Accessible AI Systems

Designing AI that works for Rwandans without a smartphone or a steady connection — from the start, not as an afterthought.

Elyssa Irankunda

Started January 15, 2026

This is an early-stage, exploratory research area. We'll share concrete results here once they exist.

Most AI products assume a recent smartphone and a stable connection. For a large share of Rwandans, neither is a safe assumption. Treating accessibility as a core design constraint — not a feature added later — is one of the areas ATAS is working through.

What we're working on

Voice-first interaction as a primary way in, not a secondary option

Interfaces reachable over USSD and SMS on basic phones

Designs that stay useful on slow or intermittent connections, and offline where possible

Abstract illustration of a shortcut bypassing a staircase to reach the same point. — The goal: reaching the same outcome without requiring every step most AI products assume.

Interested in collaborating?

Cite this work

ATAS Research Team (2026). Accessible AI Systems. ATAS Research Center. https://atas.rw/research/accessible-ai-systems

Project timeline

January 15, 2026
Research area established
Began looking at what it takes for our products to work well on basic phones and unreliable connections, not just as a fallback mode.