Glossary of Terms¶
Academic Integrity¶
The commitment to honesty, trust, fairness, and respect in all aspects of academic work, including the authentic representation of one's own intellectual contributions in assignments, assessments, and credentials.
AI dramatically complicates academic integrity by making it easy for students to generate high-quality work they did not produce. Policies, pedagogy, and assessment design must all evolve to maintain meaningful standards of honest scholarship.
Example: A school updates its academic integrity policy to specify which AI tools students may use for different assignment types, and redesigns major assessments to require demonstrated understanding that cannot be faked with AI assistance.
Academic Integrity Policy¶
A formal statement of an institution's expectations regarding honesty in student academic work, including definitions of prohibited behaviors such as plagiarism and unauthorized AI use, and consequences for violations.
Traditional academic integrity policies did not anticipate generative AI and must be substantially revised. Updated policies should define the boundary between permitted AI assistance and prohibited AI substitution across different assignment types.
Example: An updated academic integrity policy specifies that students may use AI brainstorming tools for essay prewriting but must compose all submitted drafts themselves, and must disclose any AI assistance used.
Access To Devices¶
The availability of functional computing devices—such as laptops, tablets, or smartphones—to students for use in school and at home, a prerequisite for participating in AI-enhanced learning experiences.
Unequal device access is one of the most concrete barriers to equitable AI deployment. One-to-one device programs and device lending libraries are essential infrastructure investments that must precede broad AI tool adoption.
Example: Before launching an AI tutoring pilot, a district audits device availability and discovers that 15% of students in its highest-poverty schools share devices with siblings at home, leading to a targeted device grant application.
Adaptive Content¶
Educational material that automatically adjusts its difficulty, format, vocabulary, depth, or sequencing in response to individual learner performance data, providing each student with appropriately challenging and supportive experiences.
Adaptive content addresses one of the most persistent challenges in education: that a single fixed curriculum inevitably under-serves students at both ends of the readiness spectrum. AI makes truly adaptive content feasible at scale.
Example: An adaptive reading platform lowers the Lexile level of assigned passages for a struggling reader while increasing complexity for a proficient peer, all within the same classroom session.
Adoption Versus Capability¶
The observable gap between the current technical capability of AI systems and the actual level of adoption and effective use of those systems within organizations, including schools.
This gap represents both a risk and an opportunity: districts that close the gap faster than peers gain competitive advantages in student outcomes and operational efficiency. Leaders should assess whether their adoption pace matches the capability curve.
Example: AI reading comprehension tools have been technically capable of providing individualized feedback for two years, yet only 5% of teachers in a given state regularly use them, illustrating a wide adoption gap.
Agent Governance¶
The policies, oversight structures, and accountability mechanisms specifically designed to ensure that AI agents operating with autonomy do so within defined boundaries, consistent with organizational values and legal requirements.
Agent governance is a new and evolving frontier in AI strategy. As agents take more autonomous actions on behalf of schools and students, the accountability questions of who is responsible for agent behavior become critically important.
Example: A district's agent governance policy requires that any AI agent action affecting a student's educational record—such as updating attendance or flagging for intervention—be logged, reviewable by humans, and overridable by the responsible educator.
Agent Orchestration¶
The supervisory layer or system that coordinates the activities of multiple AI agents—assigning tasks, resolving conflicts, monitoring performance, and ensuring that collective agent behavior aligns with organizational goals and policies.
Agent orchestration is the governance infrastructure for AI agent workforces. Without it, individual agents may optimize locally in ways that produce poor collective outcomes or violate institutional policies.
Example: A district's agent orchestration layer ensures that no student receives more than three automated AI agent communications per day across all channels, preventing an overwhelming flood of messages from well-intentioned but uncoordinated agents.
Agent Task Assignment¶
The process of allocating specific work items or goals to individual AI agents within a multi-agent system, ensuring that tasks are matched to agents with appropriate capabilities and that work is distributed without conflict or duplication.
Clear agent task assignment prevents AI agents from duplicating effort, taking conflicting actions, or leaving important tasks unhandled. In education, it requires explicit definition of which agents are responsible for which student-facing interactions.
Example: An agent task assignment framework specifies that the progress monitoring agent handles weekly academic check-ins, while the parent communication agent handles notification of grade changes, preventing both agents from sending conflicting messages.
Agent Workforce¶
The collection of AI agents deployed within an organization to perform defined tasks autonomously or semi-autonomously, collectively extending organizational capacity in ways analogous to a team of human workers.
An agent workforce in education could handle routine administrative tasks, freeing human staff for relationship-intensive and judgment-intensive work. Governing an agent workforce requires new policies addressing accountability, quality control, and oversight.
Example: A district's agent workforce includes a scheduling agent, an attendance monitoring agent, a parent communication agent, and a progress reporting agent, collectively saving administrative staff an estimated 20 hours per week.
Agentic Task Completion¶
The ability of an AI agent to independently carry out a defined goal through a planned sequence of actions—potentially including tool use, information gathering, and decision making—without step-by-step human direction.
Agentic task completion is what enables AI to go beyond answering questions and begin managing workflows. In education, this could mean an AI agent independently preparing progress reports or scheduling student interventions.
Example: An AI agent completes an agentic task by reviewing a student's portfolio, identifying learning gaps, retrieving relevant practice resources, and sending a personalized study plan to the student.
AI Access Inequality¶
The disparity in students' ability to benefit from AI-enhanced learning due to differences in device access, internet connectivity, AI literacy, teacher preparedness, and institutional investment across schools and communities.
AI access inequality risks creating a new dimension of educational stratification where students in well-resourced schools gain AI-enhanced advantages while peers in under-resourced schools fall further behind. Proactive policy intervention is required.
Example: A state study finds that AI tutoring tool adoption is four times higher in schools in the top income quartile than in schools in the bottom quartile, prompting a targeted adoption incentive program.
AI Agent¶
An AI system designed to pursue goals by taking a sequence of actions—such as browsing the web, writing code, or sending messages—with limited or no human intervention between steps.
AI agents are moving from research into practical educational tools, with early applications in personalized scheduling, tutoring follow-up, and administrative automation. Governance frameworks need to address agent autonomy and accountability.
Example: An AI agent automatically identifies a student's missed assignments, drafts a summary email to the student's advisor, and schedules a check-in meeting—all without manual input.
AI Agent Persona¶
The designed identity, communication style, name, and behavioral characteristics given to an AI agent to make interactions feel natural, appropriate, and aligned with the institution's values and the needs of the target user population.
Thoughtful AI agent persona design shapes how students and teachers relate to AI tools, affecting trust, engagement, and appropriate use. Personas should be honest about the AI's non-human nature and avoid manipulative relationship dynamics.
Example: A district designs its AI tutoring agent with an encouraging, patient persona named "Spark," calibrated to be supportive without being sycophantic, and clearly identified as an AI assistant rather than a human tutor.
AI Benchmark¶
A standardized test or set of tasks used to measure and compare the performance of different AI systems, providing a common reference point for evaluating capability progress over time.
Benchmarks help education leaders interpret vendor capability claims and understand how rapidly AI is improving. However, benchmarks can be gamed, so leaders should look for real-world performance evidence as well.
Example: A vendor reports that their AI tutoring model scores 85% on a math reasoning benchmark, which a district technology director uses to compare it against competing products.
AI Content Generation¶
The use of generative AI systems to automatically produce educational materials—such as lesson plans, reading passages, quiz questions, and instructional videos—based on specified parameters including subject, grade level, and standard.
AI content generation dramatically reduces the time and cost of producing educational materials but introduces quality control challenges. Districts must establish review protocols to ensure accuracy, cultural appropriateness, and pedagogical soundness.
Example: A district uses AI content generation to create 200 differentiated reading passages for a new social studies unit in three days, a task that would have taken a traditional writing team three months.
AI Detection Tools¶
Software systems that analyze submitted student work for statistical patterns associated with AI-generated text, used to identify potential violations of academic integrity policies.
AI detection tools are imperfect and produce both false positives and false negatives, creating risks of unjust accusations. Leaders should treat them as one signal among many rather than as definitive evidence, and prioritize assessment redesign over detection.
Example: A district cautions teachers that AI detection tool results alone are insufficient to accuse a student of academic dishonesty, and establishes an appeal process that requires additional evidence before any disciplinary action.
AI Driven LMS¶
A learning management system that uses artificial intelligence to automate administrative functions, personalize content delivery, surface actionable insights for teachers, and adapt learning sequences based on individual student performance data.
AI-driven learning management systems go beyond content delivery to actively guide instruction and intervention. They represent the convergence of data infrastructure, AI capability, and pedagogical design into a unified platform.
Example: An AI-driven learning management system automatically groups students by demonstrated concept mastery each week and suggests targeted activities for each group, which the teacher reviews and approves each Monday morning.
AI Ethics¶
The branch of applied ethics concerned with identifying and addressing the moral implications of artificial intelligence systems, including questions of fairness, accountability, transparency, privacy, autonomy, and human dignity.
AI ethics provides the conceptual framework for responsible AI governance in education. Leaders do not need to be philosophers, but they do need to be conversant with the core ethical questions so they can make defensible decisions.
Example: An AI ethics review flags concerns that an automated student behavior scoring system may disadvantage students with disabilities whose behavioral profiles differ systematically from the training data.
AI Governance¶
The policies, processes, roles, and accountability structures that an organization establishes to guide decisions about AI adoption, use, monitoring, and oversight in ways that are responsible, equitable, and aligned with institutional values.
AI governance is the organizational infrastructure that turns good AI intentions into consistent practice. Without formal governance, AI adoption proceeds inconsistently, creating legal, ethical, and equity risks that accumulate over time.
Example: A district's AI governance structure includes a steering committee that approves all new AI tool adoptions, a data privacy review process, and quarterly reporting to the school board on AI portfolio status and outcomes.
AI Hallucination Risk¶
The danger that AI-generated content will contain plausible-sounding but factually incorrect information, potentially misleading students or educators who trust AI outputs without independent verification.
AI hallucination risk is especially serious in educational settings where accurate information is foundational. Leaders must establish systematic fact-checking processes for AI-generated content and teach students to be critical consumers of AI output.
Example: An AI hallucination risk incident occurs when an AI-generated history lesson incorrectly attributes a famous quote to a historical figure who never said it, and no review process catches the error before the lesson is taught.
AI Literacy Program¶
A structured educational initiative designed to build foundational understanding of artificial intelligence among a defined audience—such as all district teachers, students, or parents—covering AI concepts, capabilities, limitations, and responsible use.
An AI literacy program is necessary infrastructure for sustainable AI adoption. Staff and students who lack AI literacy will misuse tools, miss opportunities, and be unable to participate meaningfully in AI governance decisions.
Example: A district AI literacy program delivers a required four-hour training for all teachers, an age-differentiated student curriculum for grades 3 through 12, and an optional parent workshop series offered at each school.
AI Literacy Training¶
Structured learning experiences designed to build participants' foundational understanding of how AI works, what it can and cannot do, its risks and benefits, and how to use AI tools effectively and responsibly.
AI literacy training is essential for all staff before deploying AI tools, not just for technology specialists. Without it, teachers and administrators may misuse AI, misinterpret its outputs, or fail to detect errors.
Example: A district requires all teachers to complete a four-hour AI literacy training module before accessing the new AI lesson-planning assistant, ensuring they understand its limitations and how to verify its outputs.
AI Safety¶
The interdisciplinary field and set of practices concerned with ensuring that AI systems behave in accordance with intended goals and do not produce harmful, unintended, or catastrophic outcomes, especially as AI systems become more autonomous.
AI safety concerns in K-12 settings include preventing inappropriate content generation, protecting student psychological well-being, and ensuring that autonomous AI agents do not take harmful actions on behalf of students or teachers.
Example: A district's AI safety policy requires that all student-facing AI tools be blocked from generating violent, sexual, or extremist content and tested against adversarial prompts before classroom deployment.
AI Strategy¶
A documented, organization-wide plan that defines how an institution will evaluate, adopt, govern, and sustain the use of artificial intelligence to advance its mission and serve its stakeholders.
A formal AI strategy prevents ad hoc tool adoption, reduces risk, and aligns resources with priorities. Without a strategy, schools often accumulate disconnected AI tools that lack coherent governance or measurable outcomes.
Example: A district's AI strategy specifies which use cases to pilot in the first year, how student data will be protected, who approves new AI tools, and how success will be measured.
AI Tutoring¶
The use of AI systems to deliver personalized instructional support to individual learners, including explanations, practice questions, feedback, and scaffolding, outside of or in addition to direct teacher instruction.
AI tutoring can extend high-quality instructional support beyond classroom hours, providing students with on-demand help regardless of access to private tutors. Evidence suggests it can significantly accelerate learning when well-implemented.
Example: A middle schooler uses an AI tutoring app after school to work through algebra problems, receiving immediate, step-by-step explanations for each mistake until she masters the concept.
AI Use Policy¶
A formal organizational document that defines acceptable and prohibited uses of artificial intelligence tools by staff and students, specifying conditions, limitations, and accountability requirements for AI interactions.
An AI use policy creates shared expectations and reduces the risk of harmful or inconsistent AI use. It should be reviewed and updated at least annually given the rapid pace of AI capability and tool development.
