A media intelligence base of 2.5 millionarticles, with humans still in the loop.
Three AI platforms for one of Africa's most important media watchdogs: Dexter, i3 Africa and WITI.
The situation
Monitoring Southern African media at scale means volume no human team can read, in languages and nuance no off-the-shelf model handles well.
MMA needed AI that amplifies its monitors' judgment instead of replacing it.
What we built
Dexter: a media intelligence platform scaled past 2.5 million articles — article ingestion, NLP analysis, tagging, monitoring and reporting in one workflow, extracting people, places, companies and quoted sources.
i3 Africa (Insights into Incitement): an AI classifier that triages potential incitement on a red-amber-green risk scale.
WITI (Who Influences The Influencer): a proof-of-concept turning TikTok and social video into transcribed, searchable, AI-analysed political-influence intelligence.
What changed
Early triage improved and the manual review burden dropped, while human monitors kept the final call on Southern African language and nuance.
A single 2.5m+ article intelligence base now underpins faster research workflows, more structured analysis and stronger reporting — work that used to require manual trawling.
“AI that amplifies the watchdog, not one that replaces it.”