Towards Agile and Transparent Data Management: New Generative AI-Driven Tools for Data Governance

Abstract

Companies and public organizations around the world are facing unprecedented challenges due to the ongoing digital transformation. These changes have raised numerous concerns related to data privacy, the ethical use of personal information, and organizational transparency. In Quebec, a region of Canada, new legislation now requires organizations to minimize the amount of data they collect, ensure strict access controls, and provide individuals with the ability to access, correct, or request the removal of their personal information held by an organization. Hydro-Québec, Quebec's largest public energy provider, aims to be a pioneer in ethical and efficient data management. In this study, we explore new generative artificial intelligence (AI) tools to enhance data transparency for Hydro-Quebec customers. We present a customized generative AI approach, leveraging GPT-4 via Azure OpenAI Studio, to automatically generate compliant French descriptions for public data requests. The tool is implemented as a new Python library, capable of producing high-quality descriptions efficiently. It achieved a 90.48% human evaluation score, outperforming Collibra AI (73.81%) and Databricks AI (71.03%). Validation through cosine similarity, METEOR scores, and LLM-as-a-Judge assessments confirmed the solution's superiority, highlighting its potential to drive transparent digital transformation and promote ethical data governance.