A child can now ask a machine for a story, a logo, a game concept or fifty ideas for saving the planet. The result may arrive before the child has decided what they care about.

This creates a strange educational problem. The ability to produce has expanded faster than the ability to choose. We can get more output with less effort, but output was never the same thing as authorship.

Future Designers is built around a capability we call creative agency: the power to make, paired with the authority to decide.

The phrase sounds simple. The science underneath it is not. Agency is a mature idea in psychology and education. Creative agency is newer and less settled. There is no single scientific definition or universally accepted measurement scale. We should say that plainly.

Still, several research traditions point in the same direction. Albert Bandura described human agency through intentionality, forethought, self-regulation and reflection. Work on causal agency asks whether a person can set a goal, act towards it and believe their actions can affect the outcome. Educational research on autonomy adds an environmental condition: people need genuine opportunities to choose and act, not decorative choices where every road leads to the teacher’s preferred answer.

Creative agency brings those ideas into making. It is the capacity to imagine possibilities, judge which one deserves attention, turn a choice into action and retain authorship as the work changes.

Creativity produces possibilities. Agency turns a possibility into a course of action. Creative agency needs both.

Four parts of creative agency

Our working definition has four connected parts.

1. Form an intention

Creative agency begins before production. A person notices a problem, frames a question or decides that something should exist.

For children, this might be a sketch for a quieter school playground, a question about local flooding or an idea for helping a classmate navigate the building. The idea does not need to be brilliant. It needs to be theirs enough that they can recognise when the process starts drifting away from it.

This is why "think first" is a serious educational move, not nostalgia for pencils. If AI supplies the first framing, children begin inside the machine’s probability distribution. They can still make choices, but the first and often most influential choice has already been made for them.

2. Generate and explore possibilities

Agency without creativity can become efficient execution of somebody else’s brief. Creative agency includes the freedom to question the frame, combine unlikely ideas and imagine alternatives.

AI can be useful here. It can offer variations, simulate perspectives or help a child visualise something that would otherwise remain stuck in their head. For a pupil who struggles to draw or write fluently, that support can open a door.

But more options do not automatically create more agency. Fifty generated concepts can leave a child browsing rather than designing. Exploration needs an intention and a stopping rule.

3. Exercise judgement

Someone must decide what is accurate, useful, original enough, fair and appropriate for the people involved. In an AI-supported process, that work becomes more important because plausible output is plentiful.

Judgement includes checking facts and spotting bias, but it goes further. It includes taste: recognising which idea fits the purpose, which image feels generic, which compromise weakens the work and which imperfection gives it character.

Two recent conversations from the frontier of AI product development make this shift visible. Max Schoening, Notion’s head of product, argues that agency matters more as AI makes the first part of building easier. He also notes that more software has not automatically produced better software. Andrew Ambrosino, who leads development of the Codex desktop app at OpenAI, describes software becoming cheaper to build while taste becomes more important.

These podcast conversations are not research about children, and they should not be used as such. They are signals from workplaces already experiencing the change. When technical production gets easier, the bottleneck moves towards choosing, directing and evaluating. Schools should notice before the same pattern is presented to children as a productivity upgrade.

4. Act, revise and own the result

Ideas become agency when they affect the world. The child plans, makes, tests, receives feedback and changes course.

The result can be modest. A cardboard prototype that makes one corridor easier to navigate is enough. What matters is the connection between intention, action and consequence: I decided, I tried, something happened, and I can learn from it.

Authorship does not require doing everything alone. Designers use tools, references, teachers, teammates and feedback. AI can join that list. Authorship means the child remains able to explain the important decisions and accept responsibility for the final result.

That last part is easy to lose. When a machine performs much of the visible work, adults may praise the polished output while missing that the child made few consequential decisions. A beautiful poster can hide a very thin learning process.

Creative agency is not unlimited freedom

There is a tempting but unhelpful interpretation of agency: let children do whatever they want.

Children need structure, domain knowledge, feedback and boundaries. A blank page can feel less liberating than a well-designed challenge. The teacher’s role is to create a frame with real room inside it.

Research on causal agency supports this active teaching role. In the Self-Determined Learning Model of Instruction, teachers help students become causal agents by teaching them to set goals, solve problems and regulate action. Agency grows through guided practice, not adult disappearance.

For children aged 9 to 12, structured autonomy might mean offering a real design problem, clear safety constraints, accessible materials and explicit checkpoints. Children choose the direction, compare options and own decisions. Teachers protect the learning process and ask the questions that a fluent AI system will not ask on its own.

This also explains why teacher agency matters. A teacher with no room to adapt a curriculum will struggle to give pupils meaningful choices. If we want creative children, teachers need permission to exercise professional judgement too.

What AI changes

AI can support creative agency when it expands a child’s ability to test an intention. It can help produce a rough visual, translate a question, suggest a counterexample or make alternatives visible.

It can erode agency when it removes the thinking that the lesson was meant to develop. If the learning goal is to form an argument, asking AI to draft the argument is not scaffolding. If the goal is to compare evidence, accepting an AI summary skips the comparison. If the goal is to generate an original direction, beginning with twenty machine ideas changes whose direction it is.

The relevant question is not "Did the child use AI?" It is "Which decisions remained with the child?"

Future Designers turns that question into a classroom routine:

• Start with your own idea, sketch or question.
• Steer the machine with a clear intention.
• Interrogate the output for errors, omissions and bias.
• Curate and change what comes back.
• Stay responsible for the final call.

Questions about truth, fairness, privacy and environmental cost run through the whole process. Responsibility cannot be postponed to an ethics slide after the exciting tools.

Why this matters now

Education has often used visible production as a rough proxy for thinking. The essay, drawing, model or presentation gave teachers something to inspect. Generative AI weakens that proxy. A polished product may now reveal very little about who formed the intention, evaluated alternatives or understood the result.

That does not make creative work less important. It makes the process harder to see and more important to design.

Schools will need to value evidence of judgement: first sketches, rejected options, source checks, explanations, revisions and the reasons behind a final choice. Assessment should ask what the learner decided and why, not only whether the final object looks accomplished.

Creative agency is therefore more than a defence against AI dependence. It is a positive educational goal. Children should leave school believing that their ideas can affect the world, knowing how to turn intention into action and having practised the judgement required to do so responsibly.

AI can increase what children are able to make. Education must increase their authority over why, when and how they make it.

That is the Future Designers wager: the most important outcome is not a generation that can prompt machines. It is a generation that can create with powerful tools and still recognise itself as the author.

Sources

• Albert Bandura, Toward a Psychology of Human Agency, 2006.
• Michael L. Wehmeyer and colleagues, Promoting Causal Agency: The Self-Determination Learning Model of Instruction, 2000.
• Future Designers raw research synthesis on agency and creative agency, drawing together agency theory, developmental research and emerging work on children and GenAI. AI-generated synthesis was used for discovery; claims were limited to named frameworks and sources.
• Lenny Rachitsky with Max Schoening, Why cultivating agency matters more than cultivating skills in the AI era, 2026.
• Lenny Rachitsky with Andrew Ambrosino, OpenAI Codex lead on the new shape of product work, 2026.