This resource is provided by ACSA Partner4Purpose Learning Genie and was written by David Ross.
Spend enough time in education, and you begin to recognize when a debate is missing the point. The current conversation about artificial intelligence is one of those moments.
Across districts, the questions sound familiar: Should students be allowed to use AI? How do we prevent misuse? What policies do we need in place? These are reasonable concerns. They are also, increasingly, the wrong place to begin.
While we are debating the tool, the nature of the work has already changed. AI can now produce polished outputs in seconds. The results are often indistinguishable from what we have historically considered proficient student work. But polish is not mastery. Speed is not ownership. And a finished product tells us less and less about the thinking that produced it.
This leads to a more consequential question, one that cuts across every classroom, every assignment, and every system we have built: Who is the agent in the learning process?
The Instructional Shift We Can No Longer Avoid
Agency is not accidental. It is designed. Agency must be systematized—designed into the structures, expectations, and measures that shape daily practice.
My team at the Buck Institute for Education (now PBLWorks) arrived at a similar realization 15 years ago when we were providing professional development focused on standards-focused project-based learning. The workshops we offered hundreds of times per year were equipping teachers with the knowledge and skills needed to implement PBL in their classrooms. Our analysis revealed a harder truth: Implementation was failing because we had not created the systemic conditions the pedagogy required for long-term success.
AI is forcing a reckoning with long-standing assumptions about teaching and learning. For years, we have organized school around the production of work. Assignments are completed. Standards are covered. Student work products are graded. For all its imperfections, the system has been internally coherent. That alignment is now beginning to fracture.
When a student can generate a well-written essay without engaging deeply in the thinking required to produce it, the value of the assignment changes. Not because writing no longer matters, but because the act of writing is no longer a reliable proxy for understanding. What cannot be outsourced is the process. The decisions students make. The strategies they choose. The way they respond when something doesn’t work.
This is the shift now in front of us:
- From product to process.
- From coverage to ownership.
- From compliance to purpose.
It is not a philosophical shift. It is an operational one. And it cannot be solved at the classroom level alone.
When AI Becomes Part of the Design
The instinct in many school systems is to treat AI as something to control. But the classrooms we have examined suggest a different path. In each case, AI did not replace the teacher. It did not replace student thinking. It changed what was possible in the design of learning.
This is where the conversation must shift for system leaders. If AI is positioned only as a student tool, the district will spend its time managing behavior. If it is positioned only as a teacher productivity tool, the instructional model remains largely unchanged. But when AI is embedded into intentional design—particularly within a UDL framework—it begins to function differently. It becomes a direct support for the very capacities we are trying to develop.
Students can use it to clarify goals, explore options, revise thinking, and reflect on their work. Teachers can use it to translate student input into structured learning experiences without losing coherence. In this role, AI is not a shortcut. In this role, AI is not a shortcut. It is an accelerator of agency, amplifying students’ ability to direct, refine, and reflect on their own learning.
What It Means to Take This to Scale
At this point, the risk is misidentifying the level of the work. It is tempting to treat agency-based learning as a set of instructional strategies, as though it were something that lives in classrooms and spreads through professional development. That approach will produce pockets of excellence. It will not produce system change. This is another hard lesson we learned at the Buck Institute.
The classrooms we have highlighted are not anomalies. They are signals. They show what becomes possible when design, measurement, and belief begin to align. For that alignment to hold, systems must take a more deliberate approach.
The first step is making agency visible. This means moving beyond abstract language and into disciplined measurement. Tools like the Hope Scale for Children allow districts to generate a Student Agency Dashboard that tracks trends across schools and grade levels and provides early signals of disengagement.
But measurement alone does not change outcomes; it simply reveals them.
The next step is building the system’s capacity to design differently. This is where many initiatives falter. Teachers are asked to change practice without being given the time, structures, or support to do so. In a systemic approach, design becomes a shared responsibility. Teams co-create units. They test co-construction protocols, such as EduProtocols. They analyze student work and iterate in short cycles. High-yield strategies such as structured inquiry, collaborative protocols, and reflection routines become part of the system’s common language.
The work becomes less about individual innovation and more about collective refinement.
