AI and ML: revolutionising financial regulatory and supervisory activities

Author: DIANA PAREDES Chief Executive Officer & Co-founder - Suade Labs

Recent events involving SVB and Credit Suisse have made the conversation around AI and ML more relevant than ever, demonstrating that their use can be crucial to efficiently process large amounts of data and perform risk assessment for supervisors. One wonders if through them, information that is difficult or impossible for humans to detect manually could have surfaced in time to avert the crisis the banking system is currently experiencing. In April 2020, the Financial Stability Board (FSB) published a report where it stated that “the use of AI and ML in financial services is at an inflection point, moving from a period of experimentation towards widespread adoption.” Indeed, the use of AI and ML in financial regulatory and supervisory activities is still in its early stages, but it is rapidly gaining momentum – notably since the COVID-19 pandemic - as data becomes increasingly available, algorithms improve, and more regulators and financial institutions recognize the benefits of these technologies. Regulatory sandboxes and innovation hubs are being established to provide a safe space for regulators, financial institutions, and technology firms to collaborate and experiment with new technologies, helping to accelerate the development and adoption of AI and ML, further driving the trend towards their increasing use in regulatory and supervisory activities. How AI and ML are bringing value to regulatory and supervisory activities Regulators and financial institutions are leveraging AI and ML to improve reporting, compliance, risk monitoring, AML and CFT checks, stress tests, capital assessments, and more. These technologies bring significant benefits to regulatory and supervisory activities, improving their risk assessments and compliance monitoring accuracy and, ultimately, the safety and stability of the financial system. They are notably very useful in identifying outliers. Thus, their more intensive use could have helped, in recent events, to detect risky lending practices and disproportionate exposure of assets to rising long-term interest rates, allowing regulators to intervene earlier and perhaps prevent a bank collapse. Similarly, the use of AI and ML could have been more effective in identifying and mitigating inadequate risk management practices, which, as we saw in March, lead to reduced client confidence and exposure to highly leveraged positions.

Lastly, the progressive automation that AI and ML enables also lowers costs by reducing the need for manual labour and frees up regulators to focus on more complex tasks that require human judgment and decision-making. However, it is important to balance the benefits of automation with the need for human oversight and expertise. Some regulatory and supervisory activities require human judgment and cannot be fully programmed. How to optimize AI and ML use for regulatory and supervisory activities To optimize the use of AI and ML for regulatory and supervisory activities, a number of conditions must be met. One key factor is data standardization, ensuring that data is properly categorized, labelled, and formatted to enable effective analysis using AI and ML algorithms. Hundreds of thousands of data points must be reported in an increasingly complex regulatory reporting system. Without a coherent structure of data in corporate repositories, a lot of time and resources are lost cleaning and interpreting that data instead of using it. Moreover, to effectively analyse data using AI algorithms, it is important to have appropriate access to data flows. This may require collaboration between regulators, financial institutions, and other stakeholders to ensure that data is shared in a secure and standardized manner. High-quality data is also essential for effective AI analysis. This includes ensuring that data is accurate, complete, and up-to-date, and that there are appropriate controls in place to ensure data integrity. Additionally, as AI continues to evolve and become more sophisticated, it will be critical to continue monitoring its impact and developing appropriate safeguards to ensure that it is used in a responsible and ethical manner. Effective governance and oversight structures, including developing clear guidelines and protocols for data use, are necessary to ensure that AI models are transparent and explainable. Finally, policymakers should also consider the potential impact of AI and ML on the workforce and develop strategies to address the replacement of jobs with machines or the lack of staff skills to work alongside these technologies. Indeed, AI and ML become more prevalent in the financial industry, some jobs may become redundant.