An AI-powered Structured Pedagogy Program’s Impact on Student Learning and Teacher Productivity

Millions of children worldwide struggle to read or do basic mathematics by age 10. While many interventions have been tested–including Teaching at the Right Level (TaRL) pioneered by Pratham, and various forms of ed-tech–scaling effective, personalized learning solutions in low-income settings remains a challenge. AI offers the potential to make these tools more adaptable, accessible, and impactful. EIDU, a Kenya-based ed-tech organization, collaborates with education systems to deliver structured pedagogy for primary and pre-primary students. Preliminary evidence shows its approach improves literacy and numeracy by 0.5 standard deviations. Through Google.org’s GenAI Accelerator, EIDU has developed a generative AI model to support targeted remediation. The tool identifies learning gaps, generates student profiles, and produces ~15-minute personalized tutoring plans for small groups. This project will evaluate two approaches: (a) the current “low-touch” model requiring minimal classroom reorganization; and (b) a more ambitious integration with TaRL, where classrooms are restructured around learning levels. The pilot will assess how students, teachers, and potentially parents interact with the tool and the feasibility of this integration. A future randomized evaluation would randomize between: (1) the current EIDU GenAI tool; (2) a GenAI+TaRL model; and (3) a control group. The researchers would measure impacts on student outcomes (literacy, numeracy, cognitive, socio-emotional skills) and teacher outcomes (time use, satisfaction). More broadly, the researchers aim to explore whether AI can shift classrooms from a one-size-fits-all model to personalized structures—amplifying teacher capacity, rather than replacing it—with scalable, pro-worker AI tools.