At our recent Spring Parent Conversation Series event, we talked about the difference between tasks and goals, and what it means when AI can handle the task. We said we are designing for students to have agency over the tool. In this article, I highlight what that looks like in practice in a few middle school classes. What these teachers share is a common instinct: use AI to help the student engage more deeply in their own thinking.
Seventh grade history teacher Maddy Richards had students share their ideas for a final project, creating an ideal nation, with an AI assistant. The assistant helped direct students to which of the five previously studied sources would support their ideas. Critically, it was instructed never to provide quotes or evidence directly. Students still had to return to the original sources on their own to gather supporting material.
This turned the assistant into a reference librarian. It helped students navigate their own prior learning while preserving the rigor of primary source engagement. The design choice to prohibit the AI assistant from giving quotes is the key move. It keeps the boundary between assistance and completion clear.
Maddy also set up an activity that evaluated student papers against her assignment requirements and the evaluation rubrics. The tool provided students with some useful feedback, but it also gave inconsistent grades and sometimes suggested they improve their work by doing something counter to one of the assignment instructions.
Rather than viewing this as a failure, Maddy turned it into a lesson about analyzing different types of feedback. "This was a really good lesson on how to parse feedback and the fact that AI makes mistakes and can be inconsistent," she reflected.
Students learned how to use AI feedback productively and how to evaluate that feedback critically. In professional settings, people rarely receive perfect feedback from a single source. They learn to synthesize input and push back when the advice doesn't hold up. Exposing students to inconsistencies within an AI tool rather than hiding them is just one of the many useful things a teacher can do right now.
AI for Content Generation and Teacher TransparencyFaced with a content-heavy new unit, Maddy used AI to generate materials for simulations and role-based activities that would have taken too long to create by hand. She has been deliberate about disclosure. "I've tried to be very transparent with my students about what has been generated by AI and what has not," Maddy stated. "I would have otherwise avoided it because it's too labor-intensive to create guide sheets for each role."
The bottleneck in ambitious teaching is often time. Maddy used AI to handle the scaffolding materials so she could focus on designing the learning experience itself and facilitating student thinking. And students who see their teacher openly discuss how they are using AI learn that AI use is something to be honest about. Students are now, as Maddy puts it, "understanding some clear boundaries about when it's okay to use and when it is not."
For 6th graders, Ian Walters and his teaching colleague have found AI most effective as a controlled study aid. But they have also found powerful applications behind the scenes.
Following Ian's attendance at the Schools of the Future Conference, he began using AI to evaluate and improve the 6th grade team's most complex history simulation: a two-week competitive game with what Ian called "a true labyrinth of rules." They asked AI to analyze the game through different lenses, suggesting changes to mechanics, gameplay, and reflection activities. The AI identified structural imbalances they may have missed and designed supporting activities and a practice quiz to better prepare students before the game started.
For students, AI has become "a tool that they can add to their toolbox to supply tailored study practice by assignment" during assessment review. Complex simulations are powerful learning experiences and notoriously difficult to balance. Using AI as a design consultant lets teachers create more sophisticated activities without the time investment that would normally make them impractical. And for younger students who may not yet have the judgment to use AI for complex writing or research, a review partner that quizzes them and explains concepts in different ways is exactly the right entry point.
What these examples share is a teacher who thought carefully about what the student should actually be doing. The AI handled something specific. The thinking stayed with the kid. That is what we mean when we say agency over the tool, and our faculty are figuring it out in real time.
