Frequently Asked Questions¶
This FAQ answers the questions educators, administrators, school board members, and parents ask most often about AI Strategy for Education. Questions are organized into six sections: getting started, AI capabilities and trends, AI strategy and the idea funnel, intelligent textbooks and learning technology, responsible AI and ethics, and governance and implementation.
If a term is unfamiliar, the Glossary defines every key concept used in the course.
Getting Started¶
What is this course about?¶
AI Strategy for Education is a practical strategy course for education leaders who need to understand what AI can do, decide which AI initiatives to fund, manage the risks, and communicate their plans to a school board or board of trustees. It is not a coding course, a product review, or a technology tour. Its core tool is a repeatable decision-making framework called the idea funnel — a six-stage process that takes AI ideas from submission through evaluation, selection, resourcing, and outcome measurement. The course works from a set of near-term, evidence-based assumptions about where AI capability is heading and asks participants to plan as if those assumptions are true. Participants leave with a draft, board-ready AI strategy for their own institution.
See the Course Description for the full overview and learning outcomes.
Who is this course for?¶
The course is written for every decision-maker and stakeholder who shapes AI policy in an educational institution. That includes K-12 superintendents, principals, curriculum directors, and department chairs; classroom teachers; parents and guardians; K-12 school board members; and higher-education counterparts — provosts, deans, CIOs and CTOs, faculty senate members, instructional-design staff, and university trustees. No technical background is assumed. Every technical term is defined before it is used.
If you set policy, approve budgets, manage staff, teach students, or represent families in a school or university, this course was written for you.
Do I need any technical background?¶
No. The course assumes only curiosity about how AI is changing teaching and learning, and a willingness to translate strategy into action for your own institution. Familiarity with your school or district's budget, staffing, and governance process will make the planning exercises more concrete, but no knowledge of computer science, data science, or programming is required.
Technical terms — large language model, xAPI, Learning Record Store, and so on — are introduced gradually and always defined in plain language before being used in a strategic context.
How is the course organized?¶
The course is organized into thirteen chapters that build on each other logically. The first three chapters establish vocabulary and strategic context. Chapters 4 through 7 explore the technology landscape — intelligent textbooks, the idea funnel, and learning telemetry. Chapters 8 through 10 address new pedagogical models and the full risk picture. Chapters 11 through 13 cover governance, agentic AI, and the capstone strategic planning process.
See the List of Chapters for the complete sequence. The Learning Graph shows how all 218 concepts in the course relate to each other.
What will I be able to do after completing the course?¶
After completing the course, you will be able to recall key AI concepts and their strategic implications, run a live idea-generation session at your institution, populate and evaluate an idea registry, draft a SWOT analysis, score AI proposals against feasibility and equity criteria, identify privacy and bias risks in a proposed initiative, and — as the capstone deliverable — produce a board-ready AI strategy document for your own school, district, college, or university.
The Course Description lists the full set of learning outcomes organized by Bloom's Taxonomy level.
How long does it take to complete the course?¶
The course is designed to be flexible. A focused reader can work through a chapter in two to three hours. A cohort working through the material together and completing the planning exercises in real time — running idea-generation sessions, building an actual idea registry, drafting a SWOT analysis for their own institution — will need roughly a semester or an intensive professional-development retreat spread over several sessions. The capstone deliverable, the board-ready strategy document, is designed to be a genuine institutional artifact, not a classroom exercise.
What is the capstone deliverable?¶
The capstone is a draft, board-ready AI strategy document for your own institution. It includes a SWOT analysis, a populated idea registry, an evaluated and selected project portfolio with assigned resources, a phased roadmap toward AI-tutored personalized learning, a data and governance plan covering xAPI and the Learning Record Store, and a risk register. The document is designed to be presented directly to a school board or board of trustees. Chapter 13 on Strategic Planning guides the assembly of this document.
What topics are NOT covered in this course?¶
The course does not cover how to build or fine-tune AI models (no machine-learning mathematics or model training), programming or prompt-engineering tutorials, product reviews or vendor endorsements, detailed legal or compliance advice beyond strategic risk awareness, higher-education research-administration strategy, or general IT infrastructure and cybersecurity beyond what an AI strategy directly requires. For terminology boundaries, see the Course Description.
How do I navigate the online version of this textbook?¶
The online textbook is organized with a sidebar navigation menu on the left. Use it to jump between chapters, the learning graph, glossary, and resources. The search bar at the top right searches all content. Each chapter page has a table of contents on the right side of the screen for navigating within long chapters. The Learning Graph viewer provides an interactive visual map of all 218 concepts and their relationships.
