Large language models have the greatest potential to improve the quality of care in settings where the baseline quality of care is low, healthcare providers have more limited training, and the scope of practice is broad. At Penda Health, in Nairobi, Kenya, non-physician healthcare providers (Clinical Officers) provide urgent care and primary care services to over 1000 low-income Kenyans per day. LLMs have a real opportunity to provide “co-pilot” advice on diagnosis and treatment across this broad spectrum of care and lead to meaningful improvements in the quality of care provided. In this talk, I will present how we implemented a ChatGPT-4-based CDS co-pilot and how we evaluated the impact of the co-pilot on the quality of care.