You are reading Week 6 of the Summer Research Series. This post examines how procurement and configuration decisions shape classroom practice before a teacher or student opens an AI tool.
A teacher decides that her students should write before they use AI. They will read the sources, develop an initial position, and produce a first draft on their own. Only then will they use AI to test their reasoning and consider revisions. It is a sound instructional decision. The sequence protects the productive friction students need to form ideas, organize evidence, and confront the limits of their first attempt.
Then the teacher opens the platform the district has provided. The tool lands on a full-draft generator. There is no coaching mode, no way to require an initial attempt, and no setting that holds the student back long enough to think. The teacher is now teaching against her own tool. The decision that put her there was made months earlier, in a room she was not in.
The Decision Was Already Made
This series has moved from the student task to educator judgment and then to the institutional conditions surrounding both. Last week I argued that those conditions are load-bearing. A teacher can ask students to show their thinking, but that practice remains fragile if the gradebook, pacing guide, policy language, and approved technology all reward only the polished product.
This week asks who designs that layer. Districts often treat procurement, configuration, data agreements, and guardrails as administrative work surrounding instruction. They are not outside instruction. They shape what instruction can become. When a district approves a platform, it decides which forms of assistance will be available. When it accepts the default settings, it decides how quickly the system may step into the work. When a policy says that AI is allowed but does not define when or how, the product establishes the practical boundaries. A pedagogy has already been chosen. Governance is a design decision whether or not the people making it describe their work as curriculum.
Compliance Is Necessary but Incomplete
Privacy, security, accessibility, age restrictions, and contractual protections all require careful review. Districts need legal, technical, and data privacy expertise at the table. The problem is not compliance. The problem is treating compliance as the final standard.
A review process may determine whether a platform meets the legal, technical, and privacy requirements for entering the school environment. It may not answer whether the platform protects or removes the cognitive work an assignment was intended to develop. A system can be secure, privacy-aligned, and contractually approved while still generating the argument a student was expected to form. It can meet district data standards while making it harder for a teacher to see what the student attempted independently. Safe and educative are not the same standard. Districts need to evaluate both.
Your Vendor Is Your De Facto AI Policy
Every AI platform arrives with design decisions already built into it. The product determines whether the opening screen offers a hint, generates an outline, or produces a finished response. It determines whether students must attempt the work first, whether prior interactions are remembered, and whether teachers can inspect the process. These are not neutral settings.
A drafting mode that immediately produces finished work makes one claim about learning. A coaching mode that asks students to explain an idea makes another. A persistent memory feature may carry earlier assistance into later work, making it harder to identify what the student understood independently. A high-autonomy setting determines how much the system can do before a teacher or student makes a decision.
The vendor does not know which forms of struggle your curriculum treats as necessary, when a student needs access support, or when that student needs to wrestle with an unfamiliar idea. Those decisions belong to educators. When districts do not make them deliberately, the defaults remain in place. You do not avoid a governance decision by skipping governance. You delegate it.
What Governance as Design Looks Like
Three changes would bring instructional judgment into AI governance more directly:
Put learning in the procurement rubric. Alongside cost, privacy, security, accessibility, and interoperability, districts should ask whether a tool can preserve an unassisted first attempt, support coaching instead of completion, show revision history, and allow features to be limited by role or use case. If the rubric has no row for learning, learning is not being evaluated.
Treat configuration as policy. Memory, permissions, automation, and sharing settings determine what the system can do in practice. Those settings should be selected intentionally and documented. A default accepted to meet a launch deadline is still a policy decision, even when no one names it as one.
Establish decision rights upstream. Teachers need professional discretion, but they should not have to invent the district’s entire AI governance framework for every assignment. The district should define which uses are permitted, which require review, which data may be entered, which autonomous functions must be disabled, and which high-stakes decisions must remain human. Clear institutional boundaries do not weaken teacher judgment. Rather they protect it.
Governance Capacity Is an Equity Issue
Interrogating vendor defaults, testing configurations, and aligning tools with instructional values require time, technical knowledge, curriculum expertise, and leadership attention. Districts with greater capacity are better positioned to configure tools around local priorities. Districts with less capacity are more likely to inherit the product as it arrives.
The next digital divide may be governance capacity. Some districts will be able to author their own AI-supported pedagogy. Others will inherit one from the vendor. That difference may shape how consistently students are asked to think, attempt, revise, and explain rather than rely on systems optimized primarily for efficient completion.
The Leadership Reframe
The leadership task is not to distrust every AI tool or write a longer acceptable-use policy. It is to recognize that districts are already shaping pedagogy at the procurement table and on the configuration screen.
They are deciding what the system may generate, when it may intervene, what it may remember, and what evidence of student thinking will remain visible. Those decisions are not administrative wrapping around instruction. They are part of the instructional design.
Open the AI tools your teachers and students have been given and examine the default settings. Whose pedagogy is running in the classroom right now? Yours or the vendor’s? If the district cannot answer that question, the default already has.
Item #1
Item #2
Item #3
