Weekend Reading: Beyond Systems and Spreadsheets — The Human Element in Banking Supervision
By: Erich Hoefer , Co-Founder & COO of Starling
Last week, I had the privilege of hosting a fireside chat during SupTech Week, an annual conference hosted by the Cambridge SupTech Lab . The Cambridge SupTech Lab is a key resource for the supervisory technology (SupTech) industry. In this capacity the SupTech Lab has done remarkable work developing programs to help accelerate the digital transformation of financial supervision and supervisory agencies.
We were fortunate to feature Simone di Castri, PhD , Co-Founder and Co-Head of the SupTech Lab, in the Closing Comments of our 2024 Compendium in which he discussed the profound impact that technology has had on the supervisory landscape.
I was excited to be joined by one of Starling’s Industry Advisors, Mirea Raaijmakers . Mirea is an independent advisor on culture transformation and behavioral risk. In addition to leading behavioral risk at ING for several years, Mirea was one of the pioneering architects behind the Dutch central bank's groundbreaking supervision program for behavior and culture. In that capacity, she has had an outsized influence on the way culture and behavioral risk are supervised globally.
Our discussion covered how advances in artificial intelligence and behavioral science are enabling new capabilities for banking supervision. Financial supervision is at an inflection point, as our CEO, Stephen Scott, has shared in recent Weekend Reading articles. It is becoming increasingly clear that we are seeing a fundamental shift in the role of the supervisor. As Stephen explains, “After the Financial Crisis, cultural configurations shifted to espouse the agenda of ‘safety and stability’ that has shaped macro-prudential policymaking over the last 15-years. Today, cultural configurations have shifted yet again, and yesterday’s policy consensus has been displaced by today’s calls for a shift in policymaking that champions ‘growth and competitiveness.’”
Supervisory bodies that fail to adapt to this new reality risk further eroding the trust of a public already frustrated by the perception of repeated financial system scandals and supervisory failures in the lead-up to the 2023 bank failures. But this change in priorities will require regulatory and supervisory bodies to embrace innovation. Simply doing more of the same is unlikely to be looked at favorably.
Elizabeth McCaul , then a Member of the Supervisory Board of the European Central Bank (ECB), contributed the Preamble to our 2024 Compendium in which she reflected on her experience in driving the adoption of innovative tools throughout Europe’s Single Supervisory Mechanism (SSM). “Effective supervision needs to be able to identify whether the good governance and risk culture that are vital for banks to successfully adapt to and effectively manage changes in risks in their operating environment are in place," McCaul wrote. "Banking supervision needs to harness the benefits that technology can offer given the rapid rate of technological progress currently underway."
While the relevance of ‘safety and stability’ has fallen in the public’s mind, history suggests that it is only a matter of time before a crisis causes the pendulum to swing the other way. This is all the more reason to carefully examine how supervision can leverage technology and innovative approaches to support ‘growth and competitiveness’ objectives without encouraging conditions that may lead to the next crisis.
Recent advances in AI and behavioral science create an opening for SupTech to fill this need.
The origins of behavioral supervision
In the aftermath of the Global Financial Crisis, regulators predictably focused on strengthening capital buffers, liquidity requirements, and resolution planning. These measures were necessary but insufficient. As Mirea noted during our discussion, De Nederlandsche Bank (DNB) reached a critical realization: "It's not just the organizational structures or the balance sheet that will give you information on where an organization stands or how healthy it is."
This insight led the DNB to take what Mirea called a "bold decision" – bringing behavioral scientists into the supervisory fold. While accountants and economists could analyze balance sheets and organizational structures, understanding the human elements of risk required an entirely different set of expertise and capabilities.
The DNB's approach started with the recognition that behavior at the executive level – particularly in decision-making and stakeholder communication – played a crucial role in organizational health. As Mirea explained, "The behaviour of key decision makers in the organization, at the executive level... the way they lead the company, the way they communicate and build trust in their organization, build trust towards the stakeholders... all these types of behaviours drive performance and integrity as well, right next to the balance sheet."
The DNB realized that simply adding more rules to existing regulations wasn’t going to make a difference in outcomes, Mirea recounted.
This ultimately led to the development of a comprehensive "supervisory toolbox" that helped regulators identify and assess behavioral risks. This included tools to recognize when decision making is unbalanced or if commercial interests are regularly superseding regulatory safeguards. In other cases, the tools could reveal how an organization responds to leaders who push too hard or pursue decisions that are unrealistic.
The evolution of culture assessment
Measuring and assessing culture presents a different set of challenges. Traditional approaches to assessing organizational culture rely heavily on employee surveys and occasional executive and board meeting observations. But, Mirea argued, these methods barely scratch the surface. The real challenge lies in understanding the informal networks and trust relationships within the body of the organization which determine how work actually gets done.