Example: A district AI use policy specifies that teachers may use AI to generate draft lesson materials but must review all content for accuracy before sharing with students, and may not use AI to make final grading decisions.
Algorithmic Bias¶
The tendency of an AI system to produce systematically unfair or inaccurate outputs for certain groups of people due to biases present in training data, model design, or evaluation criteria.
Algorithmic bias in educational AI can perpetuate or amplify existing inequities in ways that are less visible than human bias. Leaders must require bias audits as a standard part of AI procurement and ongoing monitoring.
Example: A college readiness prediction model trained primarily on data from affluent districts systematically underestimates the potential of students from low-income schools, leading to lower-quality course recommendations for those students.
Alpha School Model¶
An educational model in which AI-assisted instruction handles the delivery of core academic content efficiently, freeing significant classroom time for life skills, project-based learning, mentorship, and character development.
The Alpha School model represents one vision of how AI could restructure the school day, shifting the teacher's role from content deliverer to coach and mentor. It raises important questions about what human educators uniquely provide.
Example: In an Alpha School-inspired model, students complete two hours of AI-tutored core academics each morning, then spend the rest of the day on collaborative projects, community service, and skill workshops.
Artificial Intelligence¶
A branch of computer science focused on building systems that perform tasks requiring human-like reasoning, such as understanding language, recognizing patterns, or making decisions, without being explicitly programmed for each specific scenario.
Understanding AI's capabilities and limits helps education leaders make informed decisions about adoption, investment, and policy. It provides a common vocabulary for conversations across all stakeholder groups.
Example: A school district uses an AI system to flag students at risk of failing based on attendance and grade patterns, allowing counselors to intervene early.
Authentic Assessment¶
A form of evaluation in which students demonstrate knowledge and skills by applying them to meaningful, real-world tasks or problems, rather than through standardized tests or exercises with artificial constraints.
Authentic assessment becomes more important as AI makes it easy for students to generate polished text and calculations on demand. Tasks that require original thinking, embodied skill, or collaborative problem-solving are harder to complete with AI assistance.
Example: Instead of a written test on nutrition, students design and present a week-long meal plan for a local senior center, defending their nutritional choices to a panel that includes a registered dietitian.
Benchmark Saturation¶
The condition in which AI systems score near the maximum possible on a given benchmark, rendering that benchmark no longer useful for distinguishing differences in capability among leading models.
When benchmarks saturate, new and harder tests must be created, which can make year-over-year comparisons difficult. Education leaders should be aware that a "perfect score" on an old benchmark may not reflect the full picture.
Example: Several AI models now score above 90% on a well-known reading comprehension benchmark, so researchers have developed harder benchmarks to continue measuring meaningful differences.
Benefit Scoring¶
A quantitative or qualitative rating of the expected positive outcomes of an AI use case, considering factors such as student learning improvement, teacher time saved, equity gains, and operational efficiency.
Benefit scoring provides a structured basis for comparing diverse AI proposals on a common scale. It forces proposers to articulate and defend their expected value claims before resources are committed.
Example: An AI idea that would reduce special education paperwork time by 60% receives a high benefit score based on staff survey data documenting the current administrative burden.
Blended Learning¶
An instructional model that combines face-to-face teacher-led instruction with digital and online learning activities in a structured way, allowing students to have some control over the time, place, pace, and path of their learning.
Blended learning creates natural integration points for AI tools within existing school structures, making it one of the most practical near-term models for AI-enhanced education. It preserves human connection while leveraging AI's personalization capabilities.
Example: In a blended learning English class, students work independently with an AI writing coach three days per week and participate in whole-class Socratic seminars and one-on-one teacher conferences twice per week.
Board Ready Strategy¶
An AI strategy presentation and supporting documentation that is formatted, scoped, and written to enable school board members without technical backgrounds to understand, deliberate on, and vote to approve an AI strategic plan.
A board-ready strategy bridges the gap between technical AI complexity and the policy-making responsibilities of elected officials. Leaders who cannot communicate AI strategy in accessible, non-technical terms will struggle to secure the board authorization needed for investment.
Example: The superintendent converts the district's detailed AI strategy into a board-ready presentation using plain language, clear visuals, a budget summary, and a set of decision questions requiring board action.
Broadband Access¶
The availability of high-speed internet connectivity sufficient to support cloud-based AI applications, video streaming, and real-time data exchange, enabling full participation in digital and AI-enhanced educational experiences.
Unreliable or absent broadband access is a structural barrier to AI-enabled education, particularly in rural and low-income communities. Leaders must advocate for infrastructure investment as a prerequisite for equitable AI adoption.
Example: A rural district delays deploying cloud-based AI tutoring tools until a federal broadband expansion grant brings adequate connectivity to all 12 of its school buildings.
Build Versus Buy¶
A strategic decision about whether to develop an AI capability internally using the organization's own staff and resources, or to procure it from an external vendor, considering cost, control, and capability tradeoffs.
Most districts lack the technical staff to build sophisticated AI tools and should default to buying from reputable vendors. However, for unique use cases involving sensitive data, building or customizing may be preferable.
Example: A district deciding whether to build a custom AI attendance-intervention tool or purchase an existing vendor solution conducts a build-versus-buy analysis comparing long-term costs, data privacy implications, and maintenance requirements.
Capability Doubling¶
The phenomenon in which the measurable performance of AI systems on a specific task approximately doubles within a defined time period, reflecting the rapid pace of AI advancement.
Capability doubling means that AI tools purchased today may be significantly less capable than those available in one or two years. Strategic planning must account for this pace to avoid locking in outdated solutions.
Example: If an AI system's ability to complete autonomous tutoring tasks doubles every year, a district planning a five-year implementation must budget for substantial capability increases during that period.
Capability Forecasting¶
The systematic process of predicting future AI performance levels based on historical benchmark data, research trends, and computational scaling patterns, used to inform long-term strategic planning.
Capability forecasting introduces structured foresight into AI strategy, replacing speculation with evidence-based scenarios. Districts that forecast capability trajectories can make more informed decisions about when to pilot, scale, or pause AI initiatives.
Example: Using capability forecasting, a district projects that AI-generated personalized curriculum will reach acceptable quality thresholds within 18 months, informing a decision to begin a vendor evaluation now.
Capability Trajectory¶
The projected path of AI performance improvement over time, derived from observed benchmark trends and research progress, used to anticipate what AI systems will be able to do in future planning horizons.
Understanding capability trajectory helps education leaders avoid both premature investment and dangerous delay. A well-informed trajectory analysis should inform every multi-year technology and curriculum plan.
Example: A district technology director presents a capability trajectory chart showing that AI tutoring quality is expected to reach expert-teacher equivalence within three years, informing the board's hiring and budget decisions.
Capstone AI Strategy¶
A comprehensive, integrated strategic document that synthesizes an organization's AI readiness assessment, vision, use case portfolio, governance framework, implementation roadmap, and success metrics into a single authoritative plan.
A capstone AI strategy signals organizational seriousness and provides the foundational document against which all AI decisions and investments are evaluated. It also serves as a communication tool for staff, families, and community stakeholders.
Example: A district's capstone AI strategy, approved by the school board, defines the district's AI vision for 2026-2029, lists 15 prioritized use cases, establishes a governance structure, and commits to annual public reporting on outcomes.
Center Of Excellence¶
An internal organizational unit or cross-functional team designated to develop shared expertise, standards, tools, and practices for a specific domain—such as AI—that other departments can draw upon.
Establishing an AI center of excellence prevents duplicated effort, promotes consistent governance, and builds institutional knowledge faster than isolated departmental experiments. It is a key structure for scaling AI responsibly.
Example: A large urban district creates an AI center of excellence staffed by curriculum specialists, a data privacy officer, and a technology lead, who evaluate and vet all proposed AI tools before classroom deployment.
Centralized Governance¶
An AI governance model in which decision-making authority, standards, and oversight are held by a central organizational body—such as a district-level committee or office—rather than being delegated to individual schools or departments.
Centralized governance improves consistency, compliance, and equity in AI adoption but may reduce responsiveness to school-specific needs. It is most effective when paired with clear escalation paths and defined school-level autonomy within set boundaries.
Example: In a centralized governance model, no school may adopt a new AI tool without district-level data privacy review and curriculum alignment approval, ensuring consistent standards across all 25 schools.
Change Management¶
A structured approach to transitioning individuals, teams, and organizations from a current state to a desired future state, including stakeholder communication, training, resistance management, and cultural alignment activities.
AI adoption in education fails more often from change management failures than from technical problems. Leaders who invest in communication, professional development, and stakeholder engagement dramatically increase AI implementation success rates.
Example: A district's AI tutoring rollout includes a six-month change management plan featuring teacher co-design sessions, principal coaching, a peer ambassador program, and a parent information campaign before go-live.
Community Engagement¶
The process of involving the broader school community—including local businesses, community organizations, faith communities, and civic leaders—in understanding, shaping, and supporting the district's AI strategy and implementation.
Broad community engagement builds the social license for AI in schools and surfaces diverse perspectives that improve strategy quality. It also creates opportunities for partnerships and resources that strengthen implementation capacity.
Example: A district convenes a community AI advisory panel including a pediatrician, a local employer, a civil rights attorney, and a student representative to provide ongoing input on the district's AI governance framework.
Competency Based Education¶
An educational approach that focuses on demonstrating defined skills and knowledge to a specified performance level, rather than completing a set number of instructional hours or courses, enabling flexible pacing and multiple demonstration pathways.
Competency-based education aligns naturally with AI-enabled mastery tracking and personalized learning paths. It shifts the measure of educational success from time and compliance to demonstrated capability.
Example: In a competency-based education program, a student who already demonstrates proficiency in essay writing skips the foundational writing unit and advances directly to argumentative research writing.
Competitive Advantage¶
The meaningful edge an organization gains over peers by developing capabilities, processes, or practices that enable it to deliver superior outcomes, attract resources, or fulfill its mission more effectively.
In education, competitive advantage from AI may manifest as better student outcomes, greater teacher retention, or stronger community trust. Districts that develop AI capabilities early may establish hard-to-replicate advantages.
Example: A district that deploys effective AI-assisted personalized learning two years before neighboring districts may attract families and retain high-performing teachers who value the innovative environment.
Concept Learning Graph¶
A structured map of the relationships and dependencies among the concepts in a subject area or curriculum, used to sequence learning, identify prerequisite knowledge, and guide personalized learning path recommendations.
Concept learning graphs are the foundational data structure enabling intelligent textbooks and AI tutors to navigate curriculum intelligently. They make explicit the connections between ideas that experienced teachers hold implicitly.
Example: A concept learning graph for algebra shows that students must master linear equations before tackling systems of equations, enabling an AI tutor to automatically check prerequisite knowledge before advancing a student.
Content Cost Collapse¶
The dramatic reduction in the financial cost of producing high-quality educational content driven by generative AI, which can create texts, images, assessments, and multimedia materials in a fraction of the time and cost of traditional methods.
Content cost collapse disrupts the traditional educational publishing industry and changes the economics of curriculum development for districts. Organizations that understand this shift can redirect resources from content purchase to content curation and quality assurance.
Example: A curriculum team that previously spent $200,000 hiring writers to develop a new science unit now spends $15,000 using AI-assisted development, with the remainder redirected to teacher training.
Content Democratization¶
The process by which advances in AI reduce the cost and skill barriers to creating high-quality educational content, making professional-grade materials accessible to schools and educators regardless of budget or expertise.
Content democratization has the potential to eliminate the historic advantage of wealthy districts in accessing premium curriculum materials. It also raises quality assurance challenges, since anyone can now produce content that appears professional.
Example: A rural district with no curriculum development staff uses AI tools to create a locally relevant social studies unit aligned to state standards—work that previously would have required a professional curriculum writer.
Content Quality Assessment¶
The process of evaluating AI-generated or AI-assisted educational materials against defined standards for accuracy, grade-level appropriateness, cultural sensitivity, pedagogical effectiveness, and alignment to learning objectives.
As AI lowers the cost of content creation, content quality assessment becomes the critical bottleneck. Districts need structured review processes and trained reviewers to prevent low-quality AI content from reaching students.
Example: Before deploying AI-generated history reading passages, the district's content quality assessment team checks each passage for factual accuracy, reading level, and alignment with state standards.
Context Window¶
The maximum amount of text or information an AI model can consider at one time when generating a response, encompassing both the input provided and the output being produced.
Context window size determines how much of a student's history, a document, or a conversation an AI can "hold in mind" at once. Larger windows enable richer, more coherent tutoring interactions over longer sessions.
Example: A context window of 100,000 tokens means an AI tutor can "remember" an entire semester's worth of a student's written work when giving feedback on a final essay.
Conversational AI¶
An AI system designed to interact with users through natural, back-and-forth dialogue—mimicking the flow of human conversation—to answer questions, provide support, or guide users through tasks.
Conversational AI enables more natural and engaging interactions between students and technology than traditional menu-driven software. Its effectiveness depends on the quality of the underlying model and the design of the conversational experience.
Example: A school's conversational AI chatbot helps students navigate course registration by answering their questions in plain English, rather than directing them to a complex web portal.
