
A teacher at the Cambridge Schools Conference opened proceedings with a passage from Chinua Achebe’s Things Fall Apart: “Eneke the bird says that since men have learned to shoot without missing, he has learned to fly without perching.” The choice carried deliberate weight. In classrooms where artificial intelligence now drafts essays, condenses textbooks, and handles translations with a button press, the pressing question is no longer whether students will encounter AI—but how schools can help them adapt effectively.
Nana Amakah Otu, a teacher at Country International School in Ghana, framed the issue directly: “Like Eneke, education has to adapt” explained Nana Amakah Otu from Country International School, Ghana. “AI can generate essays, summarise texts and translate languages, so we need to refocus the key things students need to learn for their futures.”
Neuroscience findings presented by Professor Rose Luckin, CEO of the AI education consultancy Educate Research Ventures, illustrated the concern. Her research demonstrates that when students depend heavily on AI to produce finished essays or problem solutions, their cognitive engagement diminishes—not just during use, but in the aftermath. The decline persists even after they stop relying on the tools. “The danger lies not in AI itself,” Luckin stated in her keynote address, “but in unchecked AI use.”
Metacognition: The skill AI can’t replace
The proposed remedy centers on redirecting education toward what AI cannot replicate: the learning process itself. Metacognition, the capacity to analyze one’s own thinking, emerged as the most essential skill across multiple workshop discussions. Participants agreed students must cultivate three linked competencies:
- Metacognition: Understanding how learning occurs.
- Self-awareness of cognitive strengths and limitations.
- Perceived self-efficacy: Confidence in applying those skills.
Yet these are precisely the abilities educators find most difficult to teach. Research from Sarah Hughes and Silvia Pirola of Cambridge Assessment International Education revealed that while both teachers and students regard metacognition as vital, it remains the most challenging to develop. The mismatch is stark: AI delivers answers efficiently, but provides no framework for arriving at them, or when to question their validity.
A practical approach gaining support is “AI reflection mapping”, a technique demonstrated in a workshop. Students first generate an AI response to a prompt, then dissect it: identifying its strengths, weaknesses, and knowledge gaps. The critical requirement? Supporting critiques with evidence from credible sources. “The goal isn’t to reject AI,” the presenting teacher emphasized, “but to teach students to use it as a supplementary tool, not a replacement for independent thought.”
Redefining success in an AI-driven classroom
The required changes extend beyond updated lesson plans to a fundamental shift in how education defines success. As Andrew Field, Senior Product Ideation Manager at Cambridge, observed in his session: “Our fundamental aim remains unchanged, joyful, purpose-driven learning, but our tools have evolved. Now we must determine which elements must persist and which must change.”
Field’s challenge to educators, “If AI can produce answers instantly, which future skills now take precedence?”, prompted a unanimous response: metacognition. The dialogue quickly became practical. How can educators evaluate a student’s reflective capacity? How do they distinguish between passive acceptance of AI outputs and active engagement with them? Some advocates pushed for formal assessments of metacognitive abilities, while others proposed weaving reflection into existing curricula.
Assessments that measure real learning
Field’s session explored how AI could be incorporated into assessments to better gauge metacognitive development. Hughes and Pirola demonstrated that current examinations often reward rote memorization over analytical thinking, questioning, or application. Their data showed students who deeply engaged with AI-generated responses, by comparing them to their own work, identifying deficiencies, and justifying their critiques, displayed stronger metacognitive abilities than those who passively accepted AI outputs. A teacher described a pilot where students submitted both an AI-generated draft and a revised version with annotations explaining their reasoning. Initial struggles with self-assessment improved significantly after targeted feedback on reflection.
Implementation hurdles remain significant. Hughes and Pirola’s research found that resource-constrained schools faced obstacles in adoption due to training needs and curriculum revisions. Educators in low-resource settings often relied on free AI tools lacking the safeguards or customization required for structured reflection exercises. A teacher from a rural Kenyan school explained that while students had access to AI, unreliable internet and charging infrastructure limited consistent use, making steady metacognitive practice difficult. Field acknowledged the disparity but argued that even basic integration, such as using AI to generate discussion prompts or peer-review suggestions, could develop critical thinking without advanced technology.