Review by Muhammad Hammad

In recent decades, the internet, social media, and AI have transformed how knowledge is produced and accessed, with AI uniquely able to perform cognitive tasks like writing and reasoning that education systems are built to teach. This has enabled vast informal learning outside formal institutions, offering unprecedented freedom but also creating fragmented, uneven pathways with challenges around quality, validation, and equity.

While fair access to quality education remains key to social and economic progress, its role is now questioned. Unlike past technologies that merely delivered content, AI generates explanations and solutions, requiring intentional adoption grounded in evidence of real learning impact. With limited evidence and rapid change, education systems need a shared framework to filter noise, align innovation with human development goals, and ensure technology strengthens rather than destabilizes learning.

The new context for education

Throughout history, technological waves – from industrialization to computers – have reshaped work and forced education systems to expand access and teach new skills to align human capability with the economy. When that alignment holds, societies benefit; when it breaks, inequality rises. Today, AI and digitalization are driving another shift toward knowledge-intensive economies, where competitive advantage depends on human capital.

Yet investment in human capital is growing slower than IT investment – with AI capital spend forecast to double from 2024-2026 – while education spending as % of GDP has declined since 2021. The signal of academic credentials is also weakening as skills needs evolve faster than qualification frameworks, making the link between education and economic opportunity less predictable. AI adoption in education is already widespread and bottom-up – education is the second most AI-intensive sector – with student use for homework rising from 48% to 62% in 2025 and over 53% of teachers using AI for materials and admin.

This reactive adoption outpaces institutional governance; only 6% of teachers say AI policies are clear. AI offers real opportunities – automating up to 20% of admin tasks amid a projected 44 million teacher shortage by 2030, enabling personalized tutoring at scale, instant feedback, accessibility, and predictive analytics – but evidence remains limited and benefits are not automatic. Without deliberate governance, teacher training, and equity safeguards, AI risks reducing education from a public good to a commodity, undermining its core functions of developing capability, validating knowledge, enabling opportunity, and sustaining civic life.

Risks of unstructured AI adoption

Generative AI automates higher-order reasoning, not just routine tasks. Unlike a calculator that checks work, AI that reasons through a problem removes the struggle that builds understanding via experience-dependent plasticity. One study showed students who used AI performed worse after it was removed than those who never used it. With PISA scores already declining since the 2010s and willingness to learn now in 20% of EU job ads, offloading threatens metacognition, attention, and the neural foundations for complex problem-solving, creativity, and original thought.

LLMs produce fluent hallucinations – 34% more confidently phrased than true statements – training learners to accept plausibility over truth and eroding epistemic vigilance. Primary children are most susceptible, while older students and teachers may propagate errors with added authority. Academic integrity is also strained: 47% of student AI chats seek ready-made answers, and 95% of educators see misuse, breaking the social contract where grades signal genuine understanding.

At the same time, tools optimized for efficiency weaken relational learning; with 40% of US high scholars reporting persistent sadness and belonging fragile post-COVID, reliance on AI risks offloading empathy and reducing resilience, collaboration, and other human-centric skills that the Future of Jobs Report 2025 identifies as most sought-after.

Building education readiness for the age of AI

Current responses to AI in education are fragmented across regulation, pedagogy, and tools, but none alone is sufficient because learning outcomes emerge from interactions across the whole system. The proposed framework organizes a collective response around four interdependent dimensions – enabling foundations, institutional capacities, pedagogical practices, and learning experiences – arranged from system environment to learner experience, and balances standardization for safety and equity with autonomy for agility and innovation.

System-level conditions need more standards – data governance, safety, equitable infrastructure – to enable flexibility closer to teaching. The framework is a diagnostic and strategic tool, not a disruptive fix, to strengthen foundations and move beyond polarized debates. It helps stakeholders conduct readiness assessments across 16 categories, identify gaps, align policy and funding, and pilot approaches.

In practice, policymakers must ensure child-specific data protections and evaluate EdTech ROI; leaders need clear AI policies and protect social learning spaces; teachers must judge when AI supports vs. hinders learning and redesign assessments for reasoning; parents should demand transparency and watch for process vs. output; and students must distinguish AI helping them learn from doing learning for them, while preserving human relationships.

Signals of readiness for AI in education

The framework’s core proposition is that outcomes depend on alignment across its dimensions: system governance that doesn’t reach institutional practice remains aspirational; institutional policy without pedagogical support never reaches classrooms; and pedagogical innovation without learner readiness produces no durable learning. It does not prescribe one AI model or rank systems, but highlights interdependence – ministries depend on schools and vice versa.

While not a scoring system, its signals can be measured through indicators like national AI guidelines, educator training, infrastructure access, and curriculum embedding. This alignment requires foundational guardrails: enforceable safety for identity, privacy, and child-specific data; resilient cyber security and AI sovereignty; and stable, equitable infrastructure with continuous maintenance and financing. As adoption grows, systems must address AI’s environmental costs – varying energy use of models, e-waste, and cloud dependencies – favouring efficient options like small language models, and weigh total cost of ownership against learning impact to avoid diverting scarce budgets.

Beyond that, schools must establish clear IP and authorship rules for AI co-creation, protect physical spaces and time for human connection, play and well-being monitoring, engage families as partners in governance, and build AI/media literacy as foundational competencies. Finally, they must monitor educator workload and redesign assessment from recall to reasoning, reflection, and human skills, using safe innovation sandboxes to test what truly improves learning.