Emma Miller, Global Head of Strategic Partnerships, Data & Analytics, Refinitiv – a London Stock Exchange Group business.
Sophisticated technology solutions can help tackle the sheer volume of data that regulated firms manage as part of their AML endeavours.
The number of records in the Refinitiv World-Check database has increased significantly in the past two years by over 50% and the number of sanctioned entities and individuals has gone up by over 25% in a year.
Refinitiv’s 2021 Global Risk and Compliance Report shows that the appetite to invest in new technologies is strong – 86% of respondents agreed that technology has helped identify financial crime and 57% want to increase their tech spend, specifically on automation and digitisation.
Using artificial intelligence (AI) and machine learning offers significant advantages in terms of speed, efficiency and accuracy. But to work effectively, attention must first be paid to data - a critical ingredient in the success of any AI approach.
86% of respondents agreed that technology has helped identify financial crime.
To learn more about the applications and advantages of AI for financial crime compliance, download our eBook: How AI can improve the detection of evolving typologies driving financial crime. In it, Dr Janet Bastiman and Emma Miller discuss how use of technologies like AI can address current challenges in financial crime detection.
Data quality
Data quality can be one of the biggest hurdles in implementing and maximising the use of AI and machine learning driven solutions. And data quality isn’t just about completeness and accuracy, though of course they are important, it is also about understanding the provenance of your data:
- Where is it from
- How was it collected?
- What are the implications for using that data to train machine learning models?
This is especially important when sourcing data from third-party providers.
Refinitiv’s World-Check follows the most stringent guidelines for research methodology and inclusion criteria, applying rigorous quality control , and helping to ensure accuracy.
Managing or handling your data in line with regulations should also be considered.
If you are integrating different data sets, you need to understand where and how data is being stored and whether it is being transferred from one jurisdiction to another.
All this needs to be done within the boundaries of applicable data protection laws.
Data structure
A key dependency for AI and machine learning systems is the structure of the data. Ingestion, normalisation, and the combination of structured and unstructured data need to be considered - though some of the newer tools on the market are agnostic to data type and structure, which avoids the need for a costly and cumbersome data cleansing exercise.
You need to be able to specify how you want to slice and dice your data to the level of granularity and specificity you require.
For example, with the introduction of Special Interest Category feature in World-Check, you will now be able to get an even more granular view of the reason for the record’s inclusion and its associated potential risk or risks. Filtering data based on these categories will help you to focus on just what you need.
Combining structured and unstructured data can be a lengthy exercise but tools like Natural Language Processing are available to help find the most relevant pieces of information in context within unstructured data such as adverse media.
Data completeness
One of the significant advantages of tools like Napier’s Client Activity Review (CAR) is that it provides a single view of a customer. It is through the linking of data sets - in this case from know your customer (KYC) and transaction monitoring - which enables AI to spot signals that would otherwise be invisible to a human analyst or impossible to connect over multiple links.
For organisations to derive the best insights from this data, it needs to be complete.
We know that we are at an inflection point in the fight against financial crime. As an industry, we are asking difficult questions about effectiveness and the need to go beyond just meeting the technical compliance requirement.
It is easy to get lost in data, processes and documentation - though these are clearly important - and forget that the real purpose is providing actionable information to law enforcement to help them disrupt and fight crime.
Why data and machine learning are useful for AML
In AML solutions that utilise AI, usually the type of AI used is machine learning.
Machine learning is a branch of AI that is especially good at identifying things that are the same and things that are different in large patterns of multidimensional data. This makes it the perfect candidate for AML solutions, since both customer due diligence and transaction monitoring involve vast data sets.
Machine learning is particularly useful when used to identify items that cannot be detected by rules-based solutions, for reducing the white noise associated with false positive alerts. It uses algorithms to make predictions on data that it has not previously seen, which is what allows it to detect new financial crime typologies as they occur, not just known patterns of criminal behaviour.
Improve your compliance processes with an award-winning solution
Get in touch to see how our solutions can help your organisation transform your compliance; or request a demo to see it in action.
Image by Alan Warburton / © BBC / Better Images of AI / Plant / CC-BY 4.0
