Data from the World Bank is alarming: after the pandemic, it is estimated that 80% of 10-year-old children in Latin America are not capable of reading and comprehending a simple text. Faced with this “learning poverty”, the instinctive response of many governments and organizations has been to distribute devices and digitize content. But the evidence tells us that the problem is not solved with infrastructure alone; it requires pedagogy.

In this article, we explore why technological improvisation often fails and how an intentional, teacher-centered approach, supported by Artificial Intelligence, can generate systemic changes in educational quality.

The mirage of digitalization

Over the last decade, we have witnessed countless “EdTech” initiatives that promised to revolutionize the classroom. Tablets delivered without training, learning platforms that students rarely open, and interactive whiteboards that end up being used as traditional projectors.

The fundamental flaw of these approaches is treating technology as the solution, rather than as a tool to facilitate the real solution: quality interaction between the teacher and the student.

“Technology can amplify good teaching, but no technology can replace a good teacher or compensate for poor pedagogy.”

When we introduce technology without changing the pedagogical model, we are simply digitizing inefficiency.

Three principles for effective technological integration

For technology—and in particular Artificial Intelligence—to have a real impact on learning outcomes, we must follow three fundamental principles:

1. Teacher-centered design

Tools must be designed to empower the teacher, not to bypass them. If a platform requires the teacher to radically change their way of working without offering them a clear and immediate benefit (like saving planning time), adoption will fail.

2. Evidence over enthusiasm

We cannot implement tools based solely on them being “novel.” We need to measure causal impact. Does this tool actually improve reading comprehension? Does it reduce the teacher’s administrative burden so they can dedicate more time to their students? At Mentu Labs, we use Randomized Controlled Trials (RCTs) to answer these questions.

3. Context before code

An AI solution that works in Silicon Valley won’t necessarily work in a rural school in Latin America with intermittent connectivity. Development must be hyper-localized, understanding the infrastructure limitations and cultural realities of each community.

The path forward

AI offers an unprecedented opportunity to personalize learning and support educators at scale. It can help teachers differentiate instruction, generate materials adapted to each student’s level, and provide real-time formative feedback.

But to seize this opportunity, we must abandon the “throw technology at classrooms” mentality and adopt a rigorous, co-designed approach with educators, based on evidence. Only then can we begin to close the profound learning gap in our region.