Series Overview
Designing with Friction: AI, Learning, and K–12 Leadership
Generative AI is changing more than the tools students and educators use. It is changing the conditions under which work is produced, learning is demonstrated, data is remembered, and institutional decisions are made. This five-part series examines what K–12 leaders need to preserve, redesign, and govern as AI becomes embedded in classrooms, platforms, procurement, policy, and school operations.
The central argument is that schools should not respond to AI only through bans, detection tools, or broad acceptable-use language. They need a clearer framework for protecting the human work of learning: thinking, explaining, revising, questioning, judging, and taking responsibility for ideas. That requires distinguishing productive friction, which supports learning, from exclusionary friction, which blocks access without educational value.
In this series:
Part 1: The Illusion of Completion examines why polished student work may no longer be reliable evidence of learning.
Part 2: Writing Policy for the Learning Sequence argues that AI policy should address when AI enters the learning process, not only whether a tool is allowed.
Part 3: From Assistant to Actor explores why agentic AI creates new procurement and governance challenges for school districts.
Part 4: The Compliance Ghost in the Machine focuses on persistent AI memory, student data privacy, and the risks of invisible personalization.
Part 5: Tiered Decision Rights offers a framework for deciding what AI may generate, recommend, automate, and what must remain human.
You are reading Part 3: This post examines agentic AI and why procurement must account for autonomy, memory, tool access, and decision authority.
From Assistant to Actor: Navigating the K–12 Procurement Crisis in Agentic AI
For the past few years, most K–12 AI conversations have focused on generative AI assistants. These systems respond to prompts. A teacher asks for a lesson idea. A student asks for help revising a paragraph. An administrator asks for a draft communication. The human initiates the action, evaluates the response, and decides what happens next.
That model is already changing.
The next procurement challenge for school districts is agentic AI. These systems do not simply respond to prompts. They can pursue goals, retain context, use tools, call external systems, plan multi-step actions, and operate across sessions. In plain language, they move from assistant to actor.
This shift matters because school procurement processes were not built for software that acts on behalf of the institution.
A traditional generative AI assistant produces content. An agentic system may do more. It might monitor student progress, draft and send messages, recommend interventions, update records, schedule meetings, trigger workflows, summarize behavior patterns, or coordinate across multiple platforms. Some of those actions may be useful. Some may be risky. Some may cross lines districts have not yet defined.
The vendor language will not make this easy.
Products will be marketed with familiar words: personalization, efficiency, learning acceleration, workflow automation, educator support, family engagement, and student success. These terms are not meaningless. Many of the promised benefits are real. But the same words can describe very different levels of system authority.
A chatbot that drafts a parent email for teacher review is one kind of tool.
A system that identifies a concern, writes the message, selects the recipient, and sends the email without human approval is another.
A tutoring tool that suggests a hint is one kind of tool.
A tutoring agent that dynamically lowers task complexity, changes the learning path, and records inferred student weaknesses into a persistent profile is another.
These differences are not small. They change the governance problem.
Districts need to begin evaluating AI products according to autonomy, memory, tool access, and decision authority.
Autonomy asks what the system can do without a human prompt or approval. Can it only generate content, or can it initiate actions?
Memory asks what the system retains over time. Does it forget after a session, or does it build a persistent profile of a student, teacher, classroom, or family?
Tool access asks what other systems the AI can touch. Can it connect to the learning management system, student information system, email platform, assessment platform, or intervention database?
Decision authority asks whether the system only recommends action or actually takes action.
These questions need to become standard procurement questions. A district cannot evaluate agentic AI with the same checklist it used for a static instructional website or a simple productivity tool.
The reason is institutional authority.
When an AI system acts inside a school environment, it is no longer just producing text. It may be exercising delegated authority. It may influence what a student sees, what a parent receives, what a teacher notices, what an administrator prioritizes, or what gets recorded as evidence. That is not simply a technical function. It is a governance function.
This does not mean districts should reject all agentic AI. That would be too simple and likely ineffective. Some agentic workflows may reduce clerical burden, improve accessibility, help staff identify patterns sooner, or support more timely communication. But the more autonomous the system becomes, the stronger the governance must be.
Districts should require vendors to answer direct questions before pilot approval.
What actions can the system take without human review?
Can those actions be turned off by role, school, grade level, or use case?
What data does the system retain across sessions?
Can memory be inspected, corrected, exported, limited, or deleted?
What external systems can the agent access?
Are all agent actions logged in a human-readable audit trail?
Can the district set different approval rules for low-stakes, moderate-stakes, and high-stakes actions?
Can the system make or influence decisions related to placement, discipline, special education, grading, intervention, or family communication?
The final question is the most important: Where is the human decision point?
If the vendor cannot answer that clearly, the district should not proceed.
K–12 leaders are not only buying software. They are defining the boundaries of institutional delegation. Agentic AI makes this visible because it can act in ways prior tools could not.
The procurement crisis is not that these tools are coming. The crisis is that they may arrive inside familiar product categories before districts have developed the language to govern them.
We need that language now.
