My living room was a pet shop last Thursday. My daughters spent the morning writing signs, fixing price tags to stuffed animals, setting up feeding stations, and dressing the "pets" in doll clothes. They weren't just playing. They were setting the rules, inventing the story, and managing the customers. My wife and I were cast as both shoppers and shop staff. At one point I was also the delivery driver, the cleaner, and the store security.
That feeling of ownership, of steering the ship yourself, has a name. It's agency. And it's under quiet threat.
What is agency, and why does it matter in schools?
Agency is the capacity to act with purpose and to influence events in line with your own intentions. People often equate it with having options, but it runs deeper than that. Agency is about authorship. It's deciding what matters to you, working towards it, and taking responsibility for the results.
In schools, the term "student agency" appears often in policy documents. The risk is that it becomes tokenistic, offering choice without meaningful control. Real agency means steering the ship yourself.
When people experience agency, they're more motivated and persistent. They treat challenges as problems to solve rather than as reasons to stop. Without it, they tend to withdraw, avoid risk, and wait for others to decide for them.
I've seen this in school. When students are given open-ended challenges, they often run further than you'd expect. I've watched them create apps, build business pitches, and present ideas that go far beyond the brief. One of my school's clubs ended up writing code that ran on the International Space Station. The technical challenge was impressive, but what stood out most was the sense of ownership.
Bandura's four features of human agency
Albert Bandura described human agency as self-organising, proactive, self-reflecting, and self-regulating. He broke it into four features:
- Intentionality. Forming a plan and committing to it.
- Forethought. Anticipating possible outcomes and setting goals.
- Self-reactiveness. Adjusting your behaviour while pursuing the goal.
- Self-reflectiveness. Evaluating your actions and learning from them.
Bandura also introduced triadic reciprocal causation: the constant feedback loop between personal factors, behaviour, and environment. You act on the world, the world acts on you, and your own thoughts and feelings influence how you respond. Think of it as a choose-your-own-adventure book that rewrites itself each time you turn a page.
He also placed self-efficacy at the heart of agency. This is the belief that you can succeed at a specific task. Without it, even the best plan will falter.
Self-efficacy grows from four sources:
- Mastery experiences. You did something hard and succeeded.
- Vicarious experiences. You saw someone else succeed and believed you could too.
- Social persuasion. Encouragement from others.
- Physiological and emotional states. Feeling calm and ready supports action. Anxiety can undermine it.
It's also domain-specific. A student might be confident in the lab but hesitant on stage, or vice versa. This is why schools need to provide varied opportunities to build it.
Three forms of agency schools should recognise
Bandura recognised three forms:
- Individual agency. Acting yourself.
- Proxy agency. Getting others to act for you.
- Collective agency. Working with others to achieve a goal.
That ISS coding project was collective agency at its finest. The pet shop was a mix of individual and collective. Our annual Ideas Lab, where students are given a community problem, a pack of design thinking principles, and free rein with cardboard, glue guns, Lego, and Mindstorms, thrives on collective agency. Teams test, adapt, and sometimes scrap their first attempts when a better idea emerges. When time is up, they pitch their products. You can see the pride in their faces because they know it's theirs.
Agency is not a single skill. It's a blend of abilities, beliefs, and lived experiences that develops through intentional action, reflection, and the confidence that your actions matter.
Self-determination theory: the nutrients agency needs
Edward Deci and Richard Ryan's Self-Determination Theory identifies three needs for intrinsic motivation and agency to flourish:
- Autonomy. Feeling that your actions are self-directed.
- Competence. Feeling capable of success.
- Relatedness. Feeling connected to and supported by others.
When these are met, agency grows. When blocked, it withers. The ISS coding project met all three. Students had autonomy in the design, competence from skill-building, and relatedness from knowing their work would help others.
Julian Rotter's locus of control adds another layer. People who see outcomes as the result of their own actions (internal locus) show greater persistence and problem-solving. Those who see outcomes as controlled by outside forces (external locus) can slide into learned helplessness, a state Martin Seligman described where repeated failure causes people to stop trying.
