25 ChatGPT-5.5 Prompts for Role-Based AI Learning Plans: Employees, Developers, Leaders, Educators, and Students

Why role-based AI learning plans need more than a list of courses

OpenAI’s Academy expansion, announced on September 21, 2026, organizes AI learning around four role-based paths: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI. That structure is useful because a customer-support analyst, a platform engineer, a business-unit sponsor, a teacher, and a student do not need the same learning plan. They need different practice tasks, different review checkpoints, different evidence of progress, and different boundaries on what AI-generated work may be used for.

This prompts masterclass is designed to turn those four Academy paths into practical ChatGPT-5.5 learning-plan prompts for five audiences: employees, developers, leaders, educators, and students. Each prompt is meant to help a learner or program owner create a safer, more measurable plan without pretending that a course, badge, or chatbot response is a substitute for expertise, authorization, or governance. The prompts in later sections will ask ChatGPT to work from permitted materials only, identify missing information, preserve human decision-making, and create verification checkpoints tied to real tasks.

OpenAI’s Academy announcement says the portfolio snapshot included 3 courses under Apply AI at Work, 8 under Build with AI, 1 under Lead AI Adoption, and 2 under Teach and Learn with AI. Treat those numbers as a dated announcement snapshot, not as a permanent catalog guarantee. Academy courses can change as models, products, and guidance evolve, and availability or reporting can vary by organization, account configuration, rollout, region, and workspace policy. A learning plan should therefore reference the current catalog before assigning courses, and it should avoid promising badge availability, completion timelines, or reporting access unless the organization has verified those details through its own Academy environment or account team.

This article provides ChatGPT-5.5 prompt templates for AI-native company workflows, including onboarding, account intelligence, and developer integrations. The 25 ChatGPT-5.5 Prompts for AI-Native Company Workflows: Onboarding, Account Intelligence, and Developer Integrations article is a focused companion for Employee AI Onboarding because it directly supports the employee onboarding marker by expanding from role-based learning plans into practical AI-native onboarding workflows.

The most important distinction for this article is the difference between participation and applied capability. OpenAI’s Champion deployment guide distinguishes course participation and completion signals from examples of application. Completion shows that a learner engaged with course material and passed an assessment where applicable; it does not prove that the learner can safely deploy an agent, publish a policy, grade student work, ship code, run a hiring workflow, or make a regulated decision. Applied capability requires observable work samples, role-appropriate review, policy alignment, and evidence that the learner can recognize uncertainty, check sources, and escalate consequential decisions to an authorized human.

The four OpenAI Academy paths and what each one should change in a learning plan

The Academy paths should not be treated as generic “AI literacy” labels. Each path implies a different set of tasks, risks, and evidence. The prompts in this article use those distinctions to help learners move from course selection to practice design, review, and follow-up support.

Academy path	Primary audience	What OpenAI says it covers	Learning-plan implication
Apply AI at Work	Knowledge workers and employees using AI in everyday work	Clear instructions, context, reviewing responses, reusable workflows, agent delegation, checkpoints, and human review	The plan should focus on permitted business tasks, output review, reusable prompt patterns, escalation rules, and evidence that the learner can check results before using them.
Build with AI	Developers, Codex users, and technical teams building with the OpenAI API	Software-development planning and implementation, review and quality, solution design, evaluations, agents, retrieval, and production operations	The plan should include technical design review, evaluation criteria, test data boundaries, security review, retrieval quality checks, and human approval before deployment or permission changes.
Lead AI Adoption	Executives, managers, sponsors, and adoption leaders	Business value, priorities, ownership, governance, strategy, and a roadmap	The plan should connect learning to a governed initiative, not a vague productivity claim. It should define ownership, risk controls, review cadence, and evidence beyond usage metrics.
Teach and Learn with AI	Educators and students	Review against learning objectives, source material, assignment requirements, permitted materials, study, teaching, and career preparation	The plan should respect academic integrity, assignment rules, accessibility needs, and final human responsibility for submitted work, teaching materials, and learning outcomes.

For employees, the practical learning goal is not “use ChatGPT more.” It is to choose appropriate work where AI assistance is allowed, provide context without exposing sensitive information, inspect responses, reuse effective workflows, and know when not to delegate. A safe employee plan should include examples such as drafting a non-confidential meeting summary from approved notes, transforming a public policy into a checklist, or preparing questions for a manager review. It should exclude private employee records, credentials, medical records, legal advice files, undisclosed financial data, and customer secrets unless the organization has explicitly approved the environment and use case.

For developers, the learning goal is not “generate more code.” OpenAI describes Build with AI as covering the software lifecycle, including evaluations, agents, retrieval, and production operations. A developer learning plan should therefore include requirements decomposition, test planning, code review, security review, evaluation design, failure analysis, rollback criteria, and operations readiness. A course badge or completed practice exercise should not be treated as permission to connect production systems, change access controls, deploy agents, process confidential source code, or bypass established engineering review.