COPPA Compliance¶
Adherence to the Children's Online Privacy Protection Act, a U.S. federal law that restricts the online collection of personal information from children under 13, requiring parental consent and strict data minimization practices.
COPPA compliance is mandatory for any AI tool used with elementary and middle school students. Violations can result in significant federal penalties, and leadership cannot rely solely on vendors to self-certify compliance.
Example: Before deploying an AI reading app to third-grade students, the district verifies that the vendor has a COPPA-compliant privacy policy, does not collect persistent identifiers from children, and does not share data with advertisers.
Cost Estimation¶
A projection of the financial resources required to implement, operate, and sustain an AI initiative, including software licensing, hardware, staff time, training, and ongoing maintenance expenses.
Accurate cost estimation prevents budget shortfalls that derail AI pilots mid-implementation. It also enables fair comparison among competing project proposals and realistic total-cost-of-ownership calculations.
Example: A cost estimation for an AI tutoring pilot includes the vendor license fee, one day of teacher training per school, IT integration time, and an ongoing per-student subscription cost.
Cost Per Task¶
The total financial cost incurred each time an AI system completes a defined unit of work, encompassing compute, licensing, and operational expenses, used to evaluate the economic viability of AI deployment at scale.
Rapidly declining cost per task is one of the most significant economic forces in educational AI. Tasks that were too expensive to automate at scale two years ago may now be economically feasible for every district.
Example: If an AI system's cost to generate a personalized practice problem set drops from $0.50 to $0.05 per student per day, it transforms from a premium tool into a district-wide standard.
Critical Thinking Agent¶
An AI agent specifically designed to challenge students' reasoning by posing questions, presenting counterarguments, and asking for evidence, stimulating deeper intellectual engagement rather than simply providing answers.
Critical thinking agents represent a pedagogically sophisticated application of AI, using the technology to build students' reasoning capacity rather than substitute for it. They embody the principle that AI should augment thinking, not replace it.
Example: When a student submits a thesis statement, the critical thinking agent responds with "What evidence supports this claim?" and "What might someone who disagrees with you argue?" to push the student toward more rigorous reasoning.
Cross Team Collaboration¶
The practice of structured cooperation among staff from different departments, schools, or disciplines to design, implement, and evaluate AI initiatives, leveraging diverse expertise and preventing siloed decision-making.
AI projects that cross departmental boundaries—such as tools that span curriculum, technology, and student services—require deliberate cross-team collaboration structures or they will fail at the seams between departments.
Example: A cross-team collaboration between the special education department and the technology office produces an AI accommodation-recommendation tool that neither team could have designed effectively alone.
Curriculum Alignment¶
The process of ensuring that instructional materials, assessments, and learning activities are coherently mapped to established educational standards, learning objectives, and scope-and-sequence frameworks.
AI tools can accelerate curriculum alignment by automatically tagging content to standards and identifying gaps, but human expert review remains essential to ensure that alignment is substantive rather than superficial.
Example: An AI tool scans a newly developed unit and flags three state standards that are mentioned in the materials but never actually assessed, allowing the curriculum team to address gaps before publication.
Data Interoperability¶
The technical capability of different software systems to exchange, interpret, and use each other's data without manual conversion or custom integration work, enabling seamless data flow across an educational technology ecosystem.
Poor data interoperability forces schools to choose between isolated best-in-class tools and integrated mediocre platforms. Prioritizing interoperability in vendor selection prevents data silos that undermine analytics and personalization.
Example: A district requires all new AI tool vendors to support xAPI output, ensuring that every tool's student performance data can be imported into the central learning record store without custom programming.
Data Portability¶
The right and technical ability of students or institutions to export their learning data from one platform and import it into another, preventing vendor lock-in and ensuring that data collected on behalf of students remains accessible.
Data portability protects districts from being held hostage by vendors who accumulate valuable student learning histories. It should be a contractual requirement in every educational technology procurement agreement.
Example: A district contract requires that upon termination, the AI tutoring vendor provide all student learning records in a standard xAPI format within 30 days, enabling migration to a new platform.
Data Privacy¶
The right of individuals to control how information about them is collected, used, shared, and retained, and the corresponding obligation of organizations to handle personal data with appropriate security and transparency.
Data privacy in education is both a legal obligation and an ethical imperative, particularly when students are minors. Every AI tool adoption decision must include a rigorous privacy impact assessment.
Example: Before deploying an AI tutoring platform, the district's data privacy officer reviews the vendor's data use agreement to confirm that student data will not be used for advertising, model training, or sale to third parties.
Decentralized Governance¶
An AI governance model in which individual schools, departments, or teachers have significant autonomy to make decisions about AI adoption and use within broad institutional guidelines, enabling faster experimentation and local adaptation.
Decentralized governance accelerates innovation and accommodates diverse school contexts but can produce inconsistent data privacy practices, duplicated efforts, and equity gaps between schools with different capacity levels.
Example: Under a decentralized governance model, individual schools can adopt AI tools from an approved vendor list without central approval, but must complete a self-certification checklist confirming privacy compliance.
Declining AI Cost¶
The ongoing trend in which the price of accessing AI capabilities—measured per token processed, per task completed, or per user served—decreases substantially over time due to hardware and software improvements.
Declining AI cost is a key driver of educational AI adoption, making tools that were previously affordable only to elite institutions accessible to under-resourced districts. Strategic plans should model cost trajectories, not just current prices.
Example: The cost of running an AI tutoring session that cost $1.00 per hour in 2023 fell to under $0.10 by 2025, changing the district's budget calculus for a proposed pilot.
Digital Divide¶
The gap in access to digital technology—including devices, reliable internet connectivity, and digital literacy skills—between different groups of people, particularly along lines of income, geography, race, and age.
The digital divide means that AI-enhanced education could increase rather than decrease educational inequality if access is not proactively addressed. Every AI initiative must include an equity analysis of differential access implications.
Example: A district deploying AI tutoring tools discovers that 30% of its students lack reliable home internet access, leading to a partnership with the local library to provide evening access and a hotspot lending program.
Digital Transformation¶
The broad process of integrating digital technologies—including AI—into all areas of an organization's operations to fundamentally change how it delivers value and achieves its goals.
Digital transformation in education goes beyond purchasing new tools; it requires changes to culture, workflows, and professional practice. AI is accelerating the pace and depth of digital transformation in schools.
Example: A district's digital transformation initiative replaces paper-based attendance tracking with an AI-linked system that automatically alerts counselors when students show chronic absence patterns.
Doubling Time¶
The length of time required for a measured quantity—such as AI capability on a specific benchmark or the cost-efficiency of AI inference—to double in value, used to characterize the pace of change.
Shorter doubling times signal faster change and require more agile strategic responses. Education leaders can use documented doubling times in AI capability research to stress-test their multi-year technology plans.
Example: If AI long-task performance has a doubling time of approximately twelve months, a district that delays a two-year implementation decision may face a very different competitive landscape by project completion.
Early Alert System¶
A technology-enabled process that automatically identifies students who show early indicators of academic struggle, disengagement, or risk of course failure, and triggers timely notification to teachers or counselors for intervention.
Early alert systems can prevent small learning gaps from becoming large failures by enabling proactive rather than reactive intervention. AI makes it possible to monitor many more indicators simultaneously than manual monitoring allows.
Example: The district's early alert system flags a student who has missed 20% of her AI tutoring sessions and whose quiz scores have declined for three consecutive weeks, prompting her advisor to schedule a check-in.
Educational Equity¶
The principle and practice of ensuring that every student has fair access to the resources, opportunities, and support they need to achieve their full educational potential, with particular attention to eliminating barriers faced by historically marginalized groups.
Educational equity must be a design criterion, not an afterthought, in AI strategy. Tools that improve average outcomes while widening gaps between groups are not acceptable, regardless of their aggregate performance metrics.
Example: A district's AI equity policy requires that all new tools be analyzed for differential impact across student demographic groups before scaling, with equity improvements as a condition of continued investment.
Equity Impact Scoring¶
A systematic evaluation method that rates the expected effect of a proposed AI initiative on educational equity, considering differential access, potential for bias, and impact on historically underserved student populations.
Building equity impact scoring into the idea evaluation process ensures that equity is a consistent criterion in AI project selection, not an optional consideration applied only when concerns are raised.
Example: An equity impact scoring rubric deducts points from AI proposals that lack a plan for serving students without home internet access, incentivizing proposers to address the digital divide from the start.
Executive Sponsorship¶
Active, visible support from senior organizational leaders for an initiative, including public advocacy, resource authorization, barrier removal, and accountability signaling that the effort is a genuine institutional priority.
AI initiatives without executive sponsorship tend to stall at the pilot stage because they cannot secure sustained resources or overcome organizational resistance. Board-level and superintendent sponsorship is critical for district-wide AI adoption.
Example: The superintendent personally presents the district's AI strategy to the school board, allocates a dedicated budget line, and names herself the executive sponsor, signaling its priority across the organization.
Expert Review Panel¶
A group of knowledgeable individuals—drawn from relevant disciplines such as curriculum, technology, law, and data privacy—convened to evaluate AI proposals, provide specialized input, and recommend decisions.
An expert review panel brings diverse expertise to AI evaluation decisions that no single administrator could provide alone. It also distributes accountability and builds organizational trust in the selection process.
Example: The district assembles an expert review panel including a special education director, a school attorney, a data science consultant, and a parent representative to evaluate its top-five AI proposals.
Explainability¶
The capacity of an AI system to provide understandable reasons for its outputs or decisions in terms that are meaningful to the people affected by them, supporting informed oversight and appropriate trust calibration.
Explainability is particularly important when AI influences consequential educational decisions such as intervention placement or course recommendations. Decisions students and parents cannot understand are decisions they cannot effectively challenge.
Example: An AI early-alert system provides explainability by showing the three specific factors—missed assignments, declining quiz scores, and reduced platform logins—that triggered each student's alert flag.
Exponential Growth¶
A pattern of increase in which a quantity grows by a consistent percentage over each equal time period, resulting in accelerating absolute growth that can be difficult to anticipate intuitively.
AI capability and adoption are both exhibiting exponential growth patterns. Education leaders accustomed to linear change in technology adoption may significantly underestimate how different the AI landscape will look in just a few years.
Example: If AI tutoring usage doubles each year starting from 10,000 student sessions, it reaches over 300,000 sessions by year five—an outcome that linear planning would miss entirely.
Extracurricular Activities¶
Organized educational, social, athletic, artistic, or community engagement programs offered by schools outside of the formal academic curriculum, contributing to students' holistic development and sense of belonging.
Extracurricular activities develop dimensions of student growth—teamwork, creativity, resilience, leadership, and civic identity—that AI cannot replicate. Their importance grows as AI handles more of the academic content delivery function.
Example: A school that adopts AI-assisted instruction uses the time savings to expand its robotics team, debate club, and community garden program, ensuring that students develop rich human experiences alongside AI-enhanced academics.
Fairness In AI¶
The design principle and evaluation standard requiring that an AI system's outputs and impacts do not systematically disadvantage individuals or groups based on protected characteristics such as race, gender, disability, or socioeconomic status.
Fairness in AI is not a single metric but a family of related properties that can sometimes conflict with each other. Education leaders must explicitly define what fairness means for their context and hold AI vendors accountable to it.
Example: A district's AI tutoring contract requires the vendor to provide quarterly demographic performance reports disaggregated by race, disability status, and English learner status to monitor for fairness disparities.
Feasibility Assessment¶
An analysis of whether a proposed AI use case can realistically be implemented given the organization's current technical capabilities, data availability, staff capacity, and budget, within the desired time frame.
Feasibility assessment prevents organizations from committing resources to AI projects that cannot be executed with available capabilities. It complements impact scoring to identify ideas that are both valuable and achievable.
Example: A feasibility assessment determines that a proposed AI scheduling tool requires student data integration that the district's current information system cannot support, recommending the idea be deferred for 18 months.
FERPA Compliance¶
Adherence to the Family Educational Rights and Privacy Act, a U.S. federal law that restricts access to and disclosure of student educational records and grants parents and eligible students the right to inspect, correct, and control those records.
FERPA compliance is a non-negotiable legal requirement for any AI tool that accesses or generates student educational records. Leaders must ensure that vendor data agreements include appropriate school official agreements and use limitations.
Example: A district's FERPA compliance review determines that an AI grading assistant is a legitimate school official under the act, but requires a data processing agreement before the vendor can access student work.
Fifty Percent Reliability¶
The threshold at which an AI system succeeds on a given class of tasks approximately half the time, often used as a practical benchmark for determining when a capability becomes worth deploying in real workflows.
Fifty percent reliability may seem low, but for many low-stakes or easily reviewable educational tasks, it already delivers value by reducing human effort. Leaders should calibrate acceptable reliability thresholds to task criticality.
Example: An AI that drafts parent newsletter text with 50% reliability still saves staff time, because editing a draft is faster than writing from scratch even half the time.
Fine Tuning¶
A process in which a pre-trained foundation model is further trained on a smaller, domain-specific dataset to improve its performance or align its behavior for a particular subject, audience, or task.
Fine tuning allows districts or vendors to customize AI behavior for specific educational contexts, such as a particular grade level's vocabulary or a school's pedagogical approach. It typically costs less than training a model from scratch.