Is this course appropriate for both K-12 and higher education?¶
Yes. The strategy frameworks, the idea funnel, the risk categories, and the governance models apply equally to K-12 districts and higher education institutions, though the specific examples differ. The course includes worked SWOT analyses for a range of institutional archetypes — under-resourced inner-city schools, wealthy suburban districts, community colleges, and research universities — so readers from any context can see how the tools apply to their own situation. See the SWOT Case Studies page for these examples.
AI Capabilities and Trends¶
What is the METR study and why does it matter for school leaders?¶
The METR study — Measuring AI Ability to Complete Long Tasks (2025) — is a peer-reviewed research study that measured how long a task AI models can complete autonomously at 50% reliability. Its key finding is that this task horizon has doubled roughly every four to seven months for six straight years, growing from tasks measured in seconds in 2019 to multi-hour tasks by late 2025. For school leaders, this finding matters because a capability that doubles two to three times a year cannot be planned around with annual, linear planning cycles. The strategy tools in this course exist precisely because of this gap. Chapter 2 analyzes the METR data in detail.
What does "capability doubling every four to seven months" actually mean?¶
It means that roughly every four to seven months, AI systems become able to handle tasks that are twice as long and complex as before — autonomously, without human step-by-step guidance. In 2019, AI could reliably complete tasks that took a human a few seconds. By late 2025, AI models were reliably completing tasks that take a human several hours. If the doubling rate continues, within a few years AI will be capable of completing tasks that currently take days or weeks. For educators, this is not an abstraction: it means AI tutoring, curriculum planning, and administrative workflows will grow dramatically more capable on a predictable timeline.
Chapter 2: Measuring the AI Capability Curve explains how to read and use this data.
What is a large language model (LLM)?¶
A large language model is an AI system trained on vast amounts of text data to predict and generate human-like text. When you type a question into an AI chatbot and receive a fluent answer, you are interacting with a large language model. LLMs are the engine behind most of the AI writing, tutoring, and question-answering tools that schools are evaluating today. They are impressive at generating coherent text but have important limitations — most notably, they can produce confident-sounding answers that are factually wrong. Chapter 1 introduces LLMs and all the foundational AI vocabulary needed for the rest of the course.
What is generative AI?¶
Generative AI is the category of AI systems that can produce new content — text, images, audio, video, or code — in response to a prompt. Large language models are one form of generative AI. Image-generation systems that can create diagrams and illustrations on demand are another. For education, generative AI means that a single teacher with an AI assistant can now draft lesson plans, generate differentiated reading passages, create assessment rubrics, and produce visual explanations in a fraction of the time it previously required.
Chapter 4 explores how generative AI is reshaping the economics of educational content creation.
What is hallucination, and how serious a risk is it for education?¶
Hallucination occurs when an AI model generates text that sounds confident and authoritative but is factually incorrect or entirely fabricated. For example, an AI tutor might cite a nonexistent study, get a historical date wrong, or invent a scientific fact. For education, hallucination is a meaningful risk because students may accept AI-generated content without verifying it. The appropriate response is not to avoid AI entirely but to treat AI-generated content the way you treat any unverified source — with a habit of checking claims against authoritative references. Teaching students to verify AI output is itself an important critical-thinking skill.
Chapter 9 covers hallucination risk and other responsible-AI concerns in depth.
What is an AI agent, and how is it different from a chatbot?¶
A chatbot responds when you ask it something — you initiate each interaction. An AI agent is a more autonomous system that monitors conditions, makes decisions, takes actions, and reports results without requiring you to initiate each step. An AI agent might continuously monitor every student's learning progress, detect when a specific student is at risk of falling behind, draft an alert to the teacher, and suggest an intervention time — all without anyone asking it to. The course's planning assumption is that every educator, administrator, and student will have dozens of named AI agents working on their behalf within roughly two years.
Chapter 12 explores how to design and govern an institutional agent workforce.
How fast are AI costs falling?¶
AI costs have been falling dramatically. The cost of running AI inference — asking an AI model a question or generating content — has dropped by orders of magnitude over the past several years, and the trend continues. Tasks that cost dollars in 2022 now cost fractions of a cent. This cost collapse is one of the central strategic realities of the course: it means AI tools that were unaffordable for education budgets even two years ago are now or will soon be within reach of every school. Chapter 4 discusses declining AI cost and its implications for educational content procurement.