From Initiative to Infrastructure
What ultimately determines whether this work takes hold is not enthusiasm. It is alignment. Agency must be reflected in what the system values and reinforces. The values that are displayed on plaques in school offices or enumerated in the Portrait of a Graduate competencies stored on a district’s website must be operationalized.
In curriculum, this means moving from fixed pathways to flexible frameworks that maintain standards while allowing for student input. In instruction, it means designing experiences that require decision-making, not just task completion. In assessment, it means making the learning process, not just the final product, visible. In professional learning, it means shifting from workshops to ongoing design cycles grounded in real classroom work.
When these elements operate in isolation, agency remains episodic. Students experience it in one classroom and lose it in the next. When they operate together, it becomes structural. And when it becomes structural, it becomes sustainable.
My team at the Buck Institute grew comfortable sharing a difficult truth when they spoke with school administrators, district officials, and board members. To implement PBL (or a similar program, such as UDL) sustainably, a school system should dedicate 3-5 years to the process. Otherwise, everyone involved is wasting time, energy, money, resources, and most importantly, the good will of teachers and students.
The Work Beneath the Work
There is one final layer to this shift, and it is the most difficult to address directly. Agency-based learning is, at its core, a question of trust:
- Trust that students can contribute meaningfully to their learning.
- Trust that teachers can design for complexity rather than follow scripts.
- Trust that systems can evolve without losing coherence.
This is not a technical challenge. It is a cultural one. It shows up in how leaders respond to risk, how they use data, how they define success. It determines whether agency is treated as a priority or a permission. And it is the difference between a system that transforms and one that merely adjusts at the margins.
Where This Leaves Us
The path forward is clearer than we often admit. We have defined agency. We have learned how to measure it. We have seen how to design for it.
The remaining question is whether we are willing to build systems that support it.
This is not a classroom initiative. It is not a program. It is not a short-term strategy. It is a shift in how we think about learning itself.
In a world where AI will continue to evolve, where tools will become more powerful and more accessible, the ability to produce work will matter less than the ability to direct it. Students will need to set goals, identify pathways, adjust when those pathways break down, and persist when the answers are not immediate.
Those capacities do not develop by accident. They are cultivated deliberately, systematically, and over time. And if we are serious about preparing students for that world, then agency cannot remain an aspiration. It must become infrastructure.
David Ross is a writer and consultant in the U.S. and China who focuses on implementing generative AI in K-12 settings. He is the retired CEO of the Partnership for 21st Century Learning and the former Senior Director for the Buck Institute for Education.
Bibliography
Bennett, S. (2012, December 8). Information abundance and its implications for education. User Generated Education. https://usergeneratededucation.wordpress.com/2012/12/09/information-abundance-implications-for-education/
City Year. (2025, November 17). Why building trust in schools matters. https://www.cityyear.org/national/stories/education/why-building-trust-in-schools-matters/
Condliffe, B. (2017). Project-based learning: A literature review (Working paper). MDRC. https://eric.ed.gov/?id=ED578933
Doyle, W. (1983). Academic work. Review of Educational Research, 53(2), 159–199.
Elmore, R. F. (2004). School reform from the inside out: Policy, practice, and performance. Harvard Education Press.
Giray, Louie & Roe, Jasper & Diesta Espiritu, Jairus. (2026). AI writing detectors are ineffective, unreliable and harmful. English Teaching: Practice & Critique. 10.1108/ETPC-07-2025-0155.
Hattie, J. (2009). Visible learning. Routledge.
Novak Education. (2024, May 8). UDL implementation timeline: A month by month ungated resource. https://www.novakeducation.com/blog/udl-implementation-timeline
Penfold, L. (2020, January 12). Harvard Project Zero’s Mara Krechevsky talks about making learning visible. Art. Play. Children. Learning. https://louisapenfold.com/making-learning-visible/
Snyder, C. R., Hoza, B., Pelham, W. E., Rapoff, M., Ware, L., Danovsky, M., Highberger, L., Rubinstein, H., & Stahl, K. J. (1997). The development and validation of the Children’s Hope Scale. Journal of Pediatric Psychology, 22(3), 399–421. https://doi.org/10.1093/jpepsy/22.3.399
Snyder, C. R. (2002). “Hope Theory: Rainbows in the Mind.” Psychological Inquiry.
Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 776–778. https://doi.org/10.1126/science.1207745