Example: A publisher fine-tunes a foundation model on thousands of annotated elementary school writing samples so that its feedback tool uses language and expectations appropriate for young writers.
Flipped Classroom¶
An instructional model in which students engage with introductory content—often through video or AI-assisted learning—outside of class, and use in-class time for active practice, problem-solving, and discussion facilitated by the teacher.
The flipped classroom model makes particularly good use of AI tools for content delivery and initial practice, freeing precious in-person time for the high-value interactions that human teachers provide most effectively.
Example: In a flipped chemistry class, students watch AI-narrated concept videos at home and answer comprehension checks, then spend class time conducting experiments and working through problems with teacher guidance.
Formative Assessment¶
Ongoing evaluation of student understanding during the learning process—through questions, activities, or observations—used to inform instructional adjustments before summative judgment, rather than to assign grades.
AI dramatically expands the frequency and granularity of formative assessment by monitoring every student interaction with learning materials. This real-time feedback loop enables faster instructional response than traditional weekly quizzes allow.
Example: An AI tutoring platform provides teachers with a daily formative assessment dashboard showing each student's current mastery status across all active learning objectives, updated after each session.
Foundation Model¶
A large AI model trained on broad, general-purpose data that can be adapted or fine-tuned for many specific applications, serving as a reusable base rather than a single-purpose tool.
Foundation models lower the cost of building specialized educational tools because developers can start from a powerful base rather than training from scratch. This accelerates the pace of new product development.
Example: A company builds a student mental-health chatbot by fine-tuning a foundation model on counseling transcripts, rather than building an AI system from the ground up.
Frontier Model¶
The most capable AI models available at a given moment, representing the current leading edge of performance across language, reasoning, and other tasks, typically produced by a small number of well-resourced organizations.
Frontier models set the ceiling for what AI can currently do, and their capabilities are advancing rapidly. Education leaders should monitor frontier model progress to anticipate near-future impacts on teaching and learning.
Example: When researchers report that the latest frontier model can solve complex high-school-level math problems, that is a signal that AI tutoring for STEM subjects is becoming more viable.
Funnel Stage Tracking¶
The practice of recording and monitoring which stage of the idea funnel each AI proposal currently occupies, enabling leaders to identify where ideas are getting stuck and how long the overall process takes.
Funnel stage tracking reveals process bottlenecks—such as evaluation reviews that take too long or feasibility assessments that lack resources—that reduce the organization's overall innovation throughput.
Example: Funnel stage tracking reveals that 40% of submitted ideas have been waiting more than three months for an initial feasibility review, prompting the district to assign a dedicated evaluator role.
Gap Analysis¶
A systematic comparison of an organization's current state against its desired future state in a specific domain—such as AI readiness—used to identify the specific investments, capabilities, or changes needed to close the distance.
Gap analysis translates aspirational AI strategy into concrete action requirements. It prevents vague commitments to "adopt AI" by specifying exactly which capabilities are missing and what must be done to acquire them.
Example: A district's AI gap analysis reveals that it lacks xAPI-compliant learning record infrastructure, AI-literate curriculum staff, and data sharing agreements with key vendors—three specific gaps that must be closed before personalized learning pilots can begin.
Generative AI¶
A category of artificial intelligence systems capable of producing new content—such as text, images, audio, or video—that resembles human-created work, based on patterns learned during training.
Generative AI is reshaping content creation in education, raising both opportunities for personalized materials and concerns about academic honesty. Leaders need policy frameworks to govern its use.
Example: A district curriculum coordinator uses generative AI to create differentiated versions of a social studies reading at three different Lexile levels.
Hallucination¶
A phenomenon in which an AI language model generates information that is factually incorrect, fabricated, or nonsensical while presenting it with apparent confidence and fluency.
Hallucinations are a significant risk in educational settings where accuracy matters. Leaders must establish policies for human review of AI-generated content and teach students to verify AI outputs against authoritative sources.
Example: An AI research assistant tells a student that a specific court case was decided in 1987, but that case does not exist—the AI invented a plausible-sounding but false citation.
Human Agent Collaboration¶
The working relationship between human educators, administrators, or students and AI agents, characterized by complementary contributions where humans provide judgment, context, and accountability while AI agents provide scale, consistency, and data processing.
Effective human-agent collaboration design is a critical success factor for AI in education. The most effective AI deployments are those where human and AI contributions are explicitly designed to complement each other's strengths and compensate for each other's limitations.
Example: In a human-agent collaboration model for student support, the AI agent monitors all 500 students' learning metrics daily and surfaces the 12 who most need attention, while the counselor focuses her limited time on those highest-priority cases.
Human In The Loop¶
A design principle for AI systems in which a human reviews, approves, or can override AI-generated outputs or decisions at key points, maintaining human accountability for consequential actions.
Human-in-the-loop design is a critical safeguard for high-stakes AI applications in education such as grade recommendations, disciplinary flags, or learning plan modifications. It ensures accountability remains with educators, not algorithms.
Example: An AI-generated IEP goal recommendation system uses a human-in-the-loop design in which every AI suggestion must be reviewed and explicitly approved by the special education teacher before it is added to the student's plan.
Hyperpersonalized Learning¶
An approach to education in which every aspect of a student's learning experience—including content, pacing, format, feedback, and goal setting—is continuously tailored to that individual's unique profile by AI systems operating at scale.
Hyperpersonalization extends well beyond differentiating reading levels; it encompasses learning style, motivation state, social context, and long-term goals. It represents the furthest extension of personalized learning enabled by AI.
Example: A hyperpersonalized learning system adjusts not only the difficulty of problems but also the narrative framing of word problems to match each student's stated interests and cultural background.
Idea Evaluation¶
The systematic process of assessing submitted AI use-case proposals against defined criteria—such as educational impact, feasibility, cost, risk, and alignment with district priorities—to determine which ideas merit further development.
Structured idea evaluation ensures that decisions about which AI projects to pursue are transparent, consistent, and based on evidence rather than political relationships or personal enthusiasm.
Example: A five-member panel of curriculum directors, technology staff, and a data privacy officer evaluates each submitted AI idea using a common rubric before recommending a shortlist to the superintendent.
Idea Feedback Loop¶
A process by which the outcomes of implemented AI ideas—including successes, failures, and unexpected findings—are systematically captured and used to improve future idea generation, evaluation, and implementation.
Without a feedback loop, organizations repeat mistakes and miss opportunities to learn from experience. Regular retrospectives that feed findings back into the idea funnel improve the quality of the entire innovation process over time.
Example: After a pilot AI grading tool underperforms, the district captures lessons in its idea registry and updates its evaluation rubric to require efficacy evidence before approving similar future proposals.
Idea Funnel¶
A structured process for systematically collecting, evaluating, and progressing AI use-case ideas from initial submission through feasibility assessment to project approval and implementation.
An idea funnel ensures that AI innovation is inclusive and evidence-based rather than driven solely by vendor relationships or leadership preferences. It captures the distributed expertise of teachers and staff who encounter problems daily.
Example: A district's AI idea funnel receives 150 staff submissions in its first quarter, of which 12 pass feasibility review and 3 are approved as funded pilots.
Idea Generation¶
The first stage of the innovation process in which individuals or teams identify and articulate potential applications of AI to solve problems or create new value within the educational organization.
Structured idea generation practices—such as hackathons, problem-focused workshops, or regular submission opportunities—produce more and better ideas than relying on spontaneous proposals. Broad participation improves both quantity and diversity of ideas.
Example: A district hosts a half-day "AI for Education" workshop where teachers brainstorm potential AI applications for their classrooms, generating 80 candidate ideas across seven subject areas.
Idea Metadata¶
Descriptive information associated with a submitted AI use-case idea—such as proposer name, submission date, subject area, grade level, estimated impact, and evaluation status—used to organize and filter the idea registry.
Consistent metadata enables meaningful analysis of the idea funnel, such as identifying which departments generate the most ideas or which problem types attract the highest feasibility scores.
Example: Each entry in the district's idea registry includes metadata fields for grade band, subject, estimated number of students affected, and submitter department, enabling filtered reports by school board members.
Idea Recognition Awards¶
A formal program that acknowledges and celebrates staff members who submit AI ideas that are selected for piloting, successfully implemented, or produce documented educational improvements.
Recognition programs signal that leadership values staff-generated innovation and encourage broader participation in the idea funnel. They also help sustain engagement in the innovation process over multiple years.
Example: At the district's annual technology showcase, three teachers receive innovation awards and a small stipend for AI ideas that were piloted and produced measurable improvements in student engagement.
Idea Registry¶
A centralized database or repository in which all submitted AI use-case ideas are recorded, tracked, and made accessible to evaluators and stakeholders throughout the idea funnel process.
An idea registry prevents duplication of effort, enables pattern recognition across submissions, and creates organizational memory about what has been proposed and evaluated. It also demonstrates that the organization takes staff input seriously.
Example: The district's idea registry contains 200 entries from the past two years, annotated with evaluation status, and is searchable by subject area, grade level, and problem type.
Idea Submission Form¶
A standardized digital or paper form used to capture essential information about a proposed AI use case from staff, ensuring that each idea is documented consistently and completely enough to be evaluated.
A well-designed idea submission form reduces the friction of participating in the innovation process and ensures evaluators have the information they need. It signals to staff that their ideas are welcome and will be taken seriously.
Example: The district's idea submission form asks proposers to describe the problem, the proposed AI solution, the expected benefit, and any known risks, taking approximately 10 minutes to complete.
Image Generation¶
The AI capability to create original visual content—including illustrations, diagrams, charts, and artwork—from text descriptions or other inputs, without requiring human artistic skill.
Image generation can significantly reduce the cost and time required to produce engaging visual learning materials. It also raises questions about copyright, authenticity, and responsible use that schools must address in policy.
Example: A science teacher uses an AI image generator to create accurate cross-sectional diagrams of cell structures customized to match the vocabulary in her district's adopted textbook.
Implementation Roadmap¶
A time-phased plan that sequences the activities, milestones, resources, and decisions required to move an AI initiative from approval to full operational deployment, providing a shared reference for all involved parties.
An implementation roadmap converts strategic intent into actionable steps with clear accountability. It prevents the common failure mode of AI initiatives that are approved but never actually deployed due to planning gaps.
Example: The district's AI tutoring implementation roadmap specifies vendor contracting in month one, IT integration in months two and three, teacher training in month four, and phased student rollout in months five through eight.
Inclusive Design¶
An approach to creating products, services, and learning experiences that proactively considers the needs of people with diverse abilities, backgrounds, and circumstances, reducing or eliminating barriers to participation and benefit.
Inclusive design in educational AI ensures that tools are usable and effective for students with disabilities, English language learners, and other groups whose needs may not have been centered in the original design.
Example: An AI tutoring platform uses inclusive design principles to ensure that all content is accessible to screen readers, available in multiple languages, and usable with keyboard-only navigation for students with motor impairments.
Inference¶
The process by which a trained AI model applies its learned knowledge to new inputs in order to produce outputs such as answers, predictions, translations, or generated content.
Inference is what happens every time a student or teacher uses an AI tool in real time. The speed and cost of inference are key factors in whether AI-powered tools can scale across an entire district.
Example: When a student types a question into an AI tutoring app and receives an answer within seconds, that response is generated through inference.
Innovation Culture¶
An organizational environment in which staff are encouraged to propose and test new ideas, learn from failures, and continuously improve practices, supported by leadership behaviors and institutional structures.
Building an innovation culture is a prerequisite for sustained AI adoption. Without psychological safety to experiment and fail, staff will default to familiar practices even when better AI-enabled approaches are available.
Example: A principal builds innovation culture by publicly celebrating a teacher whose AI-assisted reading intervention failed to meet its goal, praising the team's learning and revised approach.
Institutional Archetype¶
A categorization of an educational organization based on its characteristic combination of current AI readiness, resource profile, governance culture, and strategic orientation, used to tailor AI strategy recommendations to organizational context.
Different institutional archetypes face different AI strategy challenges. A large urban district with strong technical infrastructure needs a different playbook than a small rural district with limited staffing, even if both have similar educational goals.
Example: A district identifies itself as a "cautious late-majority" institutional archetype—risk-averse and resource-constrained—and selects an AI strategy focused on proven tools with strong vendor support rather than experimental innovation.
Intelligence Versus Price¶
The relationship between the measurable capability of an AI system and its cost to use, capturing the trend toward increasingly powerful AI at decreasing cost over time.
Improving intelligence-to-price ratios are rapidly democratizing access to powerful AI tools. Districts that previously could not afford advanced AI capabilities may find them accessible within their existing technology budgets.
Example: An AI tutoring model that cost $20 per student per month two years ago now offers equivalent capability at $2 per student per month, making district-wide deployment financially viable.
Intelligent Textbook¶
A digital educational resource that uses AI to adapt its content, pacing, assessments, and explanations dynamically to each learner's demonstrated knowledge and needs, rather than presenting a fixed sequence of material.
Intelligent textbooks represent a fundamental shift in how curriculum is delivered and experienced. They can provide every student with a personalized learning pathway at a cost approaching that of traditional textbooks.
Example: An intelligent textbook in high school biology adjusts which sections a student reads next based on her quiz performance, skipping material she has mastered and reinforcing concepts where she shows gaps.