What is the difference between a foundation model and a frontier model?¶
A foundation model is a large, general-purpose AI model trained at massive scale on diverse data that can be adapted for many tasks. A frontier model is the most capable AI model available at any given moment — the current leading edge of what is technically possible. Frontier models are typically the first to achieve new performance benchmarks and are what the METR study tracks. For strategy purposes, the distinction matters because schools need to understand not only what AI tools can do today but what the frontier capability implies about where all AI tools will be in 18 to 36 months.
Chapter 1 defines both terms in the broader AI vocabulary framework.
What is the difference between AI capability and AI adoption?¶
Capability refers to what AI systems can technically do at their current state of development. Adoption refers to how widely and effectively those capabilities are actually being used in practice. There is typically a significant gap between the two — AI tools with impressive capabilities often sit unused or underused in education because of training gaps, policy uncertainty, budget constraints, or simple inertia. One of the key strategic insights of the course is that closing the adoption-versus-capability gap faster than peer institutions represents a competitive advantage in student outcomes and operational efficiency.
Chapter 2 discusses the adoption-versus-capability distinction and its planning implications.
Will AI replace teachers?¶
No — but it will substantially change what teachers do. The course's analysis draws on the Alpha School model and related pedagogical research to show that when AI handles the repetitive elements of direct instruction and practice, the teacher's role shifts toward mentoring, coaching, facilitating complex projects, and building the social-emotional relationships that AI cannot replicate. The course is explicit that the human relationships at the heart of education are not being replaced — they are being freed up. Chapter 8 addresses this teacher role shift in detail.
What is a reasoning model?¶
A reasoning model is a type of AI system specifically designed to tackle multi-step problems by breaking them down, working through intermediate steps, checking its own reasoning, and arriving at a conclusion. Standard large language models predict what text should come next; reasoning models are optimized to think through problems more systematically. For education, reasoning models are particularly relevant for AI tutoring on mathematics, science, and logic problems, where step-by-step reasoning rather than pattern matching is what the student needs to observe and learn. Chapter 2 introduces reasoning models in the context of the capability curve.
AI Strategy and the Idea Funnel¶
What is the idea funnel?¶
The idea funnel is the course's central decision-making framework — a six-stage process for continuously identifying, evaluating, selecting, and learning from AI initiatives. The six stages are: (1) Gather ideas from teachers, staff, students, and families through literacy sessions and open submission channels; (2) Register each idea in a structured idea registry; (3) Evaluate each idea against feasibility, risk, cost, equity, and educational benefit; (4) Select a portfolio of projects to fund; (5) Assign people, budget, and time to the selected projects; and (6) Evaluate project outcomes and feed lessons back into the funnel.
The funnel is adapted from the Generative AI Center of Excellence model. Chapter 5 and Chapter 6 walk through every stage in detail.
What is an idea registry?¶
An idea registry is a shared, searchable database where every AI idea submitted to the institution is recorded with structured metadata. A good idea registry captures: the problem the idea addresses, the proposed AI approach, the expected benefit, which stakeholders are affected, estimated cost, and identified risks. The registry is not a suggestion box — it is a living institutional asset that the review panel uses to evaluate and track ideas over time. Making the registry accessible to all staff creates transparency, reduces duplicate submissions, and signals that leadership takes AI ideas seriously.
Chapter 5 covers what to capture in an idea registry and how to structure its metadata.
How does an idea generation session work?¶
A typical idea-generation session is a one-hour AI literacy training event designed to help staff, teachers, students, or parents understand what AI can do well — and then immediately apply that understanding to generate practical, high-return ideas for their own work. The session covers basic AI capabilities in plain language (no jargon), walks through examples of AI use cases already working in similar institutions, and ends with a structured brainstorming exercise where participants submit ideas through a prepared form. The course teaches participants how to design and facilitate these sessions for their own institution.
Chapter 5 provides the complete session design and submission form templates.
What criteria are used to evaluate AI ideas?¶
Every idea in the registry is scored against five criteria: feasibility (can we actually build or deploy this given our technical and staffing constraints?), risk (what are the privacy, equity, accuracy, and dependency risks?), cost (what are the one-time and ongoing costs?), educational benefit (how much does this improve student outcomes, teacher effectiveness, or operational quality?), and equity impact (does this help or hurt students who are already underserved?). A scoring rubric standardizes how the expert review panel applies these criteria so that ideas are compared on a level playing field.
Chapter 5 covers the full evaluation framework and scoring rubric design.
What is an expert review panel?¶
An expert review panel is the group of people responsible for applying the scoring rubric to submitted ideas and recommending which ones advance to project selection. Good panels include a mix of perspectives: classroom teachers who understand instructional implications, IT or infrastructure staff who can assess technical feasibility, a finance or operations representative who understands cost, a community or parent representative who can speak to equity and family concerns, and a subject-matter expert in the relevant curriculum area when needed. Panels should operate with clear written criteria to limit the influence of personal relationships or institutional politics on funding decisions.
