In a modest classroom at Edo Boys High School in Benin City, Nigeria, students recently participated in an after-school program that may redefine the future of learning. “AI helps us to learn. It can serve as a tutor, it can be anything you want it to be, depending on the prompt you write,” said Omorogbe Uyiosa, known as “Uyi” to his classmates. His school was one of several that took part in a pilot program exploring the use of generative artificial intelligence (AI) in education.
The program, supported by the World Bank and run between June and July 2024, aimed to assess the impact of generative AI as a tool to support basic education, particularly in English language learning. Now, the results are in—and they are nothing short of transformative.
According to findings from a forthcoming randomized impact evaluation, students who were randomly assigned to participate in the six-week AI-supported intervention significantly outperformed their peers in all tested areas, including English language proficiency, AI literacy, and digital skills. The intervention’s primary focus was English, but students also showed marked improvement in their year-end curricular exams—assessments that covered content far beyond what the program taught. The evidence points to the AI tool’s role not just as a digital assistant, but as a catalyst for independent, self-guided learning beyond the classroom.
Importantly, the program showed strong equity outcomes. Students of all performance levels, including those initially behind, saw improvements. Girls in particular benefited significantly—reducing the gender performance gap and demonstrating the intervention’s potential to close long-standing learning divides.
Attendance, though affected by challenges such as seasonal flooding, teacher strikes, and domestic responsibilities, proved crucial. Each additional day of participation correlated with significant gains in learning outcomes. The data suggest that longer or more sustained programs could result in even greater impacts, as learning benefits did not level off with continued exposure.
Quantitatively, the results were striking. Students recorded a learning gain of approximately 0.3 standard deviations—equivalent to nearly two years of typical learning achieved in just six weeks. When compared against a global database of educational interventions, this program outperformed 80 percent of evaluated strategies, including widely respected models such as structured pedagogy and teaching at the right level. The scale of these gains, especially within such a short intervention period, underscores the high potential of AI-assisted learning in resource-constrained settings.
While this pilot is among the first to rigorously evaluate the impact of generative AI as a virtual tutor in a developing country, the implications stretch far beyond Nigeria. It builds on existing examples—such as coding education supported by AI, math instruction via WhatsApp in Ghana, and experimental AI learning tools in Turkey—but provides concrete, scalable evidence that AI, when paired with thoughtful implementation and teacher guidance, can enhance learning in significant ways.
Yet questions remain. What are the long-term effects of such interventions? How do students’ relationships with AI tools evolve over time? What role must teachers continue to play in guiding these interactions? And perhaps most critically, how can such programs be scaled without undermining traditional pedagogy or exacerbating inequalities?
As the World Bank prepares to release a final analysis in its ongoing series, attention is turning toward how best to integrate generative AI into national education systems—not as a shortcut, but as a supplement that empowers students and teachers alike. For now, the pilot in Edo stands as a compelling demonstration of AI’s potential to reshape learning—one prompt, and one student, at a time.
