Making AI Count: The Next Measurement Frontier
The webinar “Making AI Count: The Next Measurement Frontier” examines how artificial intelligence (AI) systems can be effectively measured and evaluated in real-world applications, with emphasis on improving accountability and interpretability of AI-driven outcomes.
The session is delivered by Dr. John Lourenze S. Poquiz, Research Associate at the Bennett School of Public Policy, University of Cambridge. He discusses emerging approaches and challenges in AI measurement, particularly the limitations of existing evaluation metrics in capturing real-world impact.
The talk highlights the need for improved frameworks that align technical performance metrics with policy and societal considerations, enabling more meaningful assessment of AI systems across domains.
Speaker:
Dr. John Lourenze S. Poquiz
Research Associate, Bennett School of Public Policy
University of Cambridge
Date: 17 November 2025
Abstract: Generative AI is reshaping how we produce, consume, and work yet our statistical system, as it is now, would struggle to capture its true economic impact. As the 2025 System of National Accounts begins recognizing AI as a distinct asset, new challenges emerge in valuing AI-related investments, outputs, and productivity gains. This paper explores the measurement gaps created by transformative AI technologies and discusses how more granular, task-based, and outcome-focused approaches can make economic statistics fit for an AI-driven economy.
Speaker: John Lourenze S. Poquiz is a Postdoctoral Research Associate at the Bennett School of Public Policy, University of Cambridge, with over seven years of experience in economic measurement, including work on national accounts at the Philippine Statistics Authority. His research focuses on macroeconomic accounting, productivity, and the digital economy, with particular expertise in platform economics, causal inference, and impact analysis. He works extensively with large-scale administrative data and unstructured sources, including web and user-generated content.
He has collaborated with international organisations such as the World Intellectual Property Organization (WIPO) and the Asian Development Bank (ADB), and has presented his research at leading conferences including the NBER Summer Institute (CRIW) and the World KLEMS Conference. He is an active member of the Economic Statistics Centre of Excellence (ESCoE) and The Productivity Institute (TPI), and is increasingly engaged in exploring the application of large language models (LLMs) in economic research and measurement.
Speaker: John Lourenze S. Poquiz is a Postdoctoral Research Associate at the Bennett School of Public Policy, University of Cambridge, with over seven years of experience in economic measurement, including work on national accounts at the Philippine Statistics Authority. His research focuses on macroeconomic accounting, productivity, and the digital economy, with particular expertise in platform economics, causal inference, and impact analysis. He works extensively with large-scale administrative data and unstructured sources, including web and user-generated content.
He has collaborated with international organisations such as the World Intellectual Property Organization (WIPO) and the Asian Development Bank (ADB), and has presented his research at leading conferences including the NBER Summer Institute (CRIW) and the World KLEMS Conference. He is an active member of the Economic Statistics Centre of Excellence (ESCoE) and The Productivity Institute (TPI), and is increasingly engaged in exploring the application of large language models (LLMs) in economic research and measurement.