Chapter 5 covers panel design and the feedback loop back to idea submitters.
How does project selection work?¶
After the review panel scores all submitted ideas, the institution selects a project portfolio — the set of projects it will actually fund and resource. Good portfolio selection balances quick wins (high-feasibility, lower-cost projects that build confidence and demonstrate progress) against strategic bets (higher-investment projects with transformational potential). Equity is explicitly a selection criterion: a project that would benefit only well-resourced students or that would widen the gap for underserved students should be weighted down, while projects that actively improve equity should be prioritized.
Chapter 6 covers portfolio construction, pipeline reporting, and the build-versus-buy decision.
What is a SWOT analysis and how is it used in this course?¶
A SWOT analysis is a structured framework for assessing an institution's Strengths, Weaknesses, Opportunities, and Threats — in this context, specifically related to AI strategy. Strengths might include strong technical staff or existing device infrastructure. Weaknesses might include limited professional development budget or high teacher turnover. Opportunities include the availability of low-cost AI tutoring tools. Threats include data privacy regulation, vendor lock-in, and the risk of peer institutions gaining a competitive advantage.
The course uses SWOT analysis as the foundation for the capstone strategy document and provides worked examples across a dozen synthetic school archetypes. Chapter 13 guides the SWOT process, and the SWOT Case Studies page provides detailed examples.
What is a build-versus-buy decision?¶
A build-versus-buy decision is the analysis of whether an institution should develop an AI tool internally, purchase an existing commercial product, use an open-source solution, or some combination. For most schools, building complex AI infrastructure from scratch is not realistic. But the choice between commercial vendors and open-source alternatives involves significant trade-offs in cost, customization, data privacy, and vendor lock-in risk. The course teaches a structured framework for making this decision rather than prescribing a specific product or vendor. Chapter 6 covers the build-versus-buy framework in the context of resource assignment.
What is a Center of Excellence and does my institution need one?¶
A Center of Excellence (CoE) is a dedicated team or organizational unit responsible for coordinating, evaluating, and supporting AI initiatives across the institution. In large districts or universities, a formal CoE provides the expertise and neutral coordination needed to prevent fragmented, duplicative, or inconsistent AI adoption. In smaller institutions, a lightweight version — perhaps a standing committee that meets quarterly — can serve the same function. The idea funnel in this course is adapted from the Generative AI CoE model, and Chapter 3 explains when and how to structure one for your context.
What is the difference between a quick win and a strategic bet?¶
A quick win is an AI project with high feasibility, manageable cost, and a clear, near-term return — something the institution can deploy successfully within a semester and use to demonstrate progress to staff, students, and the board. A strategic bet is a higher-investment, longer-horizon project that could fundamentally change how the institution operates — for example, implementing an AI-driven Learning Management System or piloting the Alpha School scheduling model. A well-constructed project portfolio includes both: quick wins build institutional confidence and political support for the strategic bets.
Chapter 6 covers portfolio balancing in detail.
What goes into resource assignment for AI projects?¶
Resource assignment means matching specific people, budget, and time to each selected project. This includes identifying who will lead the project, which departments will collaborate, what the technology infrastructure requirements are, what the timeline milestones are, and how success will be measured before the project begins rather than after. Many AI initiatives fail not because the technology is wrong but because responsibility is unclear, timelines are unrealistic, or success criteria are defined too vaguely to evaluate. Chapter 6 walks through resource assignment with concrete templates.
How does an institution know if an AI project succeeded?¶
Success is defined by the success metrics and key performance indicators (KPIs) established at the beginning of the project — not at the end. Before a project is funded, the team should specify: what will we measure, how will we measure it, what threshold constitutes success, and when will we evaluate? Metrics might include student proficiency gains on a standardized assessment, teacher time saved on administrative tasks, reduction in early-warning response time, or parent satisfaction scores. Results are reported to the executive team and the board on a quarterly basis and feed back into the funnel as lessons learned.
Chapter 6 covers the full outcome-evaluation process.
What is a problem taxonomy and why does it matter?¶
A problem taxonomy is a structured classification system for the types of problems that AI ideas address — for example, categorizing ideas as addressing instructional quality, administrative efficiency, student support, family communication, or data management. A problem taxonomy keeps the idea registry organized so that patterns become visible: if 40% of submitted ideas address the same underlying problem, that signals an institutional priority that deserves dedicated attention. Similarly, a solution taxonomy classifies the type of AI approach each idea uses, making it easier to identify when multiple ideas could be served by a single platform or tool.
