By 2050, one in three children in the world will live in Africa. Yet this demographic shift coincides with a profound learning crisis: more than 70 percent of children in low- and middle-income countries (LMICs) cannot read or understand simple text by the age of 10 – and in sub-Saharan Africa, this figure rose to 86 percent before the pandemic. Without rapid acceleration of core learning outcomes, this demographic advantage risks becoming a source of increased inequality and lost opportunities, talent and productivity in the labor market.
Artificial intelligence (AI) is transforming education, but most AI-based EdTech products are more suited to high-income countries – where infrastructure, data availability and learning conditions are very different – while the needs are greatest in LMICs. Without deliberate design and policy choices, AI risks widening learning gaps around the world. This blog explores what is needed for AI-powered EdTech to close the global learning gap and equip Africa’s next generation with the skills they will need.
AI offers opportunity – if we build with intention
Hundreds of AI-based EdTech products are being implemented in LMICs, and preliminary evidence suggests they can improve efficiency and support learning when designed well.
In Rajasthan, India, national authorities used AI-powered assessment tools to grade the paper worksheets of 4.5 million learners, while in Kenya, nearly 400,000 children are using EIDU, a structured learning solution with demonstrated learning gains. A World Bank after-school program in Edo State, Nigeria, achieved significant learning gains after just six weeks of AI tutoring and teacher coaching.
At the same time, major tech companies are adding educational features, such as Gemini Guided Learning, Claude’s Learn Mode, and OpenAI’s Study Mode. But context matters – if an AI tool in rural Tanzania generates a lesson plan with a pizza rather than a chapati, it has already failed to adapt to learners.
How can AI live up to its potential to support learning equitably and at scale?
Fab AI, the Gates Foundation and the World Bank share a common goal: shaping the world’s best technologies to help those who learn the least. Achieving this requires focusing on three priorities:
1. Build fairly – AI that works everywhere
To ensure that what works in high-income countries also works for LMICs, AI must be built with an understanding of local realities: languages, cultural context, school curricula, and pedagogical approaches for basic reading and mathematics. It should also reflect practical constraints such as infrastructure and bandwidth – emphasizing the importance of low-bandwidth solutions, offline functionality and smaller language models that can operate in resource-constrained environments.
2. Work collaboratively – local developers, educators, governments and big tech companies
AI-based EdTech developers across the world face many of the same challenges. The opportunity to share lessons, build openly, and avoid duplicating efforts is significant – particularly around assessment, security, and content quality.
Only 0.2% of the data used to train AI models comes from Africa and South America. Collaboration between local developers, educators, governments and technology companies is essential to ensure that AI systems are contextually relevant, aligned with national curricula and effective for learners in LMICs.
Joint initiatives involving educators from the start can create safe environments for testing new tools and enable responsible sharing of resources such as datasets and knowledge graphs. Realizing this potential requires new models of collaboration and governance that deliver clear value for all partners.
Encouragingly, large-scale research and skills development partnerships in LMICs are already emerging, such as Anthropic’s partnership with the Rwandan government, Microsoft’s initiative in Kenya, and OpenAI’s accelerator in India.
Programs focused specifically on improving foundational learning at scale in LMICs will be essential as these efforts expand.
3. Building evidence and ensuring quality – safe, effective and scalable AI
The World Bank, Gates Foundation and Fab AI share the common goal of supporting countries in the responsible use of AI in education by building evidence, establishing benchmarks and scaling what works in education systems.
This requires quality checks throughout the AI product lifecycle – from initial concepts and development to large-scale deployment. It also means testing AI-based EdTech in real-world settings and building evidence on learning outcomes and effectiveness at a system level. Only then can products achieve their intended goals.
An emerging and critical part of this quality assurance is the testing of AI results. Fab AI, with support from the Gates Foundation and the UK Foreign, Commonwealth and Development Office, is developing benchmarks of AI, while conducting effectiveness studies – creating a practical framework to help governments, funders and developers distinguish promising AI-based EdTech from the rest. Although few products currently report evidence, there are efforts to compile evidence on the impact of AI-based EdTech products (see EdTech for Good, EdTech Tulna and EduEvidence). An agentic tool for compiling and evaluating evidence on AI-based EdTech products will soon be available on the Fab AI website.
At the same time, many pilot projects supported by the World Bank are completed or underway in LMICs, including adaptive learning programs in Ivory Coast, Gambia and Mali; WhatsApp-based tutors in Ghana, teacher-focused solutions in Ethiopia; and upcoming youth skills programs in Tanzania and Mauritius. Together, these efforts help build the evidence needed to guide responsible adoption and scale.
Harnessing the potential of AI to help children learn – and thrive
In November 2025, more than 100 leaders from the education and technology ecosystem, including developers, governments, funders and large technology companies, gathered at the AI for Education Summit in Nairobi. The aim was to focus on what it takes for AI to improve learning outcomes in sub-Saharan Africa and beyond.
With “anchored ambition” in mind, as called for by Dr. Ben Piper, Director of Global Education at the Gates Foundation, participants explored high-impact AI use cases for teacher support, personalized learning, and assessment. Organizations in different contexts face the same challenges. For AI to make a real difference in education, solutions must be systemic, anchored in local realities and aligned between stakeholders.
Luis Benveniste, Global Director for Education at the World Bank, calls for “supporting students from foundational learning to job-relevant skills.” We must leverage responsible AI to accelerate this journey and scale, ensuring young people can thrive in a rapidly changing world. »
Meeting this challenge requires concerted action. We invite developers, educators, governments, multilateral organizations and technology companies to join us in shaping the next generation of AI-driven EdTech – tools that are equitably built, collaboratively developed and evidence-based.
Only by working together can we ensure that the AI that reaches classrooms is safe, effective and designed for the realities of LMICs, helping all learners acquire the fundamental skills they need to progress, access opportunities and thrive.
Thanks: We express our gratitude to Halil Dundar, Practice Manager of the Global Education Engagement and Knowledge Unit, for his comments and contributions to this blog. We also appreciate the contributions of Romana Kropilova (Director of EdTech, Fab AI) and Guy Benton (Communications Manager, Fab AI).