We discussed the tendency among management teams and supervisors alike to focus on policies and processes when assessing banks for risk and why this only reveals part of the picture. "You can have a perfect process on paper," Mirea explained, "but how it operates depends on how the people in that process act and interact with each other." This is particularly evident in the Three Lines of Defense model, where effective risk management depends on smooth information flow between different individuals and functions. As Mirea noted, "The trust levels between those different lines of defense can be very low and they hamper crucial information sharing, which in the end, at the end of the decision making process, you will get much lower quality in terms of the decision."
The DNB's early work focused primarily on executive-level behavior and board dynamics. The approach included interviewing managers, as well as their direct reports, and conducting systematic behavioral observations to understand how people act and respond to their peers and their leaders. While valuable, this faced scalability challenges when it came to extending such methods beyond the leadership team.
This becomes particularly challenging when dealing with complex, high-risk decisions that can't be automated. While many routine decisions can be systematized, the most important risk decisions require effective human judgment and collaboration. Understanding the behavioral dynamics that enable or inhibit good decision-making and determining how to scale that across the enterprise is the critical challenge in culture supervision.
The SupTech opportunity
The ability to scale culture assessments across the enterprise is the natural evolution of current cultural assessment approaches. Fortunately, this is where recent advances in technology can offer solutions. Mirea emphasized that what supervisors really need is “technology that is driven by behavioral science." Modern AI and machine learning technologies offer unprecedented capabilities to sift through massive internal corporate data sets to detect signals that tie to culture risk at the scale of the modern enterprise.
This represents a departure from traditional workforce monitoring and surveillance-based approaches that are designed to capture employee activities that indicate an event has already taken place. Rather, predictive behavioral analytics reveal the underlying drivers of risk so managers have an opportunity to prevent such events from happening in the first place.
Our discussion turned to the different ways such tools could be deployed in a supervisory context depending on the nature of the inquiry and whether the review is taking place offsite as part of an initial review or during an on-site risk assessment.
One method that does not necessarily require proprietary data is to analyze publicly accessible data that are regularly generated by firms and their employees. So called ‘unobtrusive data’ approaches use natural language processing to analyze public documents and employee information on social media platforms to provide general insights into the culture from an external perspective. While it is important not to draw too many conclusions from such analysis, scanning unobtrusive data sources can be a useful first step in identifying areas of culture that warrant additional analysis.
Obtaining a deeper understanding of how an organization functions generally requires gathering internal data. This may include communications, GRC, and other operational systems. In addition, as powerful as AI may be, there will always be a need for more traditional culture assessment tools like interviews and behavioral observations. Mirea emphasized that understanding culture requires both "numbers and stories" – quantitative metrics and qualitative data working in tandem.
Initially, this analysis may focus on understanding the actual patterns of interaction within the organization and what that reveals about networks of trust and collaboration. Mapping information flows both within and between departments allows managers to identify informal influence networks, assess the integration of control functions, and evaluate the effectiveness of frameworks like the three lines of defense on a continuous basis.
Further insights can be gained by using machine learning to identify signals that tie to specific behavioral patterns of interest. These patterns can then be layered onto the internal networks of the firm and tied to the bank’s formal processes and policies to understand where there may be gaps. Depending on the application and technology deployed, this can even be done without needing to access the content of communications — instead focusing on patterns of interaction and their outcomes.
Looking forward
In his Closing Comments to our 2024 Compendium, di Castri explored the potential for such tools to revolutionize culture risk governance and supervision. "SupTech tools powered by developments in computational social science now allow us to put data-driven, quantitative analyses to the qualitative challenge of managing cultural contributors to organizational conduct," di Castri explained. "Such capabilities allow for superior risk governance, at lower cost, and allow management to provide more timely and reliable reporting up to their boards and on to investors and industry overseers."
By reducing reliance on formal structures of processes and controls and incorporating visibility into informal cultural and behavioral drivers, financial supervision can offer firms more flexibility in conducting business without compromising on effective risk oversight.
For supervisory agencies looking to develop their behavioral assessment capabilities, Mirea offered practical advice: start small and focused. "Take three key behaviors," she suggested, "like sound decision-making, leading by example, or effective collaboration." Once identified, you can then seek out existing data sources that might provide insights into these behaviors, before considering what additional information might be helpful. This measured approach allows agencies to build capability and confidence gradually.
“If we think about innovation, if we think about growth, if we think about ESG, cyber risk, AI, and the geopolitical situation, it means having to deal with everything changing so quickly,” Mirea explained. “[The] capability to zoom in on culture and behavior should be part of the portfolio of capabilities that supervisors have available.”
This piece first appeared in Starling Insights' newsletter on December 15, 2024. If you are interested in receiving our thrice-weekly newsletter, among many other benefits, please consider signing up as a Member of Starling Insights.