Chapter 5 introduces both taxonomies.
Intelligent Textbooks and Learning Technology¶
What is an intelligent textbook?¶
An intelligent textbook is a digital learning resource that goes well beyond a static PDF or an online version of a print textbook. It combines adaptive content (material that adjusts to each learner's level), interactive simulations, AI tutoring, conversational question-and-answer, and learning telemetry (detailed data about how each student engages with the content). The course uses a Five Levels of Textbooks framework to classify how sophisticated a given intelligent textbook is — from a simple static text at Level 1 to a fully autonomous AI tutor at Level 5.
Chapter 4 introduces intelligent textbooks and the Five Levels framework in detail.
What are the Five Levels of Textbooks?¶
The Five Levels of Textbooks framework classifies educational content from least to most intelligent:
- Level 1 — Static: A traditional text or PDF, no interactivity.
- Level 2 — Interactive: Rich simulations, videos, and embedded exercises — students can interact, but the content is the same for everyone.
- Level 3 — Adaptive: The textbook adjusts content difficulty, sequence, or explanations based on individual student data. Requires storing student performance data.
- Level 4 — Chatbot: A conversational AI layer enables students to ask questions and receive tailored answers within the textbook context.
- Level 5 — Autonomous AI: A fully autonomous AI tutor that monitors progress, adapts in real time, and proactively intervenes — the student is in a continuous dialogue with an AI that knows their complete learning history.
Most schools today operate at Level 1 or 2. The course helps leaders plan a credible path toward Levels 3, 4, and 5 as costs fall. See Chapter 4.
What is the "10,000 intelligent textbooks" assumption?¶
The course works from the planning assumption that within roughly two years, on the order of 10,000 intelligent textbooks — interactive, AI-tutored, simulation-rich resources — will be freely or cheaply available across subjects and grade levels, collapsing the cost of high-quality adaptive content toward zero. This is not a guarantee, but a near-term scenario grounded in the economics of AI content generation and the rate at which open educational resources are being produced. For procurement and curriculum offices, it means the current model of paying large publishers for exclusive, expensive static textbooks is under serious pressure.
Chapter 4 explores the economic and institutional implications of this assumption.
What is xAPI?¶
xAPI (short for Experience API, also called Tin Can API) is an open technical standard for recording fine-grained statements about learning activity. An xAPI statement looks like: "[Alex] [completed] [Unit 3, Lesson 2 of the Algebra textbook] [with a score of 87%] [on 2026-06-10]." Unlike older learning standards, xAPI can capture activity from any learning environment — a textbook, a simulation, a game, a video, a practice exercise — and store it in a common format. When intelligent textbooks emit xAPI data, a school can build a comprehensive picture of what every student has done and learned across all their digital learning experiences.
Chapter 7 covers xAPI and the complete learning telemetry architecture.
What is a Learning Record Store (LRS)?¶
A Learning Record Store (LRS) is the database that collects, stores, and provides access to xAPI learning records. Think of it as the educational equivalent of a fitness tracker's data cloud — it aggregates all the xAPI statements emitted by every digital learning tool a student uses and makes that data available for analysis, reporting, and AI-driven personalization. An LRS is the prerequisite for any AI-recommended learning plan: without the detailed behavioral data the LRS provides, an AI system cannot effectively personalize content for each student.
Chapter 7 explains how the LRS fits into the full learning telemetry stack.
What is an AI-driven Learning Management System?¶
A traditional Learning Management System (LMS) — such as Canvas, Schoology, or Blackboard — organizes courses, tracks assignment completion, and stores grades. An AI-driven LMS goes further: it analyzes the rich behavioral data from an LRS to continuously generate a recommended learning plan for each student, suggesting what to work on next based on demonstrated mastery, pace, learning style, and curriculum requirements. The course's planning assumption is that AI-driven LMSs capable of producing individualized learning plans will be available and widely affordable by 2028.
Chapter 7 covers the AI-driven LMS and its recommended learning plan capabilities.
What is the Alpha School model?¶
The Alpha School model is a real school operating model pioneered in Austin, Texas, that structures the school day around roughly 2–3 hours of focused, AI-tutored core academics in the morning, followed by 5–6 hours of pro-social, team-based, project-based afternoon activities — robotics, theater, athletics, clubs, community service, and entrepreneurship. The AI-tutored morning block enables students to move through core academic content at their own pace with immediate feedback, while the long afternoon restores the human, collaborative, and creative dimensions of schooling that are at risk of being crowded out by screen time. The course treats Alpha as a credible target model, not a fringe experiment.
