Foundations First: Why Artificial Intelligence Cannot Solve Pakistan’s Primary Education Crisis

The global fascination with artificial intelligence in classrooms has reached a fever pitch. Ministers in Organisation for Economic Co‑operation and Development (OECD) countries have spent unprecedented time and resources on AI integration, creating departments, hosting conferences, and rewriting curricula. Yet even in wealthy systems, the results are sobering. The latest Programme for International Student Assessment (PISA) findings show that average reading scores across OECD countries have fallen sharply since 2012, dropping thirty‑five points in just over a decade. This decline is not a minor fluctuation; it is a warning that the obsession with digital solutions has distracted from the basics. If well‑resourced systems are struggling to keep children reading, what hope is there for Pakistan’s classrooms, where millions of children cannot yet recognize letters or numbers.

In Pakistan, the urgent task is not to wire classrooms with AI but to bring more than thirty‑five million children into schools and ensure that those schools deliver foundational learning. Nearly half of children aged five to sixteen remain out of school, dropout rates are high, and surveys repeatedly show that less than half of grade five students can read a grade two text. Infrastructure is insufficient, teachers are under‑trained, and cultural barriers exclude children who are economically active or who live with identities viewed as marginalised. The conventional model of expansive school buildings, rigid catchment areas, and bloated recruitment has proven expensive and ineffective. Schools are often inaccessible, misaligned with communal realities, and unsustainable in fiscal terms. They fail to welcome children who work part‑time or those from refugee, nomadic, or migrant families, and their schedules clash with the economic rhythms of poor households.

The barriers to AI integration in Pakistan’s primary schools are even more pronounced. Electricity outages remain frequent, internet connectivity in rural and remote locations is patchy, and poverty prevents families from affording devices. As one of the World Bank studies reveals, most children who are excluded from schooling belong to low‑income strata, making the idea of equipping them with AI gadgets at home unrealistic. Even for those who attend school, retaining a continuous connection with AI tools outside the classroom is improbable. Recent studies on screen time have also raised alarms: excessive exposure is linked to weakened memory consolidation, dopamine spikes that encourage addictive behaviour, and a preference for visual inputs over fine‑motor skills that are critical for early learning. Research from UNESCO and UNICEF has warned that prolonged screen use reduces attention span, fosters social seclusion, and undermines the interpersonal skills that are essential for collaborative learning. In this context, promoting AI in Pakistan’s primary schools risks compounding the crisis rather than solving it.

Decentralisation must be accompanied by institutional structures that tie fiscal flows to measurable learning outcomes, with union councils held accountable for literacy and numeracy targets.

The private sector has expanded rapidly, enrolling more than a third of school‑age children. Mohallah‑based schools provide low‑fee access, especially for girls, while larger chains cater to middle‑class families. Provincial governments have experimented with Education Foundation models, subsidising private providers through public‑private partnerships (PPP). These initiatives have expanded access, reduced gender gaps, and introduced modern teaching methodologies, but they remain under‑scaled and unevenly implemented. The dilemma of private schooling is that while it has helped raise enrolment and retention, its contribution to learning outcomes and national development remains uncertain. Many private schools focus on grades rather than comprehension, producing stereotypical performers who lack critical thinking. Others, especially low‑fee schools, replicate the weaknesses of public schools with little improvement in pedagogy or schooling facilities. And in elite chains, teachers are often overloaded with performance metrics, leaving little room for genuine learning conversations with children. The result is a generation of students who may secure high marks but lack conceptual clarity and critical thinking skills.

From a or policy point of view, large, centralised schools impose heavy recurrent costs in salaries and maintenance without delivering commensurate learning outcomes. Fiscal decentralisation to the lowest tiers, i.e. union councils, would allow resources to be allocated closer to communities, with accountability tied directly to educational attainments. As one of the World Bank studies reveals, decentralisation is most effective when authority is sequenced with capacity, ensuring that local institutions are prepared to manage resources and deliver results. Ethiopia’s gradual decentralisation illustrates how building local capacity before transferring authority mitigates fragmentation. India’s Panchayati Raj committees show how local councils can oversee school management and resource allocation, embedding accountability at the community level. Brazil’s fiscal decentralisation demonstrates both the potential and the risks: resource transfers empowered municipalities, but without strong accountability mechanisms inequalities widened. For Pakistan, the lesson is clear: decentralisation must be accompanied by institutional structures that tie fiscal flows to measurable learning outcomes, with union councils held accountable for literacy and numeracy targets.

Economic viability is equally critical. Small and community‑based schools are less costly to establish and maintain than large buildings, and they deliver higher dividends in terms of access and retention. Studies on non‑formal education models in South Asia show that one‑room schools staffed by community teachers can achieve literacy and numeracy within thirty‑six months, at a fraction of the cost of formal schools. By employing local teachers, these models also generate livelihoods, creating a virtuous cycle of economic and social dividends. Allowing multiple shifts ensures that children who are economically active can access schooling without abandoning household contributions. This flexibility is essential in Pakistan’s cultural, geographical and economic landscape, where many children juggle education with work.

The cultural dimension reinforces the governance argument. Pakistan’s schools often fail to accommodate children who are economically active, forcing them to choose between work and learning. Timetables clash with household survival strategies, and schools rarely offer flexible shifts. Gender norms further restrict access, with girls facing long commutes, unsafe environments, and parental mistrust. Community‑based schools, by contrast, embed education within the rhythms of daily life, offering short sessions at safe times and places, and involving mothers and local women in governance. This feminisation of foundational learning is not a side agenda; it is central to resilience and equity. Girls have consistently outperformed boys in recent assessments, and protecting those gains requires explicit targets, safe environments, and female leadership in classrooms and communities.

Pakistan’s education reform must therefore begin with a realistic appraisal of its landscape. Artificial intelligence integration may be fashionable, but it is not relevant to classrooms where children cannot yet read or write, and teachers lack basic pedagogical support. By scaling up public‑private partnerships through provincial education foundations, training community teachers, authorising outcome‑based assessments, and fostering genuine community ownership, Pakistan can build a primary education system that is sustainable, inclusive, and effective. As one Harvard‑affiliated school design engineer argued in the Leaders of Learning programme, the future lies not in large institutional buildings but in smaller, focused learning environments designed around learners’ specific contexts. These structures prioritise empathy and proximity over architectural grandeur, and they are precisely what Pakistan needs to overcome its foundational learning poverty. Only then will artificial intelligence and other advanced tools find meaningful application. Until that foundation is laid, pencils and copies remain more relevant than algorithms and apps. The reform agenda must be clear: get children into schools, strengthen foundational learning, and build community‑owned systems that endure.