A Voice-First AI Service for People-Centred Justice in Niger

How can voice-first AI be designed and embedded within Niger’s customary justice system? Taking Niger’s actual justice pathways as our point of departure, we report on a pilot that delivers nationally validated guidance over land and shared natural resources. The system is based on justice needs and satisfaction data and was co-created with justice seekers, national authorities, and customary and religious leaders. We present “Hadin kai–bani fondo” (“the Path for Cohesion”), a free phone service where people can call, ask a question, and hear a step-by-step answer. The system uses a Retrieval Augmented Generation (RAG) pipeline that draws exclusively on a curated set of Niger-specific texts. We describe the data, system architecture, and institutional setting, and we show (i) that retrieval-bounded AI can be embedded within existing authority structures, (ii) that corpus governance together with institutional review helps contain hallucinations and norm conflicts, and (iii) that a voice-first channel makes AI-mediated justice services workable in low-connectivity, multilingual environments.