A smarter model risk management discipline will follow from building smarter models: An abbreviated guide for designing the next generation of smart models

Abstract

What if a financial firm decided to delete its entire set of models and redevelop them from scratch? What might it do differently in the process of rebuilding its entire model ecosystem in order to avoid and leverage from some of its previous mistakes? How could such a firm make a model risk management (MRM) platform smarter and less resource intensive than it was before? This paper describes one forward-looking possibility for making the manually intensive practice of MRM smarter by building models that are smarter in the sense of having a rudimentary level of ‘self-awareness’. Similar to the ways that tech firms have tracked the usage of their smart phones, cars, laptop computers and printers for many years, active intelligent agents embedded in model source code can support the creation of a dynamic model inventory to serve as a repository of historical data that accurately describes how, when and where a firm’s models are used and to diagram firm-wide interdependencies between data and models.