AI-Powered Software Scales Model Risk Management for Nordic Bank


An unexpected result of the COVID-19 pandemic was increased model malfunctioning that heightened the importance of model risk management (MRM). As model usage becomes more widespread, reducing risk and complying with regulatory requirements demands increased testing.

A major Nordic bank needed to scale its model review capacity, so it came to Evalueserve looking for model risk management automation solutions. The bank’s MRM team lacked the capacity to take on more work and needed to hasten its process and keep up with regulatory needs. Evalueserve automated test interpretation and documentation for the client’s counterparty credit risk models.  

Our solution, MRMraptor, sped up model interpretation and documentation by more than 25% for the Nordic bank, shortening the overall model review process from a few months to a few weeks. 

The Challenge

The number of models in use at large banks is growing by more than 20% every year, according to the International Data Corporation. These models drive critical decisions for financial institutions and require rigorous stress testing, validation, reviews, and ongoing monitoring. In addition, model risk has been heavily regulated since the 2008 financial crisis.

At the Nordic bank, these trends have multiplied its MRM team’s workload and pulled experienced quants, many of whom hold PhDs, away from more valuable work. They were facing validation delays and backlogs as they simply could not keep up with demand.

The client needed to increase its testing and documentation capabilities, so it explored MRM automation in-house. After a few months of attempting to automate the documentation processes, they realized they needed external help and turned to Evalueserve. 

MRMraptor: an automated solution for model risk management

Our Solution

Evalueserve’s experts assessed the Nordic bank’s needs and implemented our MRMraptor software. We started by tackling counterparty credit risk because it was a must-have for the client. When not managed properly, the downside potential is huge, and the client could face significant regulatory fines. 

We configured the MRMraptor automation to interpret six counterparty credit risk tests and generate model review documents for their 17 forecasting models. This process used to take several months for the bank to complete manually, as each document is several hundred pages long.

The Nordic bank’s process using MRMraptor was as follows:
  • The Nordic bank’s process provided Evalueserve with the templates for their model review documents. 
  • The client ran tests in their existing environment and fed results through MRMraptor. 
  • MRMraptor interpreted their test results using domain-specific artificial intelligence and machine learning. 
  • MRMraptor autogenerated a zero-draft of model review documents in pre-approved templates that analysts could easily edit on top of using a WYSIWYG editor.

The Nordic bank found that MRMraptor was minimally intrusive, unlike many other MRM automation solutions. It’s model and environment agnostic, so the bank was able to continue to manage model testing in their existing environment and avoid the high costs associated with switching to a new one.

Business Impact

MRMraptor sped up model interpretation and documentation by more than 25% for the Nordic bank, shortening the overall model review process from a few months to a few weeks.

In addition, MRMraptor effectively upskilled more junior employees. Its domain-specific AI had information about counterparty risk models embedded into the software, so it could automate interpretation that normally requires long hours put in by highly skilled quants. Now, the clients’ experienced quants spend less time on documentation and invest time on more valuable projects.

MRMraptor provided the bank with more accurate and consistent documentation that helped ensure that the client’s model review documents met regulations.


Learn more about how our automated solutions can improve your model risk management.