For leaders, the learning goal is not “announce an AI strategy.” Lead AI Adoption should be converted into a roadmap that identifies a business priority, accountable owners, governance checkpoints, review responsibilities, and evidence standards. The Champion deployment guide recommends sponsor and manager reinforcement, office hours or application sessions, protected learning time, and an 8–12 week deployment summary. A leader’s prompt should ask ChatGPT to separate access, participation, completion, application evidence, adoption signals, quality outcomes, and business results rather than collapsing everything into a single “AI adoption” number.

For educators and students, the learning goal is not “finish assignments faster.” OpenAI’s education path emphasizes review against learning objectives, source material, and assignment requirements, and the source notes make clear that educators and students retain responsibility for the final result. A student may use AI to practice a concept, generate self-quiz questions from permitted class notes, or plan a study schedule. A student should not use it to obtain protected assessment answers, impersonate original work, conceal AI use when disclosure is required, or submit generated content that violates course rules. Educators likewise need prompts that preserve curriculum goals, accessibility, age-appropriate practice, and institutional academic-integrity policies.

The badge boundary: useful learning evidence, not authorization

OpenAI says Academy course assessments let learners demonstrate what they learned and that passing a course assessment earns an OpenAI Academy course badge. That is a learning credential for a course assessment. It should not be described as a professional license, industry certification, safety certification, authorization to deploy AI systems, proof that an organization’s systems are safe, or evidence that a learner can independently perform consequential work without review.

This distinction matters because badges can easily be misused inside organizations. A manager might be tempted to treat badge status as permission to give an employee confidential data, approve code deployment, publish public content, automate customer responses, or make a personnel decision. That is not an appropriate inference from the Academy materials. Completion and badge status can be one input into a broader learning record, but authorization should remain tied to job role, policy, access controls, technical competence, risk level, and human review.

A practical learning plan should therefore record at least three different kinds of evidence. First, it can record participation evidence, such as enrollment, attendance, course completion, and badge status where available. Second, it can record application evidence, such as a reviewed prompt workflow, a tested evaluation set, a manager-approved work sample, or an educator-approved lesson activity. Third, it can record outcome evidence, such as reduced rework on a specific process, better documentation quality, fewer review defects, or improved student study habits. The third category needs special caution because usage changes and outcome changes do not prove that the course caused the change.

Editorial rule for this prompt set: every prompt treats Academy completion and badges as learning evidence only. No prompt should use a badge as a basis for hiring, firing, promotion, grading, disciplinary action, production deployment, legal approval, medical advice, financial action, payment, publication, or any other consequential decision.

For enterprise administrators and program owners, the safest operational approach is to pair course completion with a role-specific verification artifact. An employee might submit a redacted workflow showing the task, approved inputs, prompt, review steps, and final human decision. A developer might submit an evaluation plan and a non-sensitive test result summary reviewed by a senior engineer. A leader might submit a roadmap with owners, risks, and a measurement plan. An educator might submit a lesson plan aligned with policy and learning objectives. A student might submit a study plan, reflection, or practice log that follows course rules.

Permitted-input rules every learning prompt should enforce

The prompts in this article are written for real workplaces and schools, so they must prevent a common failure mode: learners pasting the wrong material into ChatGPT. A good learning prompt does not merely say “help me make a plan.” It tells ChatGPT to use only redacted, synthetic, public, or organization-approved inputs and to refuse unnecessary sensitive data. This is especially important in learning contexts because learners may not yet know which data is safe to use.

Use the following permitted-input rule as the foundation for every learning plan in this article: provide only material that the learner is authorized to use in the chosen AI environment. Appropriate inputs may include public policy excerpts, synthetic scenarios, redacted process notes, organization-approved templates, non-confidential learning goals, or course requirements that permit AI assistance. Inappropriate inputs include passwords, tokens, account numbers, private student or employee records, medical records, legal matter files, financial records, confidential source code, protected assessment answers, undisclosed customer data, trade secrets, privileged communications, and security-sensitive details.

This rule applies even when the goal is benign. A manager building an employee coaching plan should not paste private performance notes unless the organization has explicitly approved that use. A developer should not paste proprietary production credentials, access tokens, or confidential code into a learning prompt. A teacher should not paste identifiable student records when a synthetic class profile would work. A student should not paste protected exam content or ask ChatGPT to answer an assessment that rules prohibit. A legal-technology professional should avoid privileged or client-identifying content unless the system, policy, engagement terms, and supervising attorney permit that use.

When the learner is uncertain, the prompt should ask ChatGPT to stop and request a safer substitute. For example, the model can ask for a synthetic version of the scenario, a redacted excerpt, a public policy, or a high-level description. This is not only a privacy measure; it improves learning quality because the learner practices converting real work into safe, reusable patterns without exposing unnecessary confidential content.

