This is a side event alongside the Skoll World Forum in Oxford. Registration at the larger forum is not required.
From math tutors to farmer advisory tools, generative AI is rapidly expanding in low- and middle-income countries. How do we know if these tools are working?
Evaluations can help, but there is little agreement on what they should include. Tech teams prioritize product performance, often overlooking impact, while impact evaluators focus on outcomes but may neglect the underlying technology.
This session introduces the AI Evaluation Playbook. The playbook supports organizations in evaluating AI systems from model performance through to product use, user behavior, and impact on outcomes.
Through practitioner and funder perspectives, we'll explore how organizations can build evaluation into their workflows, ask better questions about impact, and make smarter decisions about where and how to use AI. The event will feature a presentation on Digital Green’s approach to evaluating its AI farmer coach, and a panel discussion on how donors and governments are approaching evaluation.
This session is for funders and builders of AI products. It will be followed by open networking and light refreshments.
