AI for Evaluation: What It Changes, What It Breaks, and What It Enables

​Event Description

AI is beginning to transform how the social sector measures, evaluates, and learns. From analyzing large datasets to generating insights faster and at lower cost, the potential is significant. But so are the risks.

This session explores the real-world use of AI in evaluation and impact measurement, moving beyond theory to practical lessons from organizations already applying these tools at scale. Where is AI genuinely improving evaluation, making it faster, more adaptive, or more insightful? And where is it introducing new risks around bias, data quality, and over-simplification?

We will also examine how AI could reshape the broader impact data infrastructure of the sector. Could it reduce the burden of fragmented reporting, streamline data collection, and enable more meaningful learning? Or does it risk reinforcing existing challenges in new ways?

Bringing together evaluators, technologists, and funders, this session will focus on how to use AI to strengthen learning and accountability, while staying grounded in what matters most: useful, trustworthy insights that inform better decisions and real-world impact.

Speakers

• ​Paula Fabiani is CEO of the Institute for Social Development (IDIS), advancing philanthropy and giving through research, advocacy, and sector leadership in Brazil.

• ​Sam McPherson is Partner at Itad, advancing global development through evaluation and advisory services.

• ​Zak Kaufman is Co-founder and CEO of Vera Solutions, advancing technology-driven impact measurement and operations for the social sector.

• ​Zecheng Wang is Analyst at Co-Impact, supporting systems change through global philanthropy and partnerships.

• ​James Walsh is Senior Behavioral Scientist at The Agency Fund, advancing social service delivery through behavioral economics.

​Discussion Questions
Where is AI adding real value in evaluation and impact measurement today?
What risks does AI introduce in how we collect, analyze, and interpret data?
Can AI help simplify and centralize impact reporting across the sector?
How do we ensure AI strengthens learning for systems change, rather than just efficiency?