Interactive Simulation¶
A digital learning experience that allows students to manipulate variables and observe outcomes in a modeled system, enabling exploration of concepts that would be impractical, expensive, or dangerous to demonstrate in the physical world.
Interactive simulations deepen conceptual understanding by giving students agency to experiment and discover principles through direct manipulation. AI can generate and customize simulations rapidly, expanding access to this powerful pedagogy.
Example: A chemistry class uses an interactive simulation to test how changing temperature and concentration affects reaction rates, observing results that would require expensive equipment in a physical lab.
Key Performance Indicator¶
A specific, regularly tracked measure that reflects progress toward a critical organizational goal, used to monitor AI initiative health and inform leadership decisions about resource allocation and strategic direction.
Key performance indicators for AI initiatives should be reviewed on a predictable schedule and presented in dashboards accessible to school leaders. They create accountability and enable early identification of underperforming tools.
Example: A district tracks "percentage of students receiving personalized practice recommendations weekly" as a key performance indicator for its AI-powered learning platform rollout.
Knowledge Organization¶
The structured classification and arrangement of information within an institution—including curriculum content, policies, procedures, and expertise—to make it findable, shareable, and usable by both humans and AI systems.
Well-organized institutional knowledge is a prerequisite for effective AI deployment, because AI tools can only search and use information that is accessible and structured. Investing in knowledge organization now reduces AI implementation friction later.
Example: A district that organizes its curriculum materials in a tagged, searchable repository can more easily deploy an AI assistant that retrieves relevant resources for teachers by standard and grade level.
Large Language Model¶
A type of AI system trained on vast quantities of text that can understand, summarize, translate, and generate human language across a wide range of topics and styles.
Large language models power many AI tools educators encounter today, including tutoring chatbots and writing assistants. Understanding what they are helps leaders assess appropriate use and risk.
Example: ChatGPT is a large language model that a teacher might use to generate first-draft lesson plans, which the teacher then reviews and revises.
Learning Analytics¶
The measurement, collection, analysis, and reporting of data about learners and their contexts, with the goal of understanding and improving student learning outcomes and institutional performance.
Learning analytics transforms the data generated by digital learning tools from a byproduct into a strategic asset. When well-implemented, it enables early intervention, personalized support, and evidence-based curriculum improvement.
Example: Learning analytics reveals that students who complete practice problems on Friday evenings score 12% higher on Monday assessments, informing a campaign to increase access to study tools outside school hours.
Learning Management System¶
A software platform used by educational institutions to organize, deliver, track, and manage instructional content, assessments, communication, and learner progress data across courses, classes, or programs.
Learning management systems are the primary digital infrastructure for most schools and districts. Their data architecture and API capabilities determine how easily AI tools can be integrated to enhance the learning experience.
Example: Teachers use the district's learning management system to post assignments, grade submissions, communicate with students, and monitor class-level progress, all within a single application.
Learning Record¶
A digital data object that documents a specific interaction between a learner and an educational activity, capturing information such as what was attempted, how the learner responded, the time taken, and the outcome.
Learning records are the raw material of personalized education and learning analytics. Their consistent collection across tools and platforms enables AI systems to build accurate models of individual student knowledge and progress.
Example: When a student answers a practice question in an AI tutoring app, the system creates a learning record capturing the question identifier, the student's response, the correctness, and the timestamp.
Learning Record Store¶
A centralized database designed to receive, store, and retrieve xAPI-formatted learning records from multiple educational tools and platforms, serving as the unified repository of student learning activity data.
A learning record store gives districts control over their students' learning data, enabling cross-platform analytics and reducing dependency on individual vendor dashboards. It is infrastructure for data-driven personalization at scale.
Example: A district's learning record store aggregates xAPI data from its reading app, math platform, and AI tutoring tool, allowing counselors to see a student's full learning activity in a single dashboard.
Learning Telemetry¶
The continuous, automatic collection of granular behavioral and performance data from student interactions with digital learning tools—such as time-on-task, click patterns, and response sequences—used to infer learning states and progress.
Learning telemetry provides far richer information about the learning process than traditional assessments alone. However, its collection must be governed carefully to prevent surveillance overreach and protect student privacy.
Example: Learning telemetry from a math platform reveals that students who pause longest before answering fraction problems are more likely to get them wrong, informing the design of a targeted just-in-time hint feature.
Lessons Learned¶
Documented insights about what worked well, what failed, and what would be done differently, captured systematically after completing or reviewing an AI initiative and shared across the organization to improve future efforts.
Lessons-learned documentation transforms individual project experiences into organizational knowledge. Without this practice, schools repeat expensive mistakes and fail to build cumulative expertise in AI implementation.
Example: The lessons-learned report from a failed AI chatbot pilot notes that insufficient teacher training—not the technology itself—caused poor adoption, informing the design of all subsequent AI rollouts.
Lifelong Learning¶
The ongoing, self-motivated pursuit of knowledge and skill development throughout an individual's life, beyond formal schooling, driven by personal, professional, or civic goals.
In a world of rapid AI-driven change, lifelong learning is no longer optional but necessary for workforce relevance and personal flourishing. Schools that cultivate curiosity, self-direction, and learning habits are preparing students for a fundamentally different future.
Example: A high school graduation requirement includes a self-directed learning project in which each student pursues a personal interest using a combination of AI-assisted research and community mentorship.
Local AI Models¶
AI models that run entirely on hardware owned or controlled by the school or district, rather than sending data to external cloud servers, providing greater data privacy and operational independence.
Local AI models are particularly important for districts with strict data privacy requirements or unreliable internet connectivity. They represent an emerging option that may become increasingly practical as model sizes shrink.
Example: A school installs a local AI model on a dedicated server, allowing students to use an AI writing assistant during assessments without any data being transmitted outside the school building.
Long Task Rate¶
The proportion of extended, multi-step tasks—those requiring sustained effort over minutes or hours—that an AI agent completes successfully without human intervention, used as a metric of agent reliability.
Rising long task rates signal that AI is moving from a tool that handles quick queries to one capable of sustained educational work. This shift has direct implications for teacher workload and administrative automation.
Example: A long task rate of 30% means an AI agent successfully completes roughly one in three hour-long independent tasks, requiring human review for the remaining two.
Machine Learning¶
A subset of artificial intelligence in which computer systems improve their performance on a task by automatically detecting patterns in large datasets, rather than following hand-coded rules.
Machine learning underlies most modern AI tools used in education, from recommendation engines to plagiarism detectors. Knowing this helps leaders evaluate vendor claims and set realistic expectations.
Example: A reading platform uses machine learning to adjust the difficulty of passages based on a student's past quiz scores.
Mastery Based Progression¶
An educational structure in which students advance to new content only after demonstrating defined levels of competency in prerequisite material, rather than advancing with peers based on time elapsed.
Mastery-based progression ensures that students build on solid foundations rather than accumulating gaps that compound over time. AI makes it operationally feasible to track mastery individually and provide the additional practice each student needs.
Example: In a mastery-based progression math program, a student who has not yet mastered multi-digit multiplication receives additional AI-tutored practice before accessing the division unit, regardless of the class calendar.
Mastery Tracking¶
The continuous monitoring and recording of whether individual students have demonstrated sufficient proficiency in each specific skill or concept to move forward, rather than tracking progress by time spent or content covered.
Mastery tracking shifts the educational focus from seat time to learning outcomes. AI enables real-time mastery tracking across many more skills than a teacher could manually monitor, enabling timely intervention and advancement.
Example: An AI system tracks each student's mastery status across 40 arithmetic skills, automatically unlocking new content only when the student demonstrates 80% accuracy on three consecutive assessments.
Mentorship Model¶
An educational framework in which adult educators primarily serve as trusted guides, advisors, and personal development coaches for students, focusing on relationships, character, and life skills rather than content transmission.
As AI takes on more content instruction, the mentorship model may represent the future of the teacher's core value proposition. Investing in mentorship skills and structures now prepares schools for a more AI-integrated future.
Example: In a mentorship model school, each teacher is assigned to fifteen students as their personal academic and life coach, meeting weekly to discuss goals, challenges, and growth across all areas of the student's development.
METR Study¶
A research evaluation conducted by the organization METR that measures the ability of AI agents to complete long, complex, real-world software tasks, providing longitudinal data on AI capability growth.
The METR study is significant because it documents measurable, rapid growth in AI autonomous task completion. Education leaders can use these findings to anticipate when AI will be capable of handling progressively more complex educational processes.
Example: METR data showing that AI agents doubled their long-task success rate in under a year prompted one district technology team to revisit its three-year AI automation roadmap.
MicroSim¶
A small, focused interactive simulation embedded within a learning resource that illustrates a single concept or relationship through direct student manipulation, typically requiring less than five minutes to explore.
MicroSims lower the barrier to incorporating interactive learning into any lesson by being lightweight and embeddable. AI tools can now generate custom MicroSims from a concept description, democratizing access to this form of learning.
Example: A teacher embeds a MicroSim in a lesson on supply and demand that lets students drag a supply curve and immediately observe how equilibrium price changes.
Misinformation Risk¶
The danger that AI systems will generate, amplify, or lend credibility to false or misleading information, contributing to the spread of inaccurate beliefs among students and communities.
Misinformation risk is heightened in education because students are developing their epistemic habits and may not yet have the critical literacy skills to distinguish AI-generated falsehoods from accurate information.
Example: An AI social studies tool generates a passage claiming that a recent election result was fraudulent, illustrating the misinformation risk when AI tools are not constrained to verified factual sources.
MMLU Benchmark¶
A specific AI evaluation consisting of approximately 14,000 multiple-choice questions spanning 57 academic subjects, used to measure the breadth of factual and reasoning knowledge in large language models.
MMLU scores provide one widely recognized measure of general academic knowledge in AI systems. Education leaders can use MMLU trends to track how AI knowledge breadth is expanding over time.
Example: Researchers note that a new language model achieves 90% on the MMLU benchmark, surpassing the average score of human experts on the same questions.
Model Parameters¶
The internal numerical values within an AI model that are adjusted during training to encode learned patterns and relationships, collectively determining how the model responds to any given input.
Parameter count is often cited as a rough proxy for model capability. While not the only measure of quality, understanding this concept helps leaders interpret vendor comparisons and benchmark claims.
Example: When a vendor says their model has "70 billion parameters," they mean it has 70 billion adjustable values that together determine its responses.
Model Release Cadence¶
The pace at which AI developers publish new or significantly improved versions of their models, which determines how frequently the capabilities available to educational tool builders and users change.
Rapid release cadence creates both opportunity and disruption: tools built on last year's models may already be outperformed by newer alternatives. Procurement and vendor evaluation cycles must account for this pace.
Example: A vendor releases a major model update every six months, meaning a district evaluating the tool in January should request performance data from the most recent release, not the version they originally piloted.
Moores Law Analogy¶
A comparison of AI capability growth to Moore's Law, the historical observation that the number of transistors on a chip doubled approximately every two years, used to communicate the rapid and sustained pace of AI advancement.
Invoking the Moore's Law analogy helps non-technical stakeholders grasp that AI improvement is not a one-time event but a sustained trend. However, leaders should note that AI improvement rates may actually exceed the original Moore's Law pace.
Example: A superintendent uses the Moore's Law analogy to explain to her school board why an AI tutoring pilot that seems impressive today may be far more powerful by the time it scales district-wide.
Multi Agent Coordination¶
The set of protocols, communication structures, and shared information frameworks that enable multiple AI agents to work together toward a common goal without duplicating effort, creating conflicts, or losing coherence.
Multi-agent coordination becomes important as schools deploy AI agents for different functions—tutoring, scheduling, communication, and monitoring—that must share student information and align their actions without human orchestration of every interaction.
Example: Multi-agent coordination ensures that when the early-alert agent flags a student for academic risk, the tutoring agent automatically intensifies support and the parent communication agent sends a timely notification, all without staff intervention.
Multimodal AI¶
An AI system capable of processing and generating multiple types of content—such as text, images, audio, and video—within a single model, enabling richer and more flexible interactions.
Multimodal AI opens new possibilities for inclusive education, allowing students to interact with learning systems through speech, images, or diagrams rather than text alone, reducing barriers for diverse learners.
Example: A student photographs a handwritten math problem and submits it to a multimodal AI tutor, which reads the image and provides a step-by-step solution.
Neural Network¶
A computational model loosely inspired by the structure of the human brain, consisting of interconnected layers of mathematical nodes that process and transform data to recognize patterns or make predictions.
Neural networks are the foundation of most modern AI systems. A conceptual understanding helps administrators ask better questions when vendors describe how their AI tools work.
Example: The image-recognition system that grades handwritten math work uses a neural network trained on millions of examples of student handwriting.
One Shot Task¶
A task that an AI system completes in a single, uninterrupted attempt without intermediate human feedback, correction, or guidance, used to evaluate autonomous task completion under realistic conditions.
One-shot performance is a useful metric for evaluating AI tools intended to reduce teacher workload, since real-world use rarely allows for iterative prompting. Leaders should test AI tools on one-shot tasks before deployment.
Example: Asking an AI to generate a complete, standards-aligned quiz on the water cycle without any follow-up instructions is a one-shot task that tests the AI's autonomous capability.
