Jacaranda Health, a nonprofit organization, has implemented an artificial intelligence program called PROMPTS to provide pregnancy and postpartum health information via text message in Kenya. The system uses a large language model to interpret questions from users, prioritize responses, and identify potential medical emergencies that require immediate human intervention. Originally starting with fewer than 100 questions per day, the service now handles approximately 15,000 daily inquiries from an estimated 700,000 annual users.
The program was developed following the launch of an automated appointment reminder service, which prompted patients to text back with health concerns. Because the system operates entirely over SMS, it is accessible to individuals using basic mobile phones without internet or smartphone capabilities. The service currently operates in 24 Kenyan counties and has expanded to include speakers of Swahili, English, Sheng, Twi, and Hausa in parts of Ghana, Nigeria, and Tanzania.
According to Jacaranda Health, the AI identifies approximately 7% of incoming messages as potentially urgent. These messages bypass automated responses and are routed to a help desk staffed by 18 nurses who monitor the system 24 hours a day. Lisa Mushega, a policy expert for the health provider coalition HENNET, stated that the tool helps identify conditions such as pre-eclampsia and neonatal jaundice, which might otherwise go untreated due to delays in seeking care.
For the individual user, the service is free, with the $2 per-person maintenance cost covered by the nonprofit for the duration of a pregnancy and the first year after birth. Patients notice a shift from traditional one-way public health announcements to a two-way dialogue where they can ask specific questions about nutrition or symptoms in a private space on their own phones. Additionally, the system allows the nonprofit to track broader health trends, such as correlations between extreme heat and reports of dehydration among pregnant women, which can inform local clinical priorities.
However, the implementation of AI in healthcare presents specific risks and costs. Researchers like Javaid Iqbal Sofi of Virginia Tech University have raised questions regarding what happens if the system misclassifies an urgent symptom as routine. To address these concerns, Jacaranda Health employs a team of nurses to audit automated answers and manage emergency queues. While this human oversight is intended to ensure accuracy and privacy, it also increases the operational cost compared to fully automated systems. Moves to automate healthcare elsewhere, such as in U.S. insurance claims or staff scheduling, have previously faced criticism from labor unions and regulators.