AI is often framed as a global educational revolution, a rising tide expected to transform classrooms everywhere (UNESCO, 2019). Personalised learning, instant feedback, adaptive systems: the promise sounds universal. Tools like Knewton, DreamBox and Khan Academy have become shorthand for that promise, designed to adjust to each learner’s pace and close educational gaps at scale. In practice, that sense of universality rests more on assumption than on evidence. A closer look at public schools across Latin America reveals a more uneven and complex reality.
The global discourse on AI in education is substantial. Organisations such as UNESCO (2023) describe their potential to expand access to quality learning, reduce teacher workload, and personalise instruction at scale. EdTech companies and policymakers echo similar expectations (Pedro et al., 2019), presenting AI as a way to close longstanding educational gaps. Those gaps are not abstract: they include the persistent divide between students who reach higher education and those who drop out before secondary school ends, the distance between well-resourced urban schools and underfunded rural ones, and the unequal outcomes that consistently follow socioeconomic lines. The contrast is important because it shapes how we evaluate any tool that claims to address them.
Governments are drafting national strategies, and universities are beginning to adjust their curricula to include AI literacy, data skills, and computational thinking. Whether this constitutes genuine adaptation is debatable. In many cases, the institutional response resembles rebranding more than structural reform: the same content, delivered through newer platforms, without addressing the conditions that produce inequality in the first place. The momentum reflects a real possibility, and one that makes institutional investment feel urgent.
A different picture emerges when the focus shifts to who can actually participate in this transformation. In many Latin American public schools, the conditions are limited from the start. Internet access is often unstable, devices are shared, and funding constraints shape everyday practice (Pedro et al., 2019). Research mapping the obstacles to AI adoption in developing countries identifies hardware availability, internet reliability, and data costs as primary barriers, none of which are incidental or easily resolved (Nye, 2015, as cited in Pedro et al., 2019). In a classroom where five students work around a single device, tools designed for individual, high-connectivity use lose much of their intended value.
Infrastructure is only part of the issue. AI tools evolve faster than professional development systems can respond, and many teachers are expected to adapt without sufficient training, time, or institutional support. This raises a question that neither governments nor EdTech companies have answered adequately: who is actually responsible for closing that gap?
Public education systems in Latin America have historically underfunded teacher development, and the arrival of AI has not changed that pattern. Governments tend to frame digital transformation as a priority while allocating resources that reflect the opposite. Meanwhile, the companies producing these tools design them for contexts that bear little resemblance to an Argentine or Colombian public school, and rarely invest in the pedagogical scaffolding that would make adoption meaningful.