Open Educational Resources¶
Teaching and learning materials that are freely available for anyone to use, adapt, and redistribute without cost or restrictive licensing, often produced by governments, universities, and nonprofit organizations.
Open educational resources significantly reduce curriculum costs and give teachers flexibility to customize materials. AI is both consuming open educational resources as training data and creating new ones at scale.
Example: A district replaces $120 per-student commercial textbooks with a curated collection of open educational resources, freeing budget for AI-powered tutoring tools.
Open Source Models¶
AI models whose underlying code, architecture, and often weights are publicly released under licenses that allow anyone to inspect, use, modify, and redistribute them, without requiring payment to a single vendor.
Open source models give districts more control, lower costs, and reduced vendor dependency compared to proprietary alternatives. However, they may require more technical expertise to deploy and maintain effectively.
Example: A district deploys an open source language model on its own servers to power a tutoring chatbot, ensuring that student conversation data never leaves the district's infrastructure.
Opportunities Analysis¶
The component of a SWOT analysis that identifies favorable external conditions, emerging AI capabilities, funding sources, partnership possibilities, or competitive dynamics that an organization can exploit in its AI strategy.
Opportunities analysis prevents organizations from planning in a vacuum, disconnected from the rapid external changes in AI capability and policy that could dramatically affect their strategic options within a short time.
Example: An opportunities analysis identifies a new state grant program for AI literacy and an emerging low-cost AI tutoring platform that together could make a district's personalized learning vision affordable within the current budget cycle.
Over Reliance On AI¶
A condition in which students or educators depend so heavily on AI tools for tasks they should perform themselves that they fail to develop or maintain essential knowledge, skills, and judgment.
Over-reliance on AI is a genuine developmental risk, particularly for students who use AI to bypass effortful learning rather than to enhance it. Pedagogical design must ensure AI scaffolds learning rather than substituting for it.
Example: A teacher notices that her students can no longer estimate arithmetic answers mentally after a year of AI calculator dependence, illustrating how over-reliance on AI can erode foundational skills.
Parent Communication Agent¶
An AI agent designed to generate, schedule, and deliver personalized communications to parents and guardians about their child's academic progress, attendance, upcoming events, and available support resources.
Parent communication agents can dramatically increase the frequency and personalization of school-to-home communication, keeping families informed and engaged without adding to teacher workload. Privacy and tone governance are essential design requirements.
Example: A parent communication agent sends weekly personalized summaries to each family describing their child's learning activity, recent achievements, and any areas where additional practice at home might help.
Parent Engagement¶
The process of informing, consulting, and involving parents and guardians in decisions about AI tools and practices that affect their children, building transparency, trust, and shared understanding of benefits and risks.
Parents have both legal rights regarding student data and a legitimate interest in understanding how AI is shaping their children's education. Proactive engagement is more effective than reactive communication after concerns escalate.
Example: Before deploying AI writing assistants in high school English classes, a district hosts evening information sessions, publishes a parent FAQ, and creates an opt-out process for families with concerns.
Personal AI Agents¶
AI systems that serve as individualized digital assistants for specific users—such as individual students or teachers—learning their preferences, goals, and context over time to provide increasingly tailored support and automation.
Personal AI agents represent a significant near-future development with profound implications for how students learn and how teachers work. Governance frameworks must address agent autonomy, data privacy, and the preservation of human developmental agency.
Example: A high school student's personal AI agent tracks her academic goals, reminds her of upcoming deadlines, suggests study resources based on her performance data, and drafts messages to her advisor for her review.
Personalized Learning Path¶
The individualized sequence of content, activities, and assessments that a student follows based on their unique knowledge profile, learning preferences, pace, and goals, distinguished from a standard curriculum by its adaptation to the individual.
Personalized learning paths represent the holy grail of differentiated instruction, long desired but impractical to create and manage manually at scale. AI makes truly individualized pathways feasible for every student simultaneously.
Example: Two students in the same class follow different personalized learning paths through a fractions unit: one reinforces conceptual understanding through visual models while the other advances to application problems.
Pilot Program¶
A small-scale, time-limited implementation of an AI tool or initiative in a defined subset of the organization, designed to test feasibility, gather evidence, learn from experience, and inform decisions about broader adoption.
Pilot programs are the responsible middle path between doing nothing and committing to full-scale deployment before evidence is available. Well-designed pilots generate the data needed to make scaling decisions with confidence.
Example: A district pilots an AI tutoring tool in two schools for one semester before making a district-wide adoption decision, using the pilot to evaluate learning outcomes, teacher satisfaction, technical reliability, and equity impacts.
Pipeline Report¶
A regular summary document or dashboard that shows the status of all AI initiatives across every stage of the idea funnel—from submission through evaluation, selection, development, and evaluation—giving leaders a unified view of AI activity.
Pipeline reports create organizational transparency about AI investment and progress, enabling leaders to identify bottlenecks, celebrate successes, and make timely resource adjustments across the entire AI portfolio.
Example: The district's monthly AI pipeline report shows 8 ideas in evaluation, 3 projects in pilot, 1 being scaled, and 2 recently retired, giving the board a clear picture of AI activity.
Policy Framework¶
A structured set of principles, rules, procedures, and standards that guides consistent decision-making and action within an organization regarding a specific domain, providing a coherent foundation for specific policies and operational decisions.
A policy framework for AI in education ensures that individual policies—such as data privacy, academic integrity, and use guidelines—are coherent, consistent, and collectively sufficient to govern AI responsibly across all contexts.
Example: A district's AI policy framework establishes five core principles—safety, equity, transparency, effectiveness, and human oversight—that every specific AI policy must reflect and advance.
Predictive Analytics¶
The use of statistical models and machine learning to analyze historical data patterns and forecast future student behaviors or outcomes—such as likelihood of dropping a course, failing a grade, or needing additional support.
Predictive analytics can enable proactive rather than reactive educational interventions, but its use raises significant equity and ethics concerns if predictions lead to lower expectations or discriminatory treatment of flagged students.
Example: A predictive analytics model identifies ninth-grade students who are statistically likely to drop out before graduation based on early attendance and grade patterns, enabling targeted counseling interventions.
Private Knowledge¶
Information specific to an organization—such as district policies, internal assessments, or student records—that has not been published publicly and is therefore not part of an AI model's standard training data.
Accessing private knowledge typically requires additional technical steps such as retrieval-augmented generation or fine-tuning. Leaders must also ensure that sharing private knowledge with AI systems complies with privacy regulations.
Example: A district policy handbook is private knowledge; an off-the-shelf AI chatbot cannot answer questions about it unless that document is explicitly provided to the system.
Pro-Social Learning¶
Educational activities and environments intentionally designed to develop students' empathy, cooperation, conflict resolution, community contribution, and care for others, cultivating the social and moral dimensions of human development.
As AI takes on more academic instruction, pro-social learning becomes a more distinctive and essential function of human-centered education. Schools should expand pro-social learning opportunities as AI efficiency gains free up time.
Example: A school uses time freed by AI-efficient academic instruction to double its community service hours, teamwork projects, and peer mentoring programs, explicitly developing students' capacity for empathy and collaboration.
Problem Statement¶
A clear, concise description of a specific educational challenge or inefficiency that an AI solution is intended to address, written in terms of stakeholder impact rather than technology features.
A strong problem statement keeps AI development grounded in real need rather than technology enthusiasm. It also serves as the primary criterion against which a proposed solution's success will eventually be measured.
Example: "Teachers at our three middle schools spend an average of 90 minutes per week writing differentiated reading assignments, leaving less time for direct instruction" is a well-formed problem statement for an AI use case.
Problem Taxonomy¶
A structured classification system that categorizes educational challenges by type, domain, audience, and severity, used to organize the problems that AI use cases are intended to solve and to identify patterns across submissions.
A problem taxonomy helps leaders see which types of challenges are generating the most AI innovation energy, and whether high-priority institutional problems are being addressed or overlooked in the idea funnel.
Example: A district's problem taxonomy groups submitted AI ideas under categories such as "student academic support," "teacher workload," "administrative efficiency," and "family communication," enabling portfolio analysis.
Professional Development¶
Structured learning experiences designed to improve the knowledge, skills, and practices of educators and administrators, enabling them to be more effective in their roles and to implement new approaches such as AI-enhanced instruction.
Professional development is the primary mechanism for translating AI strategy into changed classroom and administrative practice. Without sustained, high-quality professional development, AI tools will be underused or misused.
Example: A district allocates 12 hours of professional development time per teacher per year specifically for AI literacy, tool-specific training, and pedagogical integration coaching, paid for through a combination of title funds and district budget.
Progress Monitoring Agent¶
An AI agent specifically designed to continuously track individual student learning activity and performance across multiple platforms, detect deviations from expected progress, and alert human educators when intervention may be needed.
Progress monitoring agents address a real capacity limitation in education: teachers cannot simultaneously track every student's progress in real time across all subjects. These agents extend teacher awareness without replacing teacher judgment.
Example: A progress monitoring agent identifies that a student's math practice session duration has declined by 50% over two weeks and her quiz scores have dropped, automatically notifying her teacher with a summary of the pattern.
Project Based Learning¶
A teaching approach in which students gain knowledge and skills by working over an extended period on a complex, real-world question or challenge, producing a tangible product or presentation as evidence of learning.
Project-based learning develops skills—such as collaboration, communication, and critical thinking—that AI cannot easily replicate or replace. As AI automates more routine academic tasks, project-based learning becomes more, not less, important.
Example: Students in a seventh-grade project-based learning unit design a proposal for improving water quality in their local river, integrating science, mathematics, and civic knowledge over six weeks.
Project Evaluation¶
A systematic review of an AI initiative's progress and outcomes at defined milestones, assessing whether it is meeting its stated objectives, producing expected benefits, and operating within acceptable risk parameters.
Regular project evaluation prevents sunk-cost continuation of failing AI initiatives and enables timely course corrections. It also generates the evidence needed to justify scaling successful pilots to additional schools or grade levels.
Example: At the six-month mark of the AI tutoring pilot, an evaluation team reviews student achievement data, teacher satisfaction surveys, and cost tracking to decide whether to continue, modify, or end the project.
Project Portfolio¶
The complete set of active and planned AI initiatives that an organization is managing at a given time, viewed collectively to balance risk, resource allocation, and strategic alignment across all efforts.
Managing AI as a portfolio prevents over-investment in a single approach and ensures a healthy mix of quick wins, medium-term improvements, and longer-horizon strategic bets. Portfolio visibility enables better resource and priority decisions.
Example: The district's AI project portfolio dashboard shows three active pilots, two in development, and five approved but not yet started, giving the superintendent a unified view of all AI investment activity.
Project Selection¶
The decision-making process by which an organization chooses which AI proposals from the evaluated idea pool to fund, resource, and move into active implementation, based on strategic fit, impact, and feasibility.
Project selection is where strategic priorities become real commitments. A transparent, criteria-based selection process ensures that the organization's finite resources are deployed on the highest-value AI opportunities.
Example: Following idea evaluation, the district's technology committee selects three AI projects for the upcoming school year: an AI writing coach pilot, an attendance prediction tool, and an AI-assisted IEP assistant.
Project Team Formation¶
The process of assembling the cross-functional group of individuals—including subject matter experts, technical staff, end users, and administrators—who will design, implement, and evaluate a specific AI initiative.
Effective team formation ensures that AI projects benefit from diverse perspectives from the start, reducing the risk that solutions are technically sound but educationally irrelevant or organizationally unacceptable.
Example: The AI grading pilot team includes two English teachers who will use the tool, an IT integration specialist, the district's data privacy coordinator, and a student representative from the high school advisory council.
Prompt¶
A text input, question, or instruction provided by a user to an AI system that initiates or guides the system's response.
The quality of a prompt strongly influences the usefulness of AI output. Teaching students and staff to write effective prompts is a foundational AI literacy skill with immediate practical value.
Example: A teacher types "Write three discussion questions about the causes of World War I suitable for eighth graders" into an AI assistant, which is the prompt.
Prompt Engineering¶
The practice of designing, refining, and structuring inputs to an AI system to elicit more accurate, useful, or appropriately formatted responses.
Prompt engineering is an emerging professional skill relevant to all educators and administrators who use AI tools. Even modest improvements in prompting can significantly increase the quality and reliability of AI outputs.
Example: A librarian discovers that adding "Explain your reasoning step by step" to research queries produces more reliable AI citations than asking questions without that instruction.
Public Knowledge¶
Information that is freely available in publicly accessible sources—such as textbooks, websites, and published research—and that AI models have likely encountered during their training.
Most large AI models have strong command of public knowledge, making them useful for general tutoring. However, school-specific policies, local context, and proprietary curriculum are not part of this knowledge base.
Example: A student asks an AI tutor to explain photosynthesis, and the AI answers accurately because the concept is thoroughly covered in public scientific literature.
Quarterly Executive Report¶
A structured summary prepared for senior leaders and board members every three months, presenting key AI portfolio metrics, notable successes and failures, upcoming decisions, and strategic implications in non-technical language.
Regular executive reporting keeps AI strategy visible at the leadership level and enables timely strategic adjustments. It also demonstrates accountability to the board and community by linking AI investments to educational outcomes.
