African Tech School

A publication on technology skills and the future of work

digital capability

French and Bambara feature differently in Kabakoo’s AI learning conversations

By · · EN

An analysis of 11,171 bilingual conversations found digital and technical questions more often in French, while entrepreneurship and endogenous knowledge topics appeared more often in Bambara.

A person in a hard hat holds a drill beside a pavilion clad in woven fifalan panels.
Kabakoo Academies

Different topics, different language patterns

At Kabakoo, an analysis of 11,171 bilingual conversations from June 2026 found a distinct relationship between language and topic: digital and technical questions came about four times more often in French, while entrepreneurship and endogenous knowledge topics ran about twice as often in Bambara.

These are patterns within the conversations analysed: they describe how language and subject matter relate, rather than establishing that either language caused learners to choose particular topics.

Feedback on actual learner work

Kabakoo has been building AI for its learners since 2019 and has operated an AI coach on WhatsApp since 2023. Its AI Mentor reviews actual learner work, such as a project video or a business plan, and returns personalized feedback in French and Bamanankan text, 24 hours a day.

Over six months in 2025, 6,618 accounts messaged the AI Mentor on WhatsApp, while 284 did so in the app. Those figures count accounts that messaged the mentor, not completed training pathways.

Testing speech recognition for Bamanankan

Kabakoo also fine-tuned Whisper medium for Bamanankan, using curated audio and existing speech datasets. In August 2026, the model reached 24% word error on held-out data and 33% on an external benchmark.

The 2025 model reached 64% word error on a different test set. Because the test sets differ, these figures are not a like-for-like measure of improvement. The August results report performance separately on held-out data and an external benchmark.

Putting technical skills into projects

In Kabakoo’s cohort program, mindset work is mandatory for all learners, while technical skills such as web design are optional. Cohorts run for 16 weeks, with mindset modules in weeks 1–6, followed by technical skills applied through KLIPs—Kabakoo Local Innovation Projects—and a final public showing of learner work. The program totals over 200 hours per learner.

Pods bring learners into groups of three to five, matched on shared interests and geographic proximity, to work on concrete projects. Kabakoo’s explanation of its learning system details this combination of project work, peer groups and AI feedback.

The Kabakoo newsletter

Research, field notes, and stories from Kabakoo in your inbox.