What AI in the Global South Reveals About Scaling Education Innovation

“A common misconception is the idea that it is possible to change an education system through ‘siloed’ interventions, silver bullets that can tackle a single shortcoming in how an education system functions.”

Fernando Reimers, Ford Foundation Professor of the Practice of International Education and Director of the Global Education Innovation Initiative at Harvard University

Artificial intelligence is becoming part of how education systems approach instruction, assessment, and administration. Its impact, however, is not determined by the sophistication of the technology alone. AI enters schools and ministries shaped by fiscal limits, governance structures, curriculum alignment, teacher capacity, and uneven access to infrastructure. These factors largely determine whether new tools strengthen learning or place additional pressure on institutions already managing competing priorities.

Nowhere are these dynamics more visible than in the Global South, where education systems serve the majority of the world’s students while facing persistent demands to expand access and improve quality. Many countries confront large class sizes, constrained public budgets, and shortages of trained educators, even as they are asked to prepare students for economies increasingly influenced by automation and digital technologies. In this context, AI is often framed as a means of accelerating progress, but its adoption also exposes the institutional limits that shape how reforms are absorbed at scale.

Decades of education reform experience suggest that technology introduced without alignment to standards, curricula, and accountability mechanisms rarely produces sustained change. Initiatives that bypass teacher preparation, governance capacity, or system coordination tend to remain fragmented, creating parallel practices rather than integrated improvement. As AI applications enter classrooms and ministries, similar patterns are emerging, particularly in systems with significant variation in institutional capacity across regions and schools.

These dynamics are central to the new work Artificial Intelligence and Education in the Global South: A Systems Perspective, which examines how AI interacts with the structural realities of education systems across diverse national contexts. Rather than treating AI as a solution in itself, the book focuses on how infrastructure, human capacity, and governance coherence determine whether technology supports learning or compounds existing challenges.

To examine what this level of coordination requires in practice, we spoke with Fernando M. Reimers, Ford Foundation Professor of the Practice of International Education at Harvard University and director of the Global Education Innovation Initiative, whose research focuses on how education systems can build the institutional capacity needed to improve learning outcomes at scale.

Fernando Reimers is the Ford Foundation Professor of the Practice of International Education and director of the Global Education Innovation Initiative at Harvard University. He is an elected member of the U.S. National Academy of Education, the International Academy of Education, and the Council on Foreign Relations. He has just published a book, co-authored with Zainab Azim, Maria Renee Palomo, and Callysta Thony, titled Artificial Intelligence and Education in the Global South: A Systems Perspective.

Professor Reimers is an expert in global education. His research and teaching focus on understanding how to educate children and youth to thrive in the 21st century. He was a member of UNESCO’s International Commission on the Futures of Education, which produced the report Reimagining Our Futures Together: A New Social Contract for Education. He has developed curricula aligned with the United Nations Sustainable Development Goals, which are used in many schools worldwide. During the Covid-19 pandemic, he led numerous comparative studies examining the pandemic’s educational consequences and identifying options to sustain educational opportunity and build back better.

At HGSE, Professor Reimers serves as co-director of the new International Education and Policy Management Pathway, an online master’s degree program focused on international education. The program launched this fall with 72 students from 33 countries.

He directs the Global Education Innovation Initiative, a cross-country research and practice collaborative focusing on education for the 21st century. He has written or edited 56 books and over 100 articles and chapters, including Artificial Intelligence and Education in the Global South: A Systems Perspective, Educating Students to Improve the World, Schools and Society During the Covid-19 Pandemic, Education to Build Back Better, Primary and Secondary Education During Covid-19, and University and School Collaborations During a Pandemic.

As part of his commitment to advancing educational opportunity and international education, he serves on multiple advisory boards and committees at Harvard, particularly those focused on advancing the university’s global mission and enhancing the effectiveness of its programs to address climate change and advance internationalization. He serves on Harvard’s University Committee on International Projects and Sites, on the executive boards of the David Rockefeller Center for Latin American Studies and the Center for African Studies, and is a faculty associate of the Weatherhead Center for International Affairs and the Center for International Development.

Professor Reimers also serves on the boards of a range of education organizations focused on improving education and promoting peace, inclusion, and sustainability, including the Fulbright Scholar Advisory Board. He has served on the Harvard faculty since 1998. Prior to that, he worked at the Universidad Central de Venezuela, the Harvard Institute for International Development, and the World Bank.

Any discussion of AI in education in the Global South begins with scale. The majority of the world’s students are educated in low- and middle-income countries, a demographic reality that shapes both the urgency of reform and the complexity of implementation. As Professor Reimers notes, “Most of the world’s students, about nine in ten, are in educational institutions in countries in the Global South.” Unlike education systems in countries with aging populations and declining enrollments, systems in the Global South face sustained demand for schooling and a constant need to expand capacity.