Example: The superintendent's quarterly AI executive report includes a one-page dashboard of pilot outcomes, a summary of lessons learned, and a recommendation to scale one successful tool to all elementary schools.
Quick Wins¶
AI use cases that can be implemented rapidly with modest resources, producing clear and demonstrable value in a short time frame, used to build organizational confidence and momentum for larger AI initiatives.
Quick wins are important because they demonstrate AI's practical value to skeptical stakeholders and provide early learning that informs larger projects. Leaders should deliberately include quick wins in every AI implementation roadmap.
Example: Deploying an AI tool that generates substitute teacher lesson plan templates in one week is a quick win that immediately reduces a common teacher pain point and builds staff enthusiasm for AI tools.
Reasoning Model¶
An AI system specifically designed or trained to perform multi-step logical reasoning, solving complex problems by working through intermediate steps rather than pattern-matching directly to a final answer.
Reasoning models represent a qualitative shift in AI capability, enabling tutoring systems to explain their thinking and walk students through problem-solving processes rather than just providing answers.
Example: A reasoning model helps a student solve an algebraic equation by explaining each transformation step and asking the student to verify each one before proceeding.
Recommended Learning Plan¶
An AI-generated, individualized sequence of learning activities, resources, and milestones tailored to a specific student's current knowledge state, learning goals, and preferred pace, presented as actionable next steps.
Recommended learning plans extend the personalization that skilled tutors provide to every student in a system, regardless of class size. Teachers should review and adjust AI-generated plans to ensure they reflect holistic knowledge of the student.
Example: After diagnosing a student's algebra readiness, an AI system generates a recommended learning plan with daily practice goals, specific video resources, and a projected timeline for mastering each standard.
Resource Assignment¶
The allocation of staff time, budget, tools, and authority to specific AI projects within the portfolio, ensuring each initiative has what it needs to succeed without depleting capacity for ongoing operations.
Poor resource assignment is a leading cause of AI pilot failure, as projects stall when staff are spread too thin or lack the necessary expertise. Leaders should explicitly name resource owners and protect their capacity for AI project work.
Example: The district assigns a dedicated half-time project coordinator and $40,000 in discretionary budget to the AI tutoring pilot, ensuring the initiative has a clear owner and sufficient resources to reach evaluation milestones.
Resource Disparity¶
The unequal distribution of financial, human, technological, and infrastructural resources across schools and districts, resulting in systematically different educational experiences and outcomes based on students' zip codes or demographic characteristics.
Resource disparity is the foundational equity challenge in American education, and AI has the potential to either narrow or widen it depending on how it is deployed and governed. Intentional equity strategies are essential.
Example: AI-generated curriculum that makes professional-quality materials universally available could address part of the resource disparity between a wealthy suburban district with a 12-person curriculum team and a rural district with none.
Responsible AI¶
The practice of designing, deploying, and governing AI systems in ways that are safe, fair, transparent, privacy-preserving, and aligned with human values and the well-being of all affected individuals and communities.
Responsible AI is not a checklist but an ongoing organizational practice. Education leaders bear special responsibility because AI in schools affects children, whose vulnerability and developmental needs warrant heightened ethical care.
Example: A district's responsible AI framework requires that every AI tool used with students be evaluated for bias, data privacy compliance, and age-appropriateness before adoption, with annual re-review.
Retrieval Augmented Generation¶
A technique in which an AI system searches a specific knowledge base or document collection to find relevant information before generating a response, grounding outputs in provided sources rather than relying solely on training data.
Retrieval augmented generation allows AI tools to be grounded in a district's specific curriculum, policies, or student data. This significantly reduces hallucination risk and makes AI outputs more relevant and verifiable.
Example: A district deploys an AI assistant that uses retrieval augmented generation to answer parent questions by searching the official parent handbook, ensuring all answers cite specific district policies.
Return On Investment¶
A measure of the financial or educational value gained from an initiative relative to its cost, used to evaluate whether an investment is justified and to prioritize among competing options.
Calculating return on investment for AI in education requires defining both financial savings and educational outcome improvements. Leaders who cannot articulate AI's return on investment will struggle to sustain budget commitments.
Example: A district calculates that an AI grading assistant saves each teacher four hours per week, equivalent to $1,200 per teacher per year in reclaimed instructional time, easily justifying the $300 annual per-teacher license cost.
Risk Register¶
A documented inventory of identified risks associated with an AI initiative or portfolio, including descriptions of each risk, its likelihood and potential impact, the planned mitigation strategy, and the responsible owner.
A risk register makes AI-related risks visible and manageable rather than implicit and ignored. Regularly reviewing and updating the register ensures that new risks are captured and that mitigation actions are actually being taken.
Example: The district's AI risk register includes data privacy breach risk (high impact, medium likelihood), AI hallucination in student-facing tools (medium impact, high likelihood), and teacher resistance to adoption (medium impact, medium likelihood).
Risk Reward Tradeoff¶
The analysis of the relationship between the potential negative consequences of an AI initiative and its expected benefits, used to determine whether the opportunity justifies the risk and what mitigations are required.
Framing AI decisions as risk-reward tradeoffs enables more balanced and defensible choices than either uncritical enthusiasm or reflexive caution. Both inaction and action carry risks; the question is which risk profile is more acceptable.
Example: A district's risk-reward analysis of deploying AI-assisted grading concludes that the time savings (reward) justify the accuracy risk (risk) if a human review step is maintained for all final grades.
Risk Scoring¶
A quantitative or qualitative rating of the potential negative consequences associated with an AI use case, considering factors such as data privacy exposure, accuracy requirements, equity impacts, and stakeholder sensitivity.
Explicit risk scoring ensures that high-risk AI applications receive appropriate scrutiny before approval, and that mitigation plans are built into project designs from the beginning.
Example: An AI proposal to analyze student social media activity receives a high risk score due to privacy concerns, triggering a requirement for legal review and parent notification planning before any pilot can proceed.
Scaling Strategy¶
A plan for expanding a successful AI pilot initiative to reach a broader population of students, schools, or use cases in a sustainable and equitable way, addressing the technical, financial, human, and governance requirements of growth.
Without a deliberate scaling strategy, successful pilots remain isolated experiments that never deliver their full potential value. Scaling introduces new challenges of consistency, quality, and equity that must be planned for in advance.
Example: Following a successful one-school AI writing coach pilot, the district's scaling strategy specifies a two-year expansion to all middle schools, with a per-school implementation checklist, a dedicated trainer, and a central monitoring dashboard.
School Board Engagement¶
The process of informing, consulting, and involving elected school board members in AI strategy decisions, ensuring they have the knowledge and context to provide effective oversight and make informed policy decisions.
School board members are accountable to the community for all district decisions, including AI. Proactive engagement prevents board members from being surprised by AI developments and builds the political support needed for sustained investment.
Example: A superintendent presents a quarterly AI strategy update at school board meetings, including progress on pilots, emerging risks, and upcoming decisions that require board authorization, building informed oversight over time.
Scoring Rubric¶
A structured evaluation tool that defines specific criteria and performance levels for assessing AI proposals or implementations, enabling consistent, transparent, and comparable judgments across reviewers.
A shared scoring rubric reduces evaluator bias and makes the idea selection process legible to all stakeholders. It also communicates to idea submitters what the organization values in AI proposals.
Example: The district's AI idea scoring rubric assigns points across five categories: educational impact, equity benefit, technical feasibility, data privacy risk, and alignment with strategic priorities.
Screen Time Concerns¶
Parental, medical, and educational concerns about the potential negative impacts of excessive time spent interacting with digital screens on children's development, sleep, physical activity, social skills, and mental health.
AI in education increases digital engagement, making screen time concerns more salient. Leaders must design AI implementations that respect recommended limits and balance screen-based learning with physical and social activities.
Example: A district's AI implementation guidelines specify that elementary students should not exceed 90 minutes of AI-assisted screen-based learning per day, with regular movement and offline activity breaks.
Self Paced Learning¶
An educational approach in which individual learners control the speed at which they progress through material, spending more time on difficult concepts and accelerating through areas of existing strength.
Self-paced learning, when combined with AI personalization, can accommodate the full range of student readiness within a single classroom. It requires strong monitoring systems to ensure students do not disengage or stall without teacher awareness.
Example: In a self-paced learning environment, an advanced student completes a semester's algebra curriculum in three months while her classmate takes the full semester, with AI tracking both students' progress continuously.
Shared Code Repository¶
A centralized, version-controlled digital location where technical team members collaboratively store, review, and manage the software code, configurations, and scripts associated with AI development or customization projects.
A shared code repository enables multiple technical staff to collaborate without overwriting each other's work, tracks the history of all changes, and makes AI tool configurations auditable and reproducible.
Example: The district's AI development team stores all custom prompt configurations and API integration scripts in a shared code repository, allowing any team member to review, modify, or restore previous versions.
Skill Atrophy¶
The gradual decline in a human ability caused by disuse, often precipitated by AI tools taking over tasks that previously required regular practice of that skill.
Skill atrophy is a subtle but serious risk of AI integration in education. Leaders must identify which skills require deliberate practice to maintain and design curricula that ensure AI augments rather than replaces essential human capabilities.
Example: Research on skill atrophy suggests that students who consistently use AI grammar checkers may lose the ability to self-correct writing errors, prompting a district to require periodic unplugged writing assessments.
Skill Development¶
The intentional process of building students' practical capabilities—including both academic competencies and life skills such as communication, collaboration, persistence, and critical thinking—through structured practice and experience.
As AI automates knowledge retrieval and application, skill development in areas that AI cannot replicate—such as empathy, creativity, ethical reasoning, and interpersonal communication—becomes the core purpose of human-centered education.
Example: A high school embeds explicit skill development in its AI-integrated curriculum by requiring students to articulate, practice, and reflect on communication and collaboration skills in every project-based learning unit.
Socratic Method¶
A teaching technique that uses guided questioning to stimulate critical thinking and draw out students' own reasoning and understanding, rather than presenting information directly, challenging students to examine their assumptions and evidence.
The Socratic method develops the reasoning and self-reflection skills that AI cannot replicate or replace. Its value increases as AI handles routine knowledge transmission, making space for deeper human-led inquiry.
Example: Instead of explaining why the American Revolution occurred, a teacher asks a series of increasingly probing questions that lead students to construct their own causal analysis from primary source evidence.
Solution Taxonomy¶
A structured classification system that categorizes AI solutions by their technical approach, data requirements, interaction mode, and deployment model, used to organize and compare options across the AI idea portfolio.
A solution taxonomy helps evaluators recognize when different teams are proposing similar technical approaches to different problems, enabling reuse of components and avoiding duplication of evaluation effort.
Example: The district's solution taxonomy reveals that five separate AI proposals all involve chatbot interfaces, prompting a decision to evaluate a single flexible chatbot platform rather than five separate vendor tools.
Stakeholder Alignment¶
The process of engaging all parties affected by an AI initiative—including teachers, parents, students, administrators, and board members—so that they share a common understanding of goals, risks, and expectations.
Unaligned stakeholders can derail even technically successful AI initiatives. Proactive communication and co-design processes build the trust and shared ownership necessary for sustainable AI adoption in schools.
Example: Before piloting an AI attendance monitoring tool, a district holds listening sessions with parent groups and the teachers' union to surface concerns and incorporate their input into the implementation plan.
Strategic Bets¶
High-uncertainty AI investments with the potential for significant long-term impact, justified by their alignment with organizational mission and the magnitude of the opportunity rather than short-term certainty of return.
Balancing strategic bets with quick wins creates a healthy AI portfolio. Organizations that only pursue quick wins may miss transformative opportunities, while those that only make strategic bets risk burning resources without tangible results.
Example: Investing in a multi-year personalized learning platform powered by AI is a strategic bet: the potential to dramatically improve student outcomes justifies the risk despite uncertainty about implementation challenges.
Strategic Planning¶
A formal process in which an organization examines its current state, defines desired future outcomes, identifies the actions needed to close the gap, and allocates resources accordingly over a defined time horizon.
Strategic planning applied to AI adoption prevents reactive, technology-driven decisions and ensures AI investments align with educational mission and equity goals. It also creates accountability structures for measuring whether AI is delivering value.
Example: A school board engages in strategic planning to determine which AI investments to fund over the next three years, starting with a SWOT analysis of the district's current technology capabilities.
Strategic Roadmap¶
A high-level, time-phased visual plan that communicates an organization's intended trajectory toward its AI strategic goals, showing key milestones, initiatives, and decision points across a multi-year horizon.
A strategic roadmap provides the shared narrative that aligns stakeholders behind a common vision and sequence. It communicates priority and pacing without prescribing every implementation detail, preserving flexibility as conditions evolve.
Example: A district's three-year AI strategic roadmap shows Year 1 focused on infrastructure and literacy, Year 2 on targeted pilots, and Year 3 on scaling proven tools and measuring system-wide impact.
Strategic Urgency¶
The condition in which the pace of external change—such as rapid AI capability growth and adoption by peers—creates meaningful risk for an organization that delays action, warranting prioritization of AI planning now rather than later.
Strategic urgency does not mean rushing without planning; it means recognizing that doing nothing carries its own risks. Districts that delay AI strategy development may find themselves significantly behind in student outcomes and teacher capacity.
