Background & Context

Agentic AI is generating interest amongst implementers for its potential to enable more personalized, responsive programming. Still, many are struggling to know how to implement it within scaling or scaled tools practically. This session draws on Jacaranda Health’s experience evolving its PROMPTS platform, which has relied on AI since 2019 to cost-effectively deliver two-way SMS-based pregnancy and postpartum support to approximately 700,000 mothers annually. It serves as a case study in integrating agentic AI, which raises a new set of implementation questions distinct from earlier AI approaches. 

Objectives

1. Present a hands-on case study of how Jacaranda Health is integrating agentic AI into a live digital health platform 

2. Discuss the broader set of questions that implementers must consider when deploying agentic systems, such as: 

   1. Hallucination risk and clinical safety, particularly how agentic systems detect, bound, and recover from unsupported or unsafe outputs in health workflows.

   2. Data sovereignty and cross-border data flows, including where sensitive data is stored, processed, and governed across jurisdictions.

   3. Security, privacy, and regulatory compliance, ensuring alignment with national health policies, consent frameworks, and evolving AI governance requirements.

Expected Outcomes

Participants will leave with a clearer understanding of the key considerations involved in integrating agentic AI into existing digital platforms, including design choices, data implications, and operational trade-offs.

Relevance to Conference Themes / Contribution

This session grounds discussion of agentic AI in a live digital health platform and focuses on real implementation decisions rather than theory, offering a practical reference point for ICT4D teams navigating similar choices