That demand intersects with a broad set of education challenges. Many systems continue to struggle with ensuring that students acquire foundational literacy and numeracy, even as they are expected to prepare young people for economies undergoing rapid technological change. Professor Reimers points to this dual pressure directly, describing the need to support “foundational skills” while also helping students “develop an expanded set of skills in response to the transformation of economies, in part resulting from AI.” In this context, AI is often seen as a potential multiplier, capable of supporting instruction, improving efficiency, and expanding access where human and financial resources are limited.

The book situates AI’s potential across multiple functions within education systems. Professor Reimers explains that underlying these imperatives to improve effectiveness and relevance are “opportunities for AI to enhance instruction, curriculum development, assessment, teacher preparation, school organization, and system governance.” This breadth matters. AI is not confined to classroom tools or tutoring applications; it increasingly touches how systems design curricula, evaluate learning, support teachers, and manage schools at scale.

At the same time, demand does not erase constraints. Professor Reimers is explicit about the barriers shaping adoption. “The barriers are financial constraints, limits to human knowledge and skills, and limited access to devices and connectivity,” he says. These limitations affect not only whether AI tools can be deployed, but also whether they can be sustained and adapted over time. In many contexts, uneven connectivity or device access creates sharp disparities between schools, reinforcing existing inequalities rather than reducing them.

Beyond material constraints, Professor Reimers highlights a design problem that cuts across regions. He points to “barriers resulting from the poor design of AI applications for education,” particularly solutions that address “problems that students, teachers, or administrators do not think they have.” When tools are developed without close attention to classroom realities or administrative needs, adoption falters, regardless of technical sophistication. Meanwhile, he notes, “there are not enough AI developments that adequately address the real problems that students and teachers face.”

For Professor Reimers, this gap between demand and design is not inevitable. “Our hope in publishing this book is to help democratize knowledge about what AI can most productively do to help transform education in the Global South,” he says, “and to help developers work on problems that really matter.” The challenge, as the research makes clear, is not simply scaling technology to large populations, but aligning innovation with the constraints, priorities, and capacities of the systems expected to use it.

This tension between scale and structure sets up a deeper question. Even when demand is clear and tools are available, what institutional conditions must be in place for AI to improve learning rather than add strain to already stretched systems? The answer, Professor Reimers argues, lies not in individual applications, but in the coherence of the systems that adopt them.

The ability of AI to improve learning depends less on the availability of tools than on the capacity of education systems to absorb and integrate them. In contexts where systems are fragmented or weakly governed, new technologies can add complexity without delivering measurable gains. Professor Reimers emphasizes that meaningful adoption begins with the basics of system design. “There needs to be an infrastructure, a system in which there are students, teachers, clear education standards, a curriculum aligned with those standards, and governance and accountability,” he explains.

This emphasis on integration reflects a broader critique of how education reform is often pursued. AI applications are frequently introduced as discrete solutions, layered onto systems without addressing underlying misalignments. Professor Reimers argues that without coherence across standards, curricula, and accountability structures, even well-designed tools struggle to influence classroom practice. “In other words,” he says, “there needs to be a system with some integration and coherence.”

In many education systems in the Global South, institutional capacity varies significantly across regions, districts, and schools. National strategies may exist on paper, but implementation often depends on local conditions, including administrative capacity and leadership stability. When AI initiatives are introduced into this uneven landscape, they can amplify differences rather than close them. Schools with stronger leadership and infrastructure are better positioned to experiment and adapt, while others struggle to move beyond basic adoption.

Professor Reimers cautions against the assumption that targeted interventions can compensate for weak system foundations. “A common misconception is the idea that it is possible to change an education system through ‘siloed’ interventions, silver bullets that can tackle a single shortcoming in how an education system functions,” he says. Education systems, he argues, “are governed by the laws that shape complex systems,” and reform efforts that fail to account for those dynamics are unlikely to produce lasting change.

This systems perspective has practical implications for the introduction of AI. Rather than asking which tools to deploy first, policymakers must consider whether the institutional environment can support sustained use. That includes clarity about learning goals, alignment between curriculum and assessment, and governance mechanisms that allow feedback from schools to inform policy adjustments. Without these elements, AI risks becoming another layer of reform that operates alongside, rather than within, the system’s core functions.

The focus on system coherence also reframes debates about speed. While AI technologies evolve rapidly, education systems change more slowly, shaped by training cycles, budget processes, and political constraints. Professor Reimers’ research suggests that attempts to accelerate adoption without attending to these realities often produce symbolic change rather than substantive improvement. The question is not how quickly AI can be introduced, but whether systems are prepared to support its use over time.

