Where AI fits and where It doesn't: A framework for problem-led AI integration in African education

Major technology companies and development partners are committing unprecedented resources to AI in education across the majority world. Anthropic, OpenAI and Google have announced significant partnerships with FCDO, UNDP, the World Bank and bilateral development agencies, positioning AI as central to closing learning gaps and expanding access to quality education. 

The question is increasingly less about whether AI will be adopted in African education systems but whether that adoption will produce genuine impact or repeat the well-funded disappointments of the past.

The history of educational technology is littered with interventions that failed because they solved the wrong problem. The "laptop per child" initiatives of the early 2000s are instructive: billions invested, negligible impact, and mounting evidence that the binding constraint was never device access but teacher capacity, curriculum quality, and the time children could spend in school. AI risks repeating this pattern at greater scale and speed.

Our Frontier Technologies and Human Development teams, working in partnership with INJINI, an African EdTech accelerator, have developed a framework to address this challenge. Rather than organising AI by its technical capabilities, a distinction that matters to engineers but rarely to practitioners, the framework is problem-led. It maps six education challenges critical to African contexts: education finance, physical infrastructure, teacher capacity, learning materials, quality and inclusive teaching, assessment, and parental engagement against five functional AI capabilities.

The framework reveals both opportunity and caution. High-impact areas remain under explored: AI for physical infrastructure planning, virtual laboratory access, and multilingual parental engagement tools are largely neglected despite clear need. Meanwhile, student tracking and personalised learning attract the largest share of investment, creating crowded spaces where marginal returns are diminishing. Most critically, the introduction of AI into classrooms requires grappling with fundamental questions about pedagogy, equity, teacher roles, and infrastructure realities that technology alone cannot answer.

Drawing on case studies from leading African EdTech implementers Mindspark and Siyavula, the framework identifies four critical considerations for responsible AI adoption: ensuring genuine learning rather than technological facades, augmenting teacher capacity rather than substituting for it, aligning automated assessment with pedagogical intent, and navigating infrastructure constraints to ensure equity rather than entrench advantage.

As AI funding flows accelerate, the risk is that countries will adopt the wrong technology, applied to the wrong constraints, crowding out interventions that would actually work.

Download the full report here or book time with us if you’d like to discuss our work.