About this web edition
Building AI Applications for Livelihoods: Lessons and tools from Kabakoo’s WhatsApp learning system is written by Kabakoo Academies for organizations building or adapting AI-supported learning tools.
Web edition 7, September 14, 2026. Ten chapters cover the work from problem framing through continued participation, peer learning, and experimentation. Five further chapters will examine outcomes, governance, teams, costs, and adaptation.
A collective effort
This playbook is a collective effort.
First and foremost, we thank the learners who trust us with their hopes and ambitions. Their questions, work, feedback, and experiences shape what we build and what we learn. We also thank those who have chosen to let us share their stories and photographs.
We thank the complete Kabakoo team, past and present, whose daily work makes this learning possible: supporting learners, designing courses, building and maintaining the technology, conducting research, and turning what we learn into changes in practice.
We are grateful to our research partners at Paderborn University—Dr. Martin Schneider, Enja Herdejürgen, and Dr. Sumaya Islam—for helping us design our evaluations and follow learners’ progress over time.
We thank Dr. Abeba Birhane and Dr. André Uhl for the sustained dialogue on AI accountability and AI governance. We also thank Dr. Hellina Hailu Nigatu and Dr. Ellen Rushe for their collaboration on Bamanankan language technologies.
Dr. Richard Sedlmayr and Catherine Thomas, PhD, have helped us sharpen our work on mindset, agency, and measurement through conversations since 2023, including the development of the Kabakoo Behavioral Index.
We thank Temina Madon, PhD, for her guidance as chair of Kabakoo’s board. We also thank Paul Wang for encouraging us to turn our experience into tools and a manual that other organizations could use.
We thank Dr. James Walsh for his behavioral analysis of onboarding and independent examination of our outcome evidence. We also thank Elia Gandolfi for his work with our team on experimentation and learner engagement, including understanding where participation breaks down and how human contact relates to continued learning.
We thank Edmund Korley for his technical mentoring since 2023 and his work alongside our engineers on the Kabakoo AI Mentor and its supporting data systems. We also thank Aman Dalmia for supporting our team’s use of Calibrate to make AI evaluation part of everyday work for learning and content colleagues as well as engineers.
We are grateful to the wider Agency Fund community, especially the teams who worked alongside us at the Bangalore and Naivasha 2024 sprints and the July 2026 Nairobi sprint, for sharing their questions, experiments, and practical experience.
Credits and reuse
The text and working tools are shared under CC BY-SA 4.0. Credit Kabakoo Academies, identify changes, and share adaptations under the same license. Supporting code uses Apache-2.0. Third-party sources retain their own terms. The text license does not grant unrestricted reuse of photographs or learner likenesses; contact Kabakoo about image reuse.
The four-level evaluation framework is credited to The Agency Fund. The playbook draws on Elia Gandolfi’s engagement analysis, Jamie Walsh’s onboarding diagnosis, Yanick Kemayou’s AI evaluation essay, and Kabakoo’s operational and research records. See Sources for the evidence and its limitations.
Citation
Kabakoo Academies. (2026). Building AI Applications for Livelihoods: Lessons and tools from Kabakoo’s WhatsApp learning system. Web edition 7. https://playbook.kabakoo.africa/