Nowhere is this more evident than in the role of educators themselves. Even in systems with clear standards and governance, teachers remain the primary link between policy and practice. As Professor Reimers notes, institutional readiness ultimately depends on the people expected to use new tools. Understanding how teacher capacity shapes AI adoption is, therefore, central to assessing whether innovation strengthens learning or adds strain.

Across education systems, teachers are the primary agents through whom policy and technology reach students. AI does not change that reality. In fact, its effectiveness often magnifies existing strengths and weaknesses in teacher preparation and support. Professor Reimers repeatedly identifies teacher capacity as a central constraint on whether AI contributes meaningfully to learning. “Teacher capacity appears repeatedly as a limiting factor,” he says, underscoring that tools alone cannot compensate for gaps in professional knowledge or institutional support.

Professor Reimers situates this challenge within a broader body of evidence on what makes professional development effective. “There is a robust body of knowledge about what forms of teacher professional development are effective,” he explains, pointing to research conducted through the Global Education Innovation Initiative. According to that work, effective programs are not episodic or symbolic. They are “integrated into the curriculum and the objectives of schools and teachers,” “reflect teachers’ needs,” and are “embedded in the routines of schools and extend over long periods.”

These features stand in contrast to many AI-related training efforts, which often take the form of short workshops or standalone courses disconnected from classroom practice. Professor Reimers argues that such approaches are unlikely to change how teachers teach or how schools function. Meaningful professional development, he notes, ideally “involve[s] multiple teachers in the same schools and enhance[s] the capacity of the school by connecting it with networks of schools and with other institutions that can bring in additional expertise.” The goal is not individual upskilling in isolation, but collective capacity building that becomes part of how schools operate.

When applied to AI, these principles suggest a shift in emphasis. Rather than training teachers to use specific tools, professional development should focus on how AI supports core instructional and organizational tasks. Professor Reimers points to practical applications that align with teachers’ daily work, including using AI to “increase their own knowledge,” “carry out administrative tasks,” “develop lessons,” and “assess students.” Framing AI as support for existing responsibilities, rather than as an added demand, makes adoption more feasible in resource-constrained settings.

This approach also recognizes the time pressures teachers face. In many education systems in the Global South, teachers manage large class sizes and heavy administrative workloads. Introducing AI without addressing these constraints risks increasing cognitive and procedural burden. By contrast, professional development that is embedded and sustained allows teachers to experiment, adapt, and integrate new tools gradually, within the rhythms of their work.

Professor Reimers’ emphasis on teacher capacity reinforces a broader lesson of the systems perspective. Lasting change depends on aligning tools with institutional realities, not on expecting individuals to compensate for structural gaps. Without investment in long-term professional development, AI initiatives are likely to remain peripheral, used unevenly or abandoned altogether. With it, technology has a greater chance of becoming part of routine practice rather than another short-lived reform.

As the discussion turns to policy, this raises a final question. If teacher capacity and system coherence are central to effective adoption, what does that imply for how policymakers approach AI-driven innovation, not only in the Global South but across education systems worldwide?

For policymakers, the lessons emerging from AI adoption in the Global South point to a familiar but often overlooked conclusion. Education systems do not change through isolated interventions, even when those interventions are technologically advanced. Professor Reimers cautions that a persistent misunderstanding continues to shape reform efforts. “A common misconception is the idea that it is possible to change an education system through ‘siloed’ interventions, silver bullets that can tackle a single shortcoming in how an education system functions,” he says.

This misconception has direct implications for how AI is governed. When policies focus narrowly on deploying tools, procuring platforms, or piloting applications, they risk bypassing the institutional conditions that determine whether those tools improve learning. Professor Reimers emphasizes that education systems operate according to the dynamics of complex systems. “Education systems are governed by the laws that shape complex systems,” he explains, and reform efforts that ignore these dynamics rarely produce sustained change.

From this perspective, governance becomes as important as infrastructure. Policymakers must consider how decisions about curriculum, assessment, teacher preparation, and accountability reinforce or undermine one another. AI introduced without clarity about learning goals or without feedback mechanisms linking classrooms to system leadership is unlikely to strengthen instruction. Instead, it may add administrative burden or operate in parallel to core teaching and learning processes.

While the book’s focus is on the Global South, the implications extend beyond it. Professor Reimers argues that “adopting a systemic perspective, rather than a piecemeal approach, to introducing AI-driven innovation is likely to be fruitful in any education system.” Differences in income level do not eliminate the risks of fragmentation or misalignment. In this sense, the experiences of lower-resourced systems surface challenges that are present, if less visible, across education systems globally.

Taken together, these lessons position AI not as a shortcut to reform but as a test of institutional readiness. Its contribution to learning depends on governance capacity, human capital, and the ability of systems to coordinate change over time. For education leaders facing pressure to modernize under tight constraints, the distinction between system-led adoption and tool-led experimentation may determine whether AI becomes part of durable improvement or another cycle of unrealized reform.