Development aid listens poorly. AI can change that.

I spent 20 years in development aid. I’ve seen thousands of surveys, focus groups, field reports. And one observation keeps coming back: we listen poorly.

We listen through questionnaires designed in Geneva. We listen in French to people who think in Wolof. We listen with checkboxes when we need open stories. We listen once, at the start of the project, and never come back.

AI won’t solve this structural problem. But it can open doors.

Imagine: a farmer in Thiafoura speaks in Wolof about his practices, his difficulties, what he observes in his field. AI transcribes, translates, structures. Without a rigid questionnaire. Without an intermediary who filters. Raw speech, organized, analyzable.

This isn’t science fiction. We’re already doing it in Senegal with the Sparks project.

The real challenge isn’t technical. It’s political. Who controls the data? Who decides the questions? Who benefits from the analysis?

AI amplifies. It amplifies the good and the bad. If we use it to extract better, we reproduce the same mistakes, faster. If we use it to listen better, we can finally build aid that starts from the people.