The transformation of national statistical systems is accelerating as Artificial Intelligence (AI) technologies mature and become more deeply embedded in data workflows. For official statistics, this shift presents significant opportunities for automation, improved metadata management, enhanced data discoverability, and more dynamic integration of statistics with digital services.

This MODS-OECD-ADB Global Workshop will explore why AI performs remarkably well in some situations, yet struggles with official statistics and complex statistical questions? In responding to this question, the workshop will provide a concise introduction to modern AI and its relevance for statistical systems. It will establish a shared understanding of key concepts such as prompts, context, tools and metadata, and will highlight why statistical information poses particular challenges for AI systems.

The standard for Statistical Data and Metadata (SDMx) has long provided the foundation for structured, harmonised, and reusable statistical data. Participants will be introduced to an AI assistant connected to SDMx resources through MCP, which will be used throughout the session to explore statistical and SDMx-related questions.

Participants will observe how context, tools and metadata affect AI responses. They will compare answers with and without SDMx context, explore tool-supported responses via an AI assistant, and test how progressively richer metadata improves answer quality.

Having experienced how metadata quality shapes AI answers, participants will turn to the producer's side of that equation. The workshop will survey tools that help producers create and maintain SDMx data and metadata, show how a general interactive coding assistant connected through MCP already does real production work in any language, and will map the opportunities still waiting to be built. One principle runs through all of it: AI drafts, the expert decides, and the model never generates the number.

The workshop will conclude with a one-day hackathon where teams will apply the workshop's ideas to a demanding statistical information request, improving metadata, wiring AI interactions, and demonstrating a measurable improvement in an AI-assisted answer. Success will be judged by the quality, transparency and trustworthiness of the result, not by technical sophistication.

1. "AI State of Play": Provides a concise introduction to modern AI and explores why AI performs remarkably well in some situations, yet struggles with official statistics and complex statistical questions.

2. "What makes (Meta)data AI-Ready?": Explores the characteristics of SDMx structural and referential metadata artefacts that enable AI systems to discover, understand and correctly interpret statistical information and then, through hands-on exercises, observes the impact from enhancing these artefacts.

3. "AI Supporting Data Producers": Examines where in the production pipeline does AI earn its place, turning messy inputs into trustworthy SDMx, and where must it stay out of the way.

1. Understanding that while language models and AI technologies continue to improve, the quality, accessibility and structure of metadata often determines whether statistical information can be discovered, interpreted and compared reliably.

2. Knowledge of what makes metadata AI-ready.

3. Ability to introduce metadata improvements and test them through immediate AI feedback.

4. Knowledge of where AI already supports statistical production.

5. Ability to use at least one tool hands-on.

This event is designed for data professionals involved in statistical modernisation and technical implementation, including data engineers, IT specialists, methodologists, metadata experts, data managers and statisticians.

This is an intermediate to advanced technical event, participants are expected to have completed prior SDMx training before attending, with familiarity of the SDMx Information Model being essential. We strongly recommend completing relevant courses available through the SDMx Learning Resources (e.g., .Stat Academy), including introductory SDMx modules where appropriate. This preparation is vital to ensure participants can fully engage with and benefit from the capacity-building sessions and hackathon activities.