How banks can use machine learning to improve risk management
INSIGHT ARTICLE |
Financial services firms are being challenged to develop innovative tools and frameworks, to adopt data-driven risk management and real-time monitoring, and to discard the ‘rearview mirror’ approach to operational risk management.
RSM has created a proven methodology for identifying and acting on emerging risks throughout a bank’s portfolio. It’s done with machine learning, which allows a bank to assess an array of risks and identify them at an early stage.
Consider the example of an RSM client, a global financial services organization that wanted to better assess its operational risk exposure. We introduced the company to predictive risk intelligence and advanced analytics in order to help managers identify areas of potential risk and the need for action.
Then, using a proven blueprint, RSM worked with the client to develop a pilot and establish an operating model. The process can be broken down into four steps:
Step 1: Establish machine learning literacy
The concept of machine learning is still new to many people, even those in the industry. Therefore, RSM trained a non-technical team in order to empower them to assess opportunities for the use of analytics.
Step 2: Assess the opportunity
The next step was to figure out the best use for the pilot program, known as an opportunity scan. Following a formal methodology, a series of workshops established these use cases. They were rated on criteria including business process fit, complexity, data availability and business value. Using a tool designed by RSM to visualize the use cases in a dashboard, RSM helped the client more easily identify the most effective projects. Then a final recommendation was made.
Step 3: Develop a pilot
Then came time to create a pilot. RSM’s engineers collaborated with the client’s engineers to develop a highly accurate model. The team focused on provisioning training and test data, selecting the right algorithm, improving data quality, executing against training data and deploying a pilot.
Step 4: Implement an operating model
The final step was to integrate the model into existing operational risk management processes. This is the payoff—where the value is created. Moreover, the operating model puts in place the governance and processes needed to create an appropriate budget, focus the investment on highest-value use cases, define required roles and responsibilities, provide a how-to guide and implement appropriate data governance. In the end, it allowed the team to build an information architecture for machine learning that offers choice and flexibility.
Outcomes and benefits
RSM’s approach and expertise in building the foundation helped the organization leverage data to unlock new value and viewpoints. Introducing new tools, an integrated data strategy and a governance framework provided the organization with the ability to adopt data-driven risk management and real-time monitoring, thereby discarding the ‘rearview mirror’ approach to operational risk management. These outcomes gave risk officers a head start in identifying potential risks and the need for action.