This session will be held in English. Simultaneous interpretation in French and Spanish will be available.
This is the third session of the “AI Hub Clinic Series” organized by the AI Hub for Social Protection under the umbrella of the Digital Convergence Initiative. The series supports social protection institutions in building the foundations for responsible and effective AI use. By pairing country experiences, enabling cross‑context peer learning, and sharing practical tools, it helps participants assess their own systems and advance their AI journey in a risk‑aware way.
Overview
We are bringing this session back. Data is too foundational a topic for one round of discussion to be enough, so we are running a second session to keep the conversation going. Building on the themes from our first session, this session zooms in on the layer that determines whether any AI system in social protection actually works: the data underneath it. We will look at how institutions handle large, often fragmented administrative datasets and why getting these foundations right has to come before AI is introduced, not after.
Together, we will unpack what “data readiness” means in practice: the quality, coverage and completeness of administrative records and the practical challenges of connecting data across systems that were never designed to talk to each other. Existing biases and gaps in institutional data don’t disappear when AI enters the picture. They get carried forward, often invisibly.
Drawing on real institutional experience and open peer exchange, participants will explore how weak data can quietly undermine even well-designed AI tools and what practical steps can reduce that risk.
Topics included:
- Data quality, coverage and completeness
- Interoperability between systems
- Bias and gaps in administrative data
- How poor data leads to biased or unreliable AI results
Moderator: Saurav Bhattarai, Team Lead, openIMIS
Speakers:
- Felipe Leão, Executive Director of Innovation and Applied AI, Social Security Information and Technology Company (DATAPREV), Brazil
Clinic Series Timeline
- Are you ready to use AI in Social Protection? | April 14, 2026
- Data foundations for AI: Weak data, weak decisions (Part One) | July 14, 2026
- Data foundations for AI: Weak data, weak decisions (Part Two) | October 20, 2026
- Responsible AI: Ethics, rights and minimum viable governance
- Beyond Efficiency: What AI can (and cannot) do in Social Protection
- Use-case Scoping: From ideas to responsible action
- Applying the AI Hub Risk Framework
Learn more about the Series