This guide explains advanced prompt engineering frameworks for 2026, including RTF, CREATE, Chain-of-Thought, ReAct, and DSPy for ChatGPT, Claude, and Codex. The Advanced Prompt Engineering Frameworks for 2026: RTF, CREATE, Chain-of-Thought, ReAct, and DSPy article is a focused companion for Prompt Verification Frameworks because it is the strongest match for a marker about prompt frameworks because it gives readers structured methods they can use to design and validate better prompts.

Human responsibility and consequential-action limits

OpenAI’s Academy materials frame AI learning as practice with review, not as a transfer of responsibility from people to models. The prompts in this masterclass must therefore keep final decisions with the authorized human. ChatGPT can help draft a study plan, structure a workflow, produce a rubric, or identify missing evidence, but it should not make final decisions about employment, grading, discipline, clinical care, legal advice, tax positions, investments, payments, purchases, system access, code deployment, publication, or public communications.

This boundary is especially important for agent delegation. Apply AI at Work includes agent delegation, checkpoints, and human review, and Build with AI includes agents and production operations. Those topics can be valuable learning areas, but they also increase operational risk. A learning prompt may ask ChatGPT to design a checkpoint plan for an agent-assisted workflow, but it should not instruct the model to send messages, submit forms, purchase services, change permissions, deploy code, or publish outputs without human approval. In a learning context, the safest default is to generate plans, checklists, dry-run outputs, and review artifacts rather than executing external actions.

For leaders and administrators, human responsibility also means that AI-generated training recommendations should be reviewed for fairness, accessibility, and role fit. A prompt can propose which Academy path may fit a role, but it should not be used to deny opportunity, assign remedial status, or infer competence from incomplete data. When training decisions affect employment, workload, evaluation, promotion, or access, use established human resources, legal, accessibility, and employee-relations processes rather than relying on a model-generated recommendation.

For educators and students, final responsibility includes academic integrity. A prompt can help a student create a study plan, explain a concept using permitted materials, or generate practice questions. It should not produce work that is submitted as the student’s own if that violates the assignment rules. A prompt can help an educator adapt lesson materials for accessibility or create discussion questions, but the educator remains responsible for accuracy, appropriateness, student privacy, and alignment with curriculum and institutional policy.

Uncertainty, missing inputs, accessibility, and integrity are not optional add-ons

Every prompt in this article should require ChatGPT to identify uncertainty and missing inputs. This is a practical safeguard, not a stylistic preference. A learning plan built from incomplete information may assign the wrong path, select inaccessible practice tasks, overlook policy restrictions, or treat a low-risk exercise as if it were ready for production. When ChatGPT lacks the learner’s role, policy constraints, permitted tools, available time, accessibility needs, or review process, it should say so and propose a conservative next step.

Accessibility should be part of the initial plan rather than a late remediation step. A useful learning prompt can ask for multiple formats, such as text summaries, checklists, audio-friendly scripts, keyboard-accessible practice flows, plain-language explanations, or additional time options. For workplace programs, accessibility also includes manager support, protected learning time, and alternatives for employees whose roles, schedules, or assistive technologies make a standard course sequence difficult. For classrooms, accessibility includes accommodations required by institutional policy and the need to avoid exposing private student information.

Academic integrity should also be explicit from the start. Students should be asked to include assignment rules, permitted materials, citation expectations, AI-disclosure requirements, and the difference between practice and submission. Educators should be asked to define what AI assistance is allowed, what must be original student work, how students should disclose use, and how learning objectives will be assessed. If those rules are missing, the prompt should not assume permissive use; it should ask the learner to consult the instructor, syllabus, institution, or program policy.

Security and privacy policies are equally central for developers and enterprise teams. A Build with AI learning plan should include non-sensitive examples, synthetic data, evaluation criteria, secure coding review, access-control review, and deployment gates. It should not treat a local prototype or course exercise as production-ready. A plan that includes retrieval, agents, or API integration should name the review steps needed before connecting real data sources, external tools, or production workflows.

A practical workflow for using the 25 prompts safely

The 25 prompts that follow are intended to be used as a workflow, not as isolated tricks. Start by establishing the learner’s role, permitted inputs, goals, available time, and policy constraints. Then select the Academy path or combination of paths that fits the learner. After that, design practice tasks using redacted, synthetic, public, or approved materials. Finally, define a human verification checkpoint that produces application evidence without turning the model into the final decision-maker.

1. Define the role and context. Identify whether the learner is an employee, developer, leader, educator, student, or a hybrid role such as a technical manager or teaching assistant.
2. Confirm permitted inputs. Use only public, synthetic, redacted, or organization-approved content, and exclude credentials, private records, confidential code, protected assessment answers