Bridging the Gap: A Scholar-Practitioner Framework for Integrating NIST Agentic GenAI RMF into Financial Risk Management

Gap between scholarly AI governance research and practical implementation in financial enterprise settings in light of Scholar-Practitioner Framework is discussed. Academic frameworks offer theoretical rigor but often lack operational specificity, while practitioner approaches prioritize immediate compliance but may miss foundational risk principles. This paper introduces a scholar-practitioner framework that systematically bridges this divide by integrating the NIST AI Risk Management Framework with quantitative risk workflows in agentic GenAI systems. Our approach makes three key contributions: First, we articulate four bridging mechanisms—dual-language communication, grounded abstraction, evidence-based pragmatism, and bidirectional knowledge flow—that translate between scholarly principles and operational imperatives. Second, we demonstrate how computational intractability research informs practical process-based governance strategies through the GOVERN, MAP, MEASURE, and MANAGE functions. Third, we validate the framework through financial services case studies showing how theoretical insights generate measurable business value (20-25% reduction in unexpected losses, 85-90% control effectiveness). Rather than treating AI governance as purely regulatory compliance or academic exercise, we position it as actionable knowledge creation where implementation experiences inform theoretical understanding and scholarly rigor enables practical innovation. While the NIST AI Risk Management Framework (AI RMF) and its Generative AI Profile provide comprehensive guidance, a significant implementation gap persists between theoretical principles and practical workflow integration. We demonstrate the framework's efficacy through financial risk management case studies and provide actionable implementation roadmaps. The framework harmonizes NIST guidelines with established enterprise risk management standards (ISO 31000, COSO ERM) while addressing verification gaps through tiered liability structures and transparent governance processes. Our contribution enables organizations to transform AI governance from compliance exercise to competitive advantage while ensuring safe, secure, and trustworthy agentic GenAI deployment. In this work we discuss both the theory and practice of AI governance by demonstrating that the scholar-practitioner model can transform abstract frameworks into operational systems.