Voxy
A multilingual AI business communication platform that converts WhatsApp voice notes and informal text in English, Nigerian Pidgin, Yoruba, Hausa, and Igbo into professional ready-to-send business replies — in seconds. Built for African small businesses. Currently live in beta.
The Problem
67% of Nigerian SMB inquiries arrive as WhatsApp voice notes in local languages. Small businesses were losing sales because of a 4-hour average response delay. On the technical side, building a multilingual NLP pipeline for low-resource languages like Yoruba and Igbo — severely underrepresented in AI training data — was genuinely hard. On the product side, the tool had to work for non-technical users with zero learning curve. A rigid subscription model would have killed adoption, so the entire billing approach needed rethinking.
The Solution
// what I learned
Voxy taught me more about product thinking than almost any other project I've built. Building for African markets means rejecting most assumptions baked into Western product design — around payment behavior, language, and trust. Technically, I developed deep respect for the complexity of multilingual NLP and the gap that still exists in AI tooling for African languages. This is a problem worth continuing to solve.