What AI agents actually are
Now the word "agent" needs a different definition. An AI agent is a software system that can:
- Perceive: take in information from its environment
- Reason: work out what that information means and decide on next steps
- Plan: set out a sequence of actions to achieve a goal
- Act: carry out those actions, not just suggest them
- Learn: improve performance over time based on results or feedback
In plain terms, it's the difference between typing "Book a venue for the school showcase" into a search engine and having a system that finds a venue, checks the dates, compares prices, sends a booking request, and updates your calendar. A search engine gives you options. An agent acts on your behalf.
These systems are growing fast. Anthropic's Claude, OpenAI's ChatGPT, and Replit's coding agents can all run tasks with minimal human prompting. They're powerful, but they've also faced pushback around security and autonomy, especially when connected to sensitive data like school records.
What humans have that machines do not
AI agents can appear to have some of Bandura's traits. They can plan ahead, adjust based on results, and learn from feedback. They can run through hundreds of options faster than any human.
But they don't experience goals, values, or ownership. They're simulating the structure of agency without the lived reality of it. Patrick Haggard draws an important distinction between the unconscious feeling of agency and the conscious judgement of agency. Both shape our sense of being actors in the world. Machines can perform actions and mimic choice. They cannot feel themselves acting.
AI agents are getting better at doing things that look like human agency, just as humans risk doing less of it. That's the paradox schools need to confront.
The danger isn't that AI will somehow gain real agency. It's that we might slowly start giving up parts of ours.
Three traps schools must avoid
Cognitive offloading
It's tempting to hand over more and more thinking. First the AI drafts your lesson plan, then it researches a task, then it writes the school improvement report. You become the approver, not the creator. Over time, those thinking muscles get flabby.
Skill erosion
If students skip the messy middle steps because the agent handles them, they miss the practice that builds mastery. No practice, no deep learning. And no foundation for complex, high-stakes work later on.
Autonomy erosion
If agents start making most of the choices, students and teachers risk becoming passive recipients. The sense of "I made this happen" fades, along with the motivation it fuels.
Five practices to protect student agency alongside AI
1. Keep humans in the loop
Give students and staff active decision points. If the agent drafts a project plan, build in a stage where the human must add, change, or remove at least three things. Turn passive checking into active co-creation.
2. Design for reflection
After any AI-supported task, ask: "Was that the right move?" and "What would I do differently?" In lessons, add a quick reflection stop where students compare the AI's answer to their own and explain which is stronger, and why.
3. Protect skill-building
Ringfence time for key skills without AI help. Structuring essays, debugging code, analysing data. Keep the muscle strong before handing students the power tools.
4. Focus on human strengths
Run challenges that need creativity, empathy, or ethics. Follow up an AI-generated marketing plan with a student-led pitch tailored to the local community. Make it obvious which bits the AI can't do.
5. Anchor in values
Ask students to set project goals based not just on success criteria but shared values: fairness, sustainability, wellbeing. When the agent proposes steps, check them against those values. Teach students that an effective plan isn't always the right one.
The question every school should be asking
Who holds the wheel? The student, or the algorithm?
AI agents are here. They're going to reshape how we learn and work. They can be powerful partners in building human agency, or subtle forces that quietly erode it. The difference will come down to how deliberately schools design around them.
Next year, our Ideas Lab teams will each have an AI bot as a companion. The bot will help with brainstorming, research, and planning, while the teacher guides and the students make the final calls. They'll experience what it's like to work alongside a non-human agent and see clearly how that differs from human agency.
Whether it's code on the ISS, the top corridor buzzing with glue guns, or a living room that has temporarily become a pet shop, the through-line is the same. Protect the conditions that let students act, choose, and matter. That's not something you can delegate to an algorithm.
Matthew Wemyss is an AIGP-certified AI in Education consultant and practising school leader. Book a discovery call to discuss building student agency alongside AI in your school.
