships.
Some data privacy regulations have also made data sharing more challenging to comply with, due to the extent to which data privacy now interplays with record-keeping, records retention, and security.
Many incumbents have been slow to adopt RegTech solutions into their AML processes, primarily as a result of legacy systems and infrastructure making implementation challenging. If relying on a third-party to provide these solutions, incumbents are typically more cautious and slower moving, with complicated procurement processes holding things up further. RegTech companies often provide solutions that focus on improving a niche area within AML, meaning that appointing multiple entities may be required to optimise the entire process.
Ultimately, collaboration is the key to achieving better AML compliance, but until these challenges are addressed, the benefits that come with using RegTech to streamline AML will remain limited.