Generative AI did not arrive in schools through a careful adoption cycle, a multi-year committee, or a neatly phased implementation plan. It simply showed up. Students began using it at home. Teachers began experimenting with it for planning and communication. And suddenly, school leaders were asked to answer big questions quickly: What counts as cheating now? How do we protect student privacy? Should we ban tools, embrace them, or do something in between?
As a Special Education Director, I see this moment through a very practical lens: we are responsible for student outcomes, legal compliance, and staffing realities. We also carry an obligation to protect the trust at the heart of schooling: the relationship between students, families, and educators. The good news is that we do not have to pretend there is one perfect answer. The more realistic goal is to build a responsive approach that can evolve.
MIT Teaching Systems Lab’s A Guide to AI in Schools: Perspectives for the Perplexed (Smith, Dukes, Sheldon, Nnamani, Esteves, & Reich, 2025) captures what many of us are living: we are “building the plane while we fly it.” The guidebook is intentionally not prescriptive. Instead, it offers educator and student perspectives, case studies, and questions that help districts craft thoughtful, values-driven decisions.
AI is an “arrival technology,” and that changes the leadership playbook
Most educational technologies enter schools through procurement, pilots, training, and board approval. Generative AI is different. It is an “arrival technology,” meaning it arrived without formal adoption. That matters because it creates uneven access and inconsistent expectations. Some students have powerful devices and unrestricted access at home; others only have school-managed devices. A strict ban in school can unintentionally widen equity gaps if students with more resources continue using AI outside of school while others cannot.
From a district perspective, this “arrival” reality means we need two things at the same time:
Short-term clarity so staff and students aren’t operating in chaos.
Long-term flexibility because the tools, research base, and norms are changing rapidly.
Start with ethics: the questions are bigger than “Is it cheating?”
One of the strongest themes in the guidebook is that educators are not only worried about academic integrity. They are worried about ethics: privacy, bias, accountability, and whether AI use supports or undermines healthy learning and childhood development.
The guidebook highlights ethical principles that are useful for district conversations. Here are a few, translated into school-leader language:
Transparency: When AI is used (by staff or students), are expectations clear? Do we disclose AI use in grading, feedback, or communications?
Justice and fairness: Will the tool produce biased outputs or create barriers for multilingual learners, students with disabilities, or students from historically marginalized groups?
Privacy: Where does the data go? How long is it stored? Can an AI tool infer personally identifiable information even if we do not explicitly provide it?
Responsibility: If something goes wrong (harmful advice, inaccurate content, data exposure), who is accountable: the vendor, the district, the educator, or the student?
Pedagogical appropriateness: Does this tool support research-based instruction, or does it replace the productive struggle students need for learning?
For Special Education leaders, the ethical conversation must explicitly include accessibility and disability rights. AI can be a powerful support for access (simplifying language, offering alternative representations, supporting communication). But a poorly chosen tool can also create discrimination if it is inaccessible, biased, or encourages inappropriate reliance.
What we know about students: benefits and risks are both real
The guidebook is candid: research on AI’s impact on student learning is still limited and mixed, and the tools evolve faster than studies can be published. In practice, districts are making decisions in uncertainty.
From interviews with educators and students, several potential benefits emerge:
Personalized scaffolding: Teachers described using AI to differentiate reading materials or break content into smaller steps for varied skill levels.
“Unstuck” support: Students can use AI like a tutor to ask follow-up questions and get explanations in different formats.
Higher-order thinking opportunities: Some educators hope that if routine tasks are streamlined, class time can shift toward deeper analysis and discussion.
But the risks are equally concrete:
Skill erosion: Writing is thinking. If AI does the composing, students may lose practice in planning, synthesizing, and revising.
Assessment validity: If AI completes work, it becomes harder to know what a student actually understands.
Misinformation: AI can generate confident but incorrect content, creating misconceptions that are hard to undo.
Screen time and attention: Many educators are already concerned about attention span and overexposure to screens; AI can intensify that trend.
What we know about teachers: AI can save time, or create new work
In district leadership meetings, I often hear two competing statements: “AI will reduce teacher workload” and “AI is one more thing.” Both can be true.
Educators in the guidebook reported time savings in areas like:
Drafting lesson ideas and outlines
Creating rubrics and checklists
Generating comprehension questions
Drafting sensitive communications to families (with careful editing)
They also reported time costs, especially around academic integrity monitoring, tool changes, and the pressure to keep up with rapidly evolving technology. Importantly, some teachers expressed concern that outsourcing feedback and grading could weaken teacher-student relationships, because reviewing student work is often how educators learn who students are and what they need.
Policy should be versioned, values-driven, and built with stakeholders
One of the most practical recommendations in the guidebook is to stop pretending policies must be permanent. Consider naming your document the “2025–2026 AI Policy” and explicitly committing to review and revision. That single move lowers anxiety and increases honesty: we are learning.
Strong policy processes described in the guidebook share common features:
Cross-functional teams: Include administrators, general education, special education, related service providers, IT/data privacy staff, and students and families.
Shared vocabulary: Define what you mean by “AI,” “generative AI,” “assistive use,” “unauthorized use,” and “disclosure.”
Case studies: Use realistic dilemmas (for example, AI feedback tools, translation supports, or student “thought partner” use) to ground decisions.
A role for the skeptic: The guidebook highlights the value of including someone who raises concerns so blind spots are surfaced early.
Academic integrity: move from “catching” to “clarifying”
Many schools are tempted to rely on AI detectors. The guidebook raises serious cautions: detectors provide probabilities, can produce false positives, and may disproportionately flag multilingual learners or students whose writing differs from expected patterns. From a compliance and equity standpoint, that should give districts pause.
More sustainable strategies discussed include:
Designing engaging assignments with student choice and ownership
Increasing in-class writing and performance tasks
Adding oral defenses or conferences for major writing assignments
Clearly stating when AI is allowed (brainstorming, outlining, image generation) and when it is not (final drafting, analysis, or skill demonstration), tied to lesson objectives
As a Special Education leader, I will add a critical nuance: “paper-only” solutions can collide with IEP accommodations (typing, speech-to-text, assistive technology). Policies must be written so they do not inadvertently restrict legally required supports.
Where TinyEYE fits: AI conversations should strengthen, not replace, human services
In many districts, the most urgent operational challenge is not whether AI can write an essay. It is whether we can staff speech-language, occupational therapy, mental health supports, and other related services consistently and compliantly.
TinyEYE’s online therapy model is a helpful example of how technology can be used in a way that remains human-centered and service-delivery focused. The goal is not to replace professional judgment with automation, but to expand access to qualified clinicians, reduce service gaps, and support schools facing therapist shortages. As districts consider AI policies, it is important not to lump all “technology” into one bucket. Teletherapy is not generative AI, but AI policies may still intersect with telepractice through privacy, data governance, and vendor vetting expectations.
A simple next step: use a district checklist before you choose tools or rules
Before adopting any AI tool or finalizing broad guidance, consider using a structured checklist like the one highlighted in the guidebook. At minimum, your district should be able to answer:
How does this align with our mission and core values?
What student data might be collected, inferred, stored, or shared?
Does it support accessibility and avoid discrimination?
What is the educational purpose, and what is the evidence base?
What is our plan when tools change overnight and AI features appear by default?
In this moment, leadership is less about having perfect answers and more about building a process your community can trust: clear expectations, transparent decision-making, and a commitment to revise as we learn.
For more information, please follow this link.