Example: A superintendent presents data on competitor districts' AI adoption rates to her school board, arguing that strategic urgency justifies accelerating the district's AI readiness timeline by one year.
Strengths Analysis¶
The component of a SWOT analysis that identifies the internal capabilities, resources, and advantages an organization already possesses that can be leveraged to support successful AI strategy implementation.
Understanding existing strengths prevents organizations from duplicating investments they have already made and identifies where AI can amplify existing advantages rather than compensating for weaknesses.
Example: A district strengths analysis reveals that it already has a mature learning management system with strong data infrastructure, making integration of AI analytics tools faster and less expensive than for peer districts.
Student Data Ownership¶
The principle that data generated by students' learning activities belongs to the students and their families—or to the institution acting in their interest—not to the vendors whose tools collected it.
Student data ownership has significant implications for vendor contracts, privacy policies, and the governance of AI systems that learn from student behavior. It should be a foundational principle in every district's AI governance framework.
Example: A district's AI governance policy states that all student learning records are owned by the district on behalf of students and may not be used by vendors for model training or product improvement without explicit consent.
Student Data Protection¶
The collection of legal, contractual, technical, and organizational safeguards designed to prevent unauthorized access to, misuse of, or harm from data generated by students in the course of their education.
Student data protection goes beyond legal compliance to encompass the ethical stewardship of information about children. Districts should treat student data as a liability to be minimized, not an asset to be maximized.
Example: The district's student data protection policy limits AI tools to collecting only the minimum data necessary for their educational purpose and requires automatic deletion of all student data within one year of enrollment end.
Student Well Being¶
The holistic state of a student's physical, emotional, social, and psychological health and happiness, considered alongside academic achievement as a core outcome of education.
AI integration must be evaluated for its impact on student well-being, not just academic performance. Tools that improve test scores while increasing anxiety or social isolation are not net positive for students.
Example: A district's AI tool evaluation process includes a student well-being impact assessment, including surveys of students about their stress levels and sense of connection when using AI-intensive learning programs.
Success Metrics¶
Predefined, measurable indicators used to assess whether an AI initiative has achieved its intended outcomes, agreed upon before implementation begins to prevent post-hoc rationalization of results.
Defining success metrics before deployment prevents moving goalposts and ensures that AI projects are accountable to educational outcomes rather than technology adoption rates. Metrics should be linked directly to the original problem statement.
Example: The success metrics for an AI writing coach pilot include a 15% improvement in student essay scores on the district rubric and a 20% reduction in teacher time spent on written feedback, measured at semester end.
SWOT Analysis¶
A strategic planning tool that systematically examines an organization's internal Strengths and Weaknesses alongside external Opportunities and Threats, providing a structured basis for strategic decision-making.
SWOT analysis applied to AI strategy helps education leaders understand their starting position honestly and identify where to focus resources. It prevents both overconfident adoption and unnecessarily cautious inaction.
Example: A district's AI SWOT analysis identifies strong data infrastructure as a strength, limited technical staff as a weakness, rapidly improving AI tutoring quality as an opportunity, and student data privacy regulations as a threat.
Task Horizon¶
A measure of the maximum duration or complexity of tasks that an AI agent can complete reliably without human assistance, used to characterize the practical autonomy of AI systems.
As task horizons extend, AI systems can take on increasingly complex educational workflows such as multi-step tutoring sequences or curriculum development projects. Monitoring task horizon growth helps leaders plan future automation opportunities.
Example: An AI agent with a one-hour task horizon can independently complete a full diagnostic assessment session with a student and compile a summary report without human check-ins.
Teacher Role Shift¶
The transformation of the teacher's primary function from content deliverer and assessor to learning designer, coach, mentor, and facilitator of human development, as AI systems increasingly handle routine instructional and administrative tasks.
The teacher role shift is one of the most significant and sensitive implications of AI in education. Leaders must invest in professional development, role redesign, and cultural change to support teachers through this transition effectively.
Example: As AI handles initial math concept explanations and practice feedback, a teacher shifts her time toward small-group problem-solving discussions, student goal-setting conferences, and creative project mentorship.
Team Based Learning¶
An active learning instructional strategy in which small student teams work through structured sequences of individual preparation, group application, and public accountability, developing both content mastery and collaboration skills.
Team-based learning counterbalances potential isolation from AI-personalized instruction by requiring students to negotiate, communicate, and build shared understanding. It develops interpersonal competencies that remain distinctly human.
Example: After individual AI tutoring on cell biology, students form teams of four to diagnose a fictional patient's illness using symptoms and lab results, applying their knowledge to a shared challenge.
Ten Thousand Textbooks¶
A conceptual scenario illustrating the potential of AI to generate a massive library of high-quality, customized educational materials at near-zero marginal cost, making highly personalized curriculum universally accessible.
The ten-thousand-textbooks concept challenges the assumption that curriculum must be scarce and expensive. It invites education leaders to reimagine procurement, quality assurance, and teacher support in a world of content abundance.
Example: In the ten-thousand-textbooks scenario, a middle school math teacher selects from hundreds of AI-generated unit options tailored to her students' specific prior knowledge and cultural backgrounds.
Term Planning Agent¶
An AI agent that assists teachers or administrators in creating, organizing, and adjusting academic term plans—including scope and sequence, resource allocation, and assessment scheduling—based on curriculum standards and institutional calendars.
Term planning agents can reduce the significant time teachers spend on planning logistics, freeing them to focus on the instructional and relational aspects of their work. Human review of agent-generated plans remains essential.
Example: A term planning agent analyzes a teacher's curriculum map, the school calendar, and pacing data from the previous year to generate a suggested week-by-week plan for the spring semester, which the teacher then reviews and modifies.
Text Generation¶
The AI capability to produce coherent, contextually appropriate written content—such as lesson plans, feedback, summaries, or assessments—based on a given prompt or set of instructions.
Text generation is one of the most immediately applicable AI capabilities for education, reducing the time teachers spend on routine writing tasks and enabling rapid production of differentiated instructional materials.
Example: A teacher uses AI text generation to produce five versions of the same history reading, each written at a different reading level, in under five minutes.
Textbook Procurement¶
The institutional process of selecting, licensing, funding, and distributing educational materials to students and teachers, typically governed by state adoption cycles, budget constraints, and curriculum standards alignment requirements.
AI-driven content cost collapse is disrupting traditional textbook procurement cycles. Districts that understand this shift can negotiate better vendor terms, reduce dependency on six-year adoption cycles, and accelerate curriculum updates.
Example: A state textbook procurement committee adds AI-generated content quality standards to its evaluation rubric after several districts submit AI-authored materials for state adoption consideration.
Threats Analysis¶
The component of a SWOT analysis that identifies external risks, competitive pressures, regulatory changes, or capability advances that could undermine an organization's AI strategy or expose it to harm if not addressed.
Threats analysis surfaces risks that leaders might prefer not to acknowledge, from competitive disadvantage if peers adopt AI faster to legal liability if privacy regulations tighten. Early identification enables proactive mitigation.
Example: A threats analysis flags that a state legislature is considering new AI disclosure requirements for student-facing tools, prompting the district to begin vendor contract renegotiations in advance of possible regulatory change.
Title I Schools¶
Schools in the United States that receive federal funding under Title I of the Elementary and Secondary Education Act because they serve a high proportion of students from low-income families, entitling them to additional resources for improving academic achievement.
Title I schools face both the greatest potential benefit and the greatest access barriers to AI-enhanced education. Leaders should prioritize AI pilots and investments in Title I schools and use Title I funding where permissible to support AI adoption.
Example: A state education agency provides targeted AI literacy training grants to Title I schools, recognizing that without proactive investment, the AI capability gap between wealthy and under-resourced schools will widen.
Token¶
A unit of text—roughly equivalent to a word or part of a word—used by AI language models to process and generate language, forming the basic building blocks of model input and output.
AI usage is often priced by the number of tokens processed, making token awareness important for budgeting at scale. It also helps educators understand why very long documents may be truncated or summarized.
Example: The sentence "The student passed the test" contains approximately six tokens, each of which the model processes individually when generating a reply.
Total Cost Of Ownership¶
The comprehensive long-term cost of an AI system including not only licensing fees but also implementation, training, integration, maintenance, data storage, support, and eventual replacement expenses.
Focusing only on license cost when evaluating AI tools leads to budget surprises. A full total-cost-of-ownership analysis ensures that the true financial commitment is understood before making a procurement commitment.
Example: An AI grading tool quoted at $10 per student per year has a total cost of ownership of $25 per student when implementation consulting, staff training, and IT integration are included.
Training Data¶
The collection of examples, texts, images, or other information that an AI system processes during development to learn patterns and build its internal knowledge and capabilities.
The quality, diversity, and recency of training data directly affect AI reliability and potential bias. Education leaders should ask vendors about the sources and composition of training data for any tool they adopt.
Example: An AI essay-feedback tool was trained on millions of student essays rated by expert teachers, so its feedback reflects the values embedded in those ratings.
Transparency¶
The quality of an AI system or process being open about how it works, what data it uses, what decisions it makes, and what its limitations are, enabling users and overseers to understand and appropriately trust or question its outputs.
Transparency is essential for AI governance in education because parents, students, and teachers have a right to understand how AI systems are influencing learning decisions. Opaque AI systems undermine trust and make accountability impossible.
Example: A district requires all AI vendors to provide a plain-language transparency statement explaining what data their tool collects, how the AI makes recommendations, and what its known error rates are.
Two Hour Learning¶
A pedagogical concept in which AI-powered individualized instruction enables students to master the essential academic content of a school day in approximately two hours, freeing the remaining time for other forms of development.
Two-hour learning challenges the assumption that academic instruction must fill the majority of the school day. If validated at scale, it would require fundamental rethinking of teacher roles, schedules, and definitions of school success.
Example: A pilot program testing two-hour learning reports that students using AI-personalized instruction achieve equivalent academic gains to peers in traditional six-hour academic schedules.
Under Resourced Schools¶
Schools that lack adequate funding, staffing, facilities, technology infrastructure, or community support to provide all students with a high-quality education comparable to that available in better-funded peer institutions.
Under-resourced schools have the most to gain from AI tools that can supplement limited human capacity, but they also have the least capacity to evaluate, procure, and implement those tools responsibly. Targeted support structures are essential.
Example: An under-resourced school uses an AI lesson-planning assistant to compensate for its lack of a dedicated curriculum coordinator, enabling teachers to access standards-aligned materials without central support.
Use Case Identification¶
The systematic process of defining specific, bounded applications of AI within an organization that address real problems, are technically feasible, and align with institutional priorities.
Effective use case identification prevents AI pilots from being technology-first rather than problem-first. A structured process ensures limited resources are applied where AI can create the greatest educational value.
Example: Through staff surveys and interviews, a district identifies "drafting individualized education program goals" as a high-priority AI use case because it consumes significant special education coordinator time.
Vendor Lock In¶
A situation in which an organization becomes excessively dependent on a single technology vendor due to proprietary data formats, switching costs, or contractual terms, reducing the ability to change providers or negotiate effectively.
Vendor lock-in is a significant risk in educational AI procurement because student learning records and curriculum investments can become trapped in proprietary systems. Leaders should require data portability and open standards in all contracts.
Example: A district discovers it cannot migrate years of student learning records from its AI tutoring platform because the vendor uses a proprietary data format not exportable to any other system.
Vendor Selection¶
The process of evaluating and choosing among competing external providers of AI products or services based on defined criteria including capability, cost, data privacy practices, interoperability, and support.
Rigorous vendor selection protects districts from costly mistakes, data privacy violations, and vendor lock-in. Education leaders should use standardized rubrics and involve privacy, curriculum, and technology staff in every vendor evaluation.
Example: A district evaluates four AI tutoring vendors using a scoring rubric that includes FERPA compliance, student data deletion policies, evidence of efficacy, and total cost of ownership.
Weaknesses Analysis¶
The component of a SWOT analysis that identifies internal deficiencies, gaps, or constraints within an organization that may impede effective AI strategy implementation and must be addressed or mitigated.
Honest identification of weaknesses is essential for realistic AI planning. Organizations that ignore their weaknesses build AI strategies on false assumptions and are repeatedly surprised by preventable failures.
Example: A district weaknesses analysis reveals that only 10% of teachers have completed any AI professional development, identifying a critical capability gap that must be addressed before tool deployment can succeed.
World Model¶
An AI system's internal representation of how the world works, built from training data, which it uses to make predictions, draw inferences, and generate contextually appropriate responses.
A richer world model enables AI to provide more nuanced and contextually accurate educational support. However, if the world model contains biases or gaps from training data, those flaws will appear in the AI's outputs.
Example: An AI geography tutor draws on its world model to correctly explain that landlocked countries face specific trade challenges, without having been explicitly taught that particular fact.
xAPI Standard¶
An open technical specification for recording and sharing learning data across different educational systems, enabling interoperable learning records that follow a standardized "actor-verb-object" structure.
xAPI adoption allows districts to consolidate learning data from multiple tools into a single learning record store, enabling richer analytics than any single platform can provide. It is a key enabler of data portability and vendor independence.
Example: When a student watches a video in one platform and completes a quiz in another, both tools send xAPI statements to the district's learning record store, creating a unified picture of the student's activity.