Leveraging generative AI for official statistics: opportunities, challenges and use cases
Abstract
In this work, we delve into the opportunities and challenges of generative AI for official statistics, with particular attention to the process and dissemination phases of the Generic Statistical Business Process Model. We also take stock of the international statistical initiatives in this domain to which the BIS contributes. Drawing upon use cases from the BIS context, we show how AI-powered assistants can be leveraged for metadata editing to enhance accuracy and consistency. Additionally, we illustrate how AI can be useful for metadata dissemination, particularly to simplify the accessibility and discoverability of statistical content, including methodologies, glossaries, and taxonomies. We show our findings as regard to process efficiency, metadata quality as well the challenges associated to performance, costs and operational risks. Finally, we illustrate the benefits of leveraging SDMX to build end-to-end AI-powered statistical workflows.