Chapter 8 explains the model and discusses phased paths toward it.
What is a MicroSim?¶
A MicroSim is a small, focused interactive simulation embedded in a digital textbook or course page to help learners experience and explore a concept rather than just read about it. For example, a MicroSim on exponential growth might let a learner adjust a doubling time slider and watch the resulting curve change in real time. MicroSims are a hallmark of Level 2 and higher intelligent textbooks and represent a kind of educational content that generative AI now makes feasible to produce at scale. This textbook itself includes MicroSims in the MicroSims section.
Chapter 4 introduces MicroSims in the context of interactive learning content.
What is personalized learning, and what makes it different from differentiated instruction?¶
Personalized learning uses data about each individual learner — their mastery of specific concepts, their pace, their preferred explanation formats — to adapt the content, sequence, and support they receive in real time and automatically. Differentiated instruction is the practice of a teacher deliberately designing multiple versions of a lesson for different readiness levels — a high-skill, high-effort practice that scales poorly when one teacher faces thirty students. AI-driven personalized learning automates what differentiated instruction tries to achieve, making it feasible for every student in every class, not just those in classes with unusually skilled and well-resourced teachers.
Chapter 7 and Chapter 8 together explain how personalized learning works and what it requires.
What is mastery-based progression?¶
Mastery-based progression is the practice of allowing students to advance to new material only after demonstrating sufficient mastery of prerequisite concepts, rather than advancing by calendar time regardless of performance. In a mastery-based system, a student who masters the multiplication table in three weeks moves on; a student who needs six weeks gets six weeks. AI tutoring makes mastery-based progression operationally feasible at scale because the AI can provide unlimited practice, immediate feedback, and varied explanations until mastery is achieved — without requiring the teacher to individually monitor and respond to each student's progress in real time.
Chapter 8 covers mastery-based progression and related pedagogical models.
What is student data ownership?¶
Student data ownership refers to the principle that students (and their families) should have meaningful control over the data generated by their learning activities — including the right to access it, understand how it is used, export it to other systems, and request its deletion. xAPI and LRS architectures enable genuinely portable student data, but portability only benefits students if the governance policies and vendor contracts actually preserve their rights. Schools need explicit data-ownership and data-portability policies before deploying any AI-driven learning platform that stores individual student behavioral data.
Chapter 7 covers student data ownership and data interoperability requirements.
Responsible AI, Ethics, and Equity¶
What does "responsible AI" mean for schools?¶
Responsible AI is the practice of deploying and governing AI systems in ways that are safe, ethical, transparent, and accountable — that actively identify and manage risks rather than hoping they do not materialize. For schools, responsible AI means: conducting privacy reviews before deploying any student-facing tool, auditing for algorithmic bias, setting and enforcing academic integrity policies, maintaining human oversight of high-stakes decisions, and being transparent with families about what AI tools are in use and what data they collect. Responsible AI is not a barrier to innovation — it is what makes AI adoption politically and institutionally sustainable.
Chapter 9 covers the full responsible AI framework.
What is FERPA and what does it require of AI tools?¶
FERPA (the Family Educational Rights and Privacy Act) is the federal law that protects the privacy of student education records in schools receiving federal funding — which includes virtually every K-12 school and most universities. FERPA requires, among other things, that schools obtain consent before disclosing student records to third parties and that families can access and correct records. For AI tools, FERPA means that any vendor whose system stores, processes, or shares identifiable student data must operate under a FERPA-compliant data-sharing agreement with the school. Reviewing vendor contracts for FERPA compliance before signing is a baseline governance requirement, not optional.
Chapter 9 covers FERPA and COPPA compliance in the context of responsible AI deployment.
What is algorithmic bias and how does it show up in education?¶
Algorithmic bias occurs when an AI system produces systematically different outcomes for different demographic groups — often disadvantaging students of color, students with disabilities, students from low-income families, or English-language learners — as a result of patterns in the data the model was trained on or in the design of the system itself. In education, algorithmic bias can show up in automated essay scoring that disadvantages non-standard English dialects, tutoring systems calibrated to middle-class cultural references, predictive models that over-identify minority students as "at risk" based on demographic rather than academic signals, or facial-recognition systems with higher error rates for darker-skinned faces.
Chapter 9 covers algorithmic bias detection and mitigation strategies.
What is the digital divide and why does it matter for AI strategy?¶
The digital divide refers to the gap between students and families who have reliable access to devices and high-speed internet — and thus to AI-powered learning tools — and those who do not. For AI strategy, the digital divide is not a background concern but a central equity risk: well-funded schools that adopt powerful AI tutoring tools will accelerate their students' learning, while under-resourced schools that cannot afford the same tools or cannot reach students without home internet will fall further behind. A responsible AI strategy explicitly scores every proposed initiative for its equity impact and prioritizes projects that help close the digital divide rather than widen it.
