As artificial intelligence reshapes education globally, a new framework is emerging to separate promising tools from expensive failures: every AI investment in schools must pass three critical tests. According to the Global Partnership for Education, technology must be effective at improving learning outcomes, safe for protecting children's data and development, and equitable in reaching historically excluded students. Without meeting all three standards, governments risk squandering resources on tools that sound innovative but deliver little real benefit to classrooms.

The stakes are enormous. Millions of children worldwide remain out of school or attend classes where they aren't actually learning foundational skills like reading and math. When these basics are mastered, the effects compound throughout a lifetime, supporting employment, well-being, and economic growth. Yet the pressure to adopt AI quickly is intense, with tech companies and vendors promoting solutions faster than evidence can validate them.

What Makes AI Education Tools Actually Work?

The research is clear on one point: the technology itself matters far less than how it's used within a school system. A sophisticated AI tutor won't help students who lack basic classroom infrastructure, trained teachers, or a clear curriculum. Instead, experts emphasize that AI works best when it reinforces proven teaching methods, keeps teachers at the center of decisions, and complements rather than replaces human instruction.

This insight challenges the common assumption that more advanced AI automatically means better learning. A Google Deepmind learning expert explained the misconception plainly: expecting a large language model (LLM), a type of AI trained on vast amounts of text, to teach effectively without intentional design is like "taking any stranger off the street, putting them in front of a classroom and saying, 'great, now you're a teacher'".

"It requires intentional building of LLMs to basically be able to tune it to be good at pedagogy to reinforce what we know about learning science and not assume that will happen without some kind of intervention," stated Miriam Schneider, director of learning initiatives at Google Deepmind.

Miriam Schneider, Director of Learning Initiatives, Google Deepmind

The implication is significant: companies developing AI for education must ground their tools in learning science, not just in raw computational power. This means testing, learning, and adapting as tools evolve, ensuring interventions remain effective and relevant to actual classroom needs.

How to Evaluate AI Education Investments

• Effectiveness Standard: Does the technology improve learning outcomes at scale? The ultimate measure is whether children are actually learning more, with sustained improvements in early grade literacy and numeracy that reinforce proven teaching approaches.
• Safety Standard: Does it protect children's data and prevent exposure to harmful content? As AI tools operate at scale and speed, safeguarding becomes increasingly urgent, requiring deliberate action to consider longer-term impacts on learning and development.
• Equity Standard: Does it expand opportunity for historically excluded learners? Access alone isn't enough; technology must be intentionally designed with strong system capacity and sustained investment to support inclusive learning for children with disabilities, in remote areas, or learning in different languages.

Beyond these three core tests, governments should also demand that investments be human-centered, localized, and cost-effective. Human-centered means technology strengthens teachers and reduces administrative burden rather than replacing judgment. Localized means adapting to curriculum, language, and infrastructure realities specific to each region. Cost-effective means understanding total system costs and long-term sustainability, not just adoption rates.

Why the "All or Nothing" Debate Is Slowing Progress

Some universities and schools have responded to AI concerns by banning the technology outright. But experts warn this approach may backfire. Rather than shutting down conversations about innovation, the debate around AI should focus on how to embrace more holistic teaching methods that develop critical thinking, relationships, and motivation alongside knowledge transfer.

"I think the risk is we don't end up having them if we make it all or nothing," Schneider noted, referring to important conversations about the future design of education systems.

Miriam Schneider, Director of Learning Initiatives, Google Deepmind

The real opportunity lies in using AI debates as a reflective pause to ask bigger questions about what education should accomplish. Should schools focus primarily on knowledge transfer, or should they develop skills like creativity, collaboration, and emotional intelligence that AI cannot easily replicate? Technology can play a role in answering these questions, but only if the conversation remains open and grounded in evidence rather than fear or hype.

Three Tactical Priorities for Getting AI Right

To move beyond debate and toward real impact, the Global Coalition for Foundational Learning identifies three concrete priorities for governments and education partners. First, strengthen the evidence base through rigorous field research on what improves learning at scale, generating timely evidence throughout each stage of technology development. Second, invest in digital public goods like shared platforms and content that reduce costs and expand equitable access while improving quality for local languages and low-resource contexts. Third, build a coordinated ecosystem that aligns global and local partners, including the private sector, around shared priorities and country needs.

This moment presents a limited window of opportunity. As new technologies are being developed and adopted, there is still scope to shape how they integrate into education in ways that prioritize equity, effectiveness, and safety. The decisions made today will shape education systems for years to come, making it crucial that governments move forward with intention rather than haste or hesitation.