AI-in-the-Loop Infrastructure for Scaling
Across the world, millions of people lack access to high-quality services — whether in healthcare, agriculture, or justice systems. The challenge is not only access, but capacity: workforce shortages, uneven quality, fragmented data, and delivery models that struggle to adapt to diverse and resource-constrained contexts.
Even where services exist, maintaining consistent quality at scale requires ongoing training, supervision, monitoring, and adaptation — functions that are difficult to sustain in under-resourced systems, regardless of sector.
This session explores how AI-in-the-loop infrastructure can help address these challenges.
Rather than replacing human decision-making, this approach focuses on how AI can strengthen systems by augmenting human capability and improving adaptability. This includes supporting training and supervision, enhancing quality assurance, enabling better use of data, and helping tailor services to different populations and contexts.
At the same time, the session will engage directly with the risks. Poorly designed AI systems can compromise safety, amplify bias, and widen inequities, particularly when tools developed in high-income settings are applied without adaptation.
Through live demos, speakers will share real-world experience building and deploying AI-enabled systems in low-resource environments, exploring both the potential and the practical dilemmas.
Key questions will include:
What does “AI-in-the-loop” mean in practice?
How do we measure whether AI improves outcomes, not just efficiency?
When does AI support human judgment, and when might it undermine it?
What level of evidence is needed before deploying AI?
How do we ensure tools are contextually appropriate and equitable?
Who this is for
This session is for leaders, technologists, funders, and practitioners interested in scaling, particularly in low-resource settings.
What you will get out of it
A clearer understanding of how AI can support, not replace, human-led systems
Real-world insights into deploying AI in complex, resource-constrained environments
A deeper view of the risks, trade-offs, and ethical considerations involved
Practical questions and frameworks to guide responsible AI adoption
Speakers
Tom Osborn and Shadrack Lilan, Shamiri. Will speak to work building AI-enabled training, supervision and quality control infrastructure while navigating real-world deployment dilemmas (data quality, code-switching, fidelity, trust).
Utkarsh Saxena, Adalat AI. Will speak to building of tech solutions to courts and judicial systems in India to tackle widespread backlogs and delays.
This session offers a grounded exploration of how AI can acclerate scaling ambitions in ways that strengthen program quality, equity, and human agency at scale.