Giving every student a chatbot is not the same as giving them a future they can shape. Across EdTech UK, access to AI tools is expanding quickly. What is slower to arrive is the harder work: helping learners decide when a tool is useful, when it is noisy, and when the most important move is still their own judgement.
At Cambridge University EdTech Society, we keep returning to a simple distinction. Access is the door. Agency is the walk through it — the capacity to question, make, refuse and revise. Meaningful AI education is not a product rollout. It is a practice of thinking with, and sometimes against, the systems now sitting in classrooms, studios and lecture halls.
That is why “access to AI is not the same as agency” is more than a slogan. A student who can open a model but cannot explain its limits, locate its sources, or choose a better method has been given a shortcut, not a craft. The more powerful the tool, the more the curriculum has to protect curiosity, evidence and responsibility.
Builders already know this from the other side of the table. In hackathons and classroom pilots, prototypes look impressive until they meet a real timetable, a real safeguarding rule, or a learner who does not share the designer’s assumptions. From prototypes to practice, the lesson is consistent: tools earn their place when teachers and students can use them without losing the plot of the lesson.
Young people are not asking for more dashboards. They are asking for work and learning that still feels human — for spaces where they can try, fail, collaborate, and be taken seriously. The future they describe is less about automation as destiny, and more about having a say in how technology is used around them.
Rethinking assessment
If AI can generate a first draft in seconds, assessment cannot stay a contest of fluent first drafts. A more human future for assessment would value process, oral reasoning, making in public, and the ability to interrogate a generated answer. That does not mean abandoning rigour. It means measuring the thinking we actually want to keep.
The skills that will matter
The skills that will matter most in 2030 are not a secret list of prompts. They look closer to old academic virtues under new pressure: asking a better question, checking a claim, designing with other people, and noticing when a system is wrong. AI education should make those muscles stronger, not optional.
Meaningful AI education, then, is not about keeping up with every new model. It is about keeping learners in charge of their own learning — with better tools, clearer judgement, and a community that still believes ideas are worth arguing over.