What school is for when answers are everywhere
A student can finish an assignment without understanding it. A student can also understand something important without that understanding showing up neatly in an assignment.
That distinction is one reason I find the conversation about AI and education both promising and difficult. It asks us to look more carefully at what we mean by learning, what we expect a school to provide, and how the adults around a student share responsibility.
I am interested in the possibility of more individual support. I am also wary of jumping from that possibility to the conclusion that education has become a software problem.
More ways to learn, not one final answer
The appeal of an AI tutor is easy to understand. A student might need another explanation, a different example, or more practice with a particular step. A well-designed system could make that support available without requiring an entire class to move at one person's pace.
There is evidence worth taking seriously. A randomized study conducted in a Harvard undergraduate physics course in 2023 and published in 2025 found stronger immediate learning gains with a specially designed AI tutor than with the comparison active-learning lessons. The study involved 194 eligible students and two lessons. It did not establish that any chatbot could replace a teacher or reproduce those results across every age and subject.1
Other findings show why the design matters. A study conducted in a Turkish high school in 2023 and published in 2025 found that students using an unrestricted AI interface improved during assisted practice but performed worse than the control group when the assistance was removed. A version designed with learning safeguards largely avoided that harm, without demonstrating an independent-test advantage over the control group.2
I read those results as a reason to be ambitious about educational design and cautious about blanket claims. Producing a correct-looking answer, helping with a task, and building lasting understanding are different outcomes.
I would want the tools we introduce to be evaluated on the last of those, not merely on the first two.
The work I would want teachers to have more room for
I do not see a useful future in which a teacher watches a room of students interact with screens and is told that their role has become supervision.
The future I would hope for gives teachers more room to exercise judgment. That could mean noticing where a student is struggling, choosing an appropriate challenge, making a connection between subjects, or deciding when the useful next step is a conversation rather than another exercise.
There is also the shared work of a classroom. Students need opportunities to explain themselves, listen to someone else's reasoning, disagree without withdrawing, and make something with other people. I would not want personalization to mean that every student becomes more isolated.
Teachers should have a meaningful role in selecting, shaping, and questioning the tools. Their experience should not be treated as an implementation problem after the important decisions have already been made.
UNESCO's guidance on generative AI in education emphasizes human agency, age-appropriate use, privacy, and educational validation. Those are useful considerations for deciding what an educational tool should be allowed to do.3
They also remind me that convenience is not the only standard. A system that generates a fluent explanation can still be wrong. A tool that works well for one learner might be unsuitable for another.
A more workable relationship with parents
I am concerned about the strain that can develop between families and schools. I do not think it is fair to explain that strain by deciding that one side has stopped caring.
A parent may be trying to understand an unfamiliar situation with incomplete information. A teacher may be managing many students and several competing demands. The student may have a perspective that neither adult has fully heard.
In that situation, more messages do not automatically create a better relationship. A stream of notifications can still leave the important questions unanswered: what happened, what does it mean, and what should we do next?
I would like communication to make responsibilities more understandable. Parents should have meaningful opportunities to participate in their child's learning and help establish habits around technology. Schools should be able to explain their expectations without requiring every teacher to become continuously available. Students should increasingly be included in conversations about their own progress.
Responsibility also needs to be realistic. A parent working unpredictable hours may need a different way to participate from someone who can attend a meeting in the afternoon. A well-designed relationship should accommodate that difference rather than mistake availability for commitment.
Seeing more without watching everything
Grades and assignments provide useful information, but I do not think they should be the only things adults have to talk about.
A student might be becoming more confident at asking questions, showing persistence in a project, helping a teammate, or discovering that they enjoy making something. Those experiences are worth noticing even when they do not fit neatly into a report card.
I would like tools that help students share appropriate parts of that experience and help adults respond constructively. The purpose would be to create better conversations, not a more comprehensive record of a child's every move.
More visibility is not always better. Young people need privacy, room to make ordinary mistakes, and relationships in which a question does not automatically become a permanent entry in a system. Families also differ, so access should not be designed around the assumption that every disclosure is safe in every household.
For me, useful transparency has to include boundaries: who sees something, why they need it, and what the student understands about that access.
Making room for direction
Alongside academic learning, I want schools to remain places where students encounter possibilities they would not have known to search for.
The OECD's 2025 report on teenage career preparation, drawing on PISA data through 2022, describes rising career uncertainty and uneven participation in career-development activities. It also reports associations between career-related experiences and later employment outcomes. That does not prove that phones caused uncertainty, but it does support paying attention to the opportunities students have to explore work and meet people doing it.4
I would not expect every teenager to have a settled plan. Exploration should leave room to change direction. But there is a difference between being undecided after trying things and being unsure because the possibilities have remained abstract.
A project, a placement, a conversation with someone in a trade, or responsibility for a real event could provide a starting point. AI might help a student prepare, reflect, or investigate what comes next. I would want it to extend the experience rather than replace the encounter.
The educational future that interests me combines better tools with stronger human participation. I do not know exactly how the balance will develop. I do know that I would rather judge it by what students come to understand and feel able to do than by how much of the school day we manage to automate.
Sources and notes
- Greg Kestin, Kelly Miller, Anna Klales, Timothy Milbourne, and Gregorio Ponti. AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports 15, 17458 (2025). ↩
- Hamsa Bastani and colleagues. Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences 122(26), e2422633122 (2025). ↩
- Fengchun Miao and Wayne Holmes, UNESCO. Guidance for generative AI in education and research. 2023 guidance; landing page updated January 16, 2026. ↩
- OECD. The State of Global Teenage Career Preparation. 2025. ↩