Keith Berry, General Manager Know Your Customer Solutions, Moody’s Analytics
We are entering a new era of technical innovation in compliance, with further advancements in Artificial Intelligence (AI) bolstering solutions designed to fight money laundering. Financial crime is an ever-changing beast, and AI – in its various forms – provides the power to tackle it head-on.
In a global, digital economy, the magnitude of data required to effectively identify and mitigate money laundering risks can be daunting. However, through machine learning, able to process vast datasets, and AI, increasingly able to perform quotidian tasks, compliance teams can gain more control, answer risk-relevant questions, and enhance AML due diligence processes.
Machine learning (ML) algorithms have been in commercial use since the early 2000s, supporting data processing and modelling. ML and pattern-recognition AI, with the ability to swiftly analyze large datasets, has prompted faster alerts to compliance professionals. Regarding specific Anti-Money Laundering (AML) efforts, AI has already begun enhancing screening activity. And the advent of Generative AI (GenAI) is providing the next set of assets for compliance teams to improve AML procedures – co-piloting between humans and machines in Q&A investigation threads.
Crucial to ML and machine learning driven AML solutions is the data they are trained on. Data needs to be current, accurate, and unbiased, to ensure compliance, ethical outcomes, and good decisions are made. Embracing ML and AI innovations trained on quality data and with humans fully embedded in the process will enable anti-financial crime teams to forge a formidable defense against money laundering.
AI in action
AI has already been deployed for practical application in many instances. It’s used to automate mundane tasks such as data capture and document translation. For these tasks, Optical Character Recognition (OCR) scans documents, and Natural Language Processing (NLP) extracts key information from them, which streamlines operations.
One area where AI is increasingly being adopted is the entity identification processes for ultimate beneficial owners and other know your customer (KYC) checks. Leveraging Machine Learning classification and scoring methods, AI is playing a pivotal role in categorizing false positives – often the bane of a compliance professional’s life.
Large Language Models (LLM) present new possibilities for AI’s use in compliance processes due to their ability to predict and verify text. This means vast amounts of text, for example held in regulatory documentation, can be summarized and cross-referenced with other text to find differences in an instant, reducing a compliance team’s workload.
One interesting and exciting use case in AML investigations is the “AI assistant”, a research function that can make enhanced due diligence processes smoother, more efficient, and more seamless during onboarding and ongoing reviews.
Transaction monitoring is another space where AI is a valuable tool, as ML classification models can aid in identifying suspicious and non-suspicious transactions. This process used to be done using scenarios and human judgement, but now with ML utilizing data sets the results are automated and more accurate.
The impact of AI is undeniable, particularly in alleviating the burden of time-consuming tasks, allowing compliance professionals to focus on critical decision-making. As AI continues to advance, its transformative influence on compliance processes will grow, opening new frontiers in efficiency and accuracy around AML activity.
Human judgment and AI efficiency
AI undoubtedly brings a host of advantages to the fight against financial crime. Nevertheless, to maintain accountability, upholding ethics, and provide the most intelligent AML solutions, AI needs to be complemented by human judgement and oversight. People are more instinctive; able to identify new types of criminal activity – things that don’t pass the “sniff test” – and they understand nuances that a machine may not see in human behavior. Also, unlike AI, people will be held accountable and face consequences for AML failures, and they are responsible for providing evidence of compliance procedure.
As AI is increasingly integrated into compliance automation, there is a greater need to foster trust, transparency, explainability, and governance. This means not accepting AI at face value but ensuring it is being used where there is evidence-based proof that it is effective. Evidence-based proof will be essential for explaining how models are trained, make decisions and avoid bias. As well as regulators, customers will also need to be assured of the accuracy of financial compliance. It allows responsible AI usage, ensuring greater efficiency while safeguarding the integrity of the financial system.
Relying solely on AI models without adequate human oversight poses potential problems. Striking a balance between compliance efficiency and human expertise is crucial. This involves recognizing the mutual learning potential between humans and AI models.
AI’s true strength lies in its ability to simplify tasks and enhance human expertise within the complex world of AML. Embracing innovation can enable compliance professionals to better apply their skillsets to nuanced investigations and important decisions. Ethical reasoning and contextual understanding, vital in intricate AML scenarios, are best delivered through a hybrid solution.
The role of regulators
AI is drawing huge attention in the compliance space, while regulation pertaining to its use is still evolving. Regulators are, by and large, however encouraging of firms exploring AI’s potential to detect financial crime. The appeal lies in AI’s ability to gather information, uncover risk, reduce false positives, and highlight suspicious activity amidst the number of alerts generated in AML procedures.
Although specific regulations governing AI in compliance are yet to be formalized, in the US, regulators have issued guidelines supporting institutions in exploring AI’s capabilities. The guidance highlights that such exploration does not automatically trigger heightened regulatory scrutiny, even if areas for improvement are revealed. Moreover, the Financial Action Task Force (FATF) acknowledges the value of AI in real-time data analysis.
To earn regulators’ trust, transparency is critical. Organizations must be able to explain their use of AI, and regulators must have visibility into AI models and decision-making processes. Naturally, institutions seek to understand how AI reaches decisions, its training process, and the measures taken to mitigate bias. Properly executing this balance and demonstrating control is vital, as failure to do so could lead to fines and reputational damage.
Optimizing AI in AML
Within AML compliance, recent developments in AI have emerged as game-changing. It holds the promise of elevating risk management and screening capabilities to new heights. By integrating AI, organizations can access and process risk-relevant data, swiftly detect threats, and navigate complexity.
The efficacy of AI lies in the quality of data it leverages, alongside the symbiosis between artificial intelligence and human expertise. It’s in this partnership that the financial industry can rise to the ever-changing face of financial crime.