Abstract

The Domain-AI Governance framework, designed to fill the considerable gap between business objectives for ai and explicit (and validated) data quality requirements, was created in response to the fact that 75% of all AI initiative will fail. A Unified Framework featuring four levels has been proposed for Domain-AI governance; the Data Foundation Level, Quality Gate Levels, AI Orchestration Levels, and Governance Dashboards which incorporate appropriate controls to enhance data quality and allow for subsequent traceability when implementing any ML solution, will provide a more structured way to implement ML solutions and reduce costs associated with re-training ML in the range of 50%, enabling companies to obtain regulatory compliance quicker. Through successful implementation of the Domain-AI Governance framework, many organizations have reported higher audit pass rates and an extremely high ROI due to reductions in the length of time required to re-train ML for regulatory purposes, as evidenced by the successful implementation of the framework by several organizations in both financial services and healthcare sectors. Furthermore, The Domain-AI governance framework focuses heavily on creating cross-functional centers of excellence for sharing knowledge and best practices, developing an advanced trust scoring system, and providing a broad range of technological solutions for organizations to provide an environment compatible with existing AI risk management frameworks. Future enhancements to this framework will emphasize both federated learning and flexibility in order to meet the ever-evolving global standards being established.