Chapter 10 covers the digital divide, access to devices, broadband access, and equity impact scoring.
How should schools handle academic integrity when students have access to AI?¶
There is no single answer — but there are several necessary elements of a sustainable response. Schools need updated academic integrity policies that clearly distinguish between permitted AI assistance (brainstorming, research, draft review) and prohibited AI substitution (submitting AI-generated work as one's own). Assessment design needs to shift toward tasks that require demonstrated understanding that AI cannot fake — oral presentations, in-person problem-solving, project defenses, and performance assessments. AI detection tools exist but are imperfect and can produce false positives; they should be used as one signal among many, not as definitive proof of violation.
Chapter 9 and Chapter 10 together address academic integrity policy design.
What is skill atrophy and is it a real risk?¶
Skill atrophy is the risk that students who rely on AI to perform cognitive tasks will fail to develop or will lose the underlying skills those tasks were meant to build. For example, a student who always uses AI to structure an argument may never develop the analytical reasoning the writing process is meant to teach. This is a genuine and well-documented educational risk — the same concern that arose with calculators in mathematics education, though at a larger scale and with less clarity about which skills are worth preserving. The course does not argue for avoiding AI; it argues for being deliberate about which tasks should remain unassisted in order to preserve developmental outcomes.
Chapter 9 discusses skill atrophy alongside over-reliance as a paired risk.
What is vendor lock-in and why is it a strategic risk for schools?¶
Vendor lock-in occurs when a school becomes so dependent on a single vendor's platform that switching to a different provider becomes prohibitively expensive or operationally impossible — even if the vendor raises prices, reduces quality, or introduces practices the school objects to. In AI, lock-in risks include: proprietary data formats that make it hard to export student learning records, contracts that transfer data rights to the vendor, and integrated systems whose removal would disrupt all connected workflows. The mitigation strategies include insisting on xAPI-standard data export, avoiding exclusive long-term contracts for rapidly evolving tools, and maintaining multiple vendor relationships for critical capabilities.
Chapter 10 covers vendor lock-in as part of the institutional risk register.
Is there an ethical case for giving students access to AI tools even if risks exist?¶
Yes — and the course takes it seriously. The argument is that if AI tutoring tools can demonstrably improve student outcomes (and the evidence is growing that they can), then denying students access to those tools is itself a harm, particularly for students in under-resourced schools who cannot pay for private tutoring. The ethical question is not simply "what are the risks of AI?" but "what are the costs of not providing AI support to students who need it?" This does not mean adopting AI recklessly, but it does mean that inaction has a moral cost that responsible leaders must weigh alongside the risks of adoption.
Chapter 10 and Chapter 11 both address the ethics of AI access.
What is a risk register?¶
A risk register is a structured document that identifies, describes, and tracks each significant risk associated with an AI initiative — including its probability, potential impact, who owns the risk, and what mitigation actions are in place. The course teaches participants how to populate a risk register as part of both the idea-evaluation process and the capstone strategy document. A risk register is not a list of fears — it is a management tool that converts vague anxiety about AI risks into specific, assignable, trackable actions.
Chapter 10 covers risk register construction and the risk-reward trade-off framework.
What about student well-being and screen time?¶
The course takes student well-being seriously as a design constraint, not an afterthought. Concerns about excessive screen time, social isolation, reduced physical activity, and the replacement of human relationships with AI interactions are legitimate risks that responsible AI deployment must address. The Alpha School model is, in part, a structural answer to this concern: by concentrating AI-tutored instruction in a focused 2–3 hour morning block, it explicitly protects large portions of the school day for physical activity, team collaboration, and human mentorship. Any AI strategy should include clear guidelines on screen time limits and monitoring for signs of AI over-dependence in student behavior.
Chapter 8 and Chapter 10 both address student well-being design considerations.
Governance, Policy, and Implementation¶
What is AI governance and who should own it?¶
AI governance is the system of authorities, processes, policies, and accountabilities that determines how an institution makes decisions about AI — what to adopt, who decides, what rules apply, how outcomes are evaluated, and who is responsible when something goes wrong. The question of who owns AI governance depends on institutional size and structure. Large districts and universities typically need a centralized governance body with representation from instruction, technology, legal or compliance, finance, and the community. Smaller institutions may delegate more decisions to building principals or department chairs while maintaining central oversight of data privacy and equity.
Chapter 11 covers both centralized and decentralized governance models and the trade-offs of each.
Should AI governance be centralized or decentralized?¶
Both approaches have merit, and most institutions end up with a hybrid model. Centralizing decisions about data privacy, vendor contracting, security, and equity standards makes sense because these decisions have institution-wide consequences and require specialized expertise. Decentralizing decisions about which AI tools individual teachers use in their classrooms, or how a specific department experiments with AI-assisted feedback, allows responsiveness and innovation. The risk of over-centralization is bureaucratic slowness that frustrates early adopters. The risk of over-decentralization is inconsistent data practices, vendor proliferation, and equity gaps between well- and poorly-resourced departments.
Chapter 11 provides a framework for deciding which decisions to centralize versus delegate.
How should schools approach professional development for AI?¶
Effective professional development for AI in education should do three things: build sufficient AI literacy for staff to understand the tools they are asked to use, build pedagogical confidence to redesign instruction around AI-assisted workflows, and build the critical-thinking skills to evaluate AI outputs rather than accept them uncritically. One-hour idea-generation sessions (covered in Chapter 5) are one professional-development format. Sustained multi-session cohorts, peer coaching, and AI pilot programs with structured reflection cycles are more effective for deep adoption. Professional development should be ongoing — not a single workshop — because AI capabilities are evolving faster than any single training event can address.
Chapter 11 covers AI literacy programs and professional development structures.
How do schools engage school board members about AI strategy?¶
School board engagement around AI should begin before any major AI initiative is launched, not after. Board members need enough AI literacy to understand what they are approving and to ask informed oversight questions — but they do not need to become technical experts. The course recommends a brief AI literacy orientation for the board as the first step, followed by a regular reporting cadence on AI project outcomes (the quarterly executive/board report from the idea funnel's evaluation stage). The capstone strategy document is explicitly designed to be board-ready — structured to inform board deliberation and formal approval.
Chapter 11 and Chapter 13 both address school board engagement and the board-ready strategy format.
What is a pilot program and how should schools run one?¶
A pilot program is a small-scale, time-limited deployment of an AI tool or model with explicit evaluation criteria, designed to test assumptions before committing to full-scale adoption. Good pilot programs: define success metrics in advance, select a representative sample of students and teachers (not just enthusiastic early adopters), include a comparison group if possible, run long enough to observe meaningful outcomes (typically one semester), and include structured data collection and reflection. Pilot results feed directly into the idea funnel's evaluation stage and inform the institution's scaling strategy.
Chapter 11 covers pilot design and scaling strategy.
What does an AI implementation roadmap look like?¶
An implementation roadmap is a phased plan that sequences AI initiatives over time, aligning each phase with the institution's current capacity, budget cycle, and governance readiness. A typical roadmap includes: a Phase 1 of foundational infrastructure and professional development (building AI literacy, establishing governance, deploying basic tools); a Phase 2 of targeted pilots with evaluation and iteration; and a Phase 3 of scaling proven initiatives and beginning the more complex work of xAPI/LRS integration and AI-driven personalization. The roadmap is a living document that is updated as pilot results and capability changes arrive.
Chapter 11 and Chapter 13 cover roadmap construction together.
How should schools communicate with parents about AI?¶
Parent engagement around AI should be transparent, proactive, and ongoing. Parents have legitimate questions about what data their children's AI tools collect, how it is used, whether it is shared with third parties, and what opt-out options exist. Schools should publish plain-language summaries of their AI tools, data practices, and privacy protections — not just legal policy documents — and create accessible channels for families to ask questions and raise concerns. Involving parents in idea-generation sessions is also valuable: families often surface practical ideas and concerns that staff do not see. Some families will be skeptical; the appropriate response is honest engagement, not dismissal.
Chapter 11 covers parent and community engagement as part of the governance framework.
What is the agentic AI workforce and why does it require new governance?¶
The agent workforce is the growing set of autonomous AI agents operating on behalf of individual administrators, teachers, and students — monitoring data, drafting communications, recommending actions, and completing tasks without being manually prompted for each step. Unlike a chatbot that does what you ask and stops, agents act continuously. This creates governance challenges that do not apply to passive AI tools: Who authorized an agent to send a message to a parent? What happens when two agents give contradictory advice? Who is accountable when an agent takes a wrong action? Agent governance — the policies and oversight structures for AI agents specifically — must be part of any comprehensive AI governance framework.
Chapter 12 covers agent governance design in detail.
More questions? Use the search bar at the top of the page, explore the Glossary, or consult the relevant chapter directly. The Learning Graph shows how all concepts in the course connect.