Human Centred AI | Designing AI Health Tools for Safety, Failure and Accountability

Designing Safer AI for African Healthcare | Human-Centred AI What happens when an AI health tool gives the wrong answer, misunderstands a patient or fails in the middle of a clinical workflow? As AI becomes more integrated into healthcare, building accurate models is only part of the challenge. We also need to design for uncertainty, failure, patient safety and accountability. In this webinar, AI HealthBeat Africa, in collaboration with Spindle Design, brings together voices from clinical practice, AI development and product design to explore how we can build AI health tools that work safely in the real world. Speakers 1. Dr Sarah Kiptinness — Shares lessons from Penda Health’s experience developing and evaluating AI Consult for clinical decision support. 2. Stephen Obonyo — Explores how Jacaranda Health builds safety checks, human escalation and quality monitoring into AI-powered maternal health information services. 3. Wanjiru Kigathi — Discusses accessibility, inclusive design, privacy and the realities of building voice-based AI health tools for blind and visually impaired users. 4. Kayuyu Mwaura — Shares a product design perspective on integrating AI into existing healthcare workflows and anticipating failure before it happens. In this conversation, we discuss: Why AI tools can fail even when the underlying technology appears capable. How to design safeguards, human oversight and escalation pathways. Why local context, language, accessibility and privacy matter. How to monitor AI performance and learn from errors after deployment. Who should be accountable when an AI health tool fails. #AIHealth #DigitalHealth #ResponsibleAI #HealthInnovation #AfricaTech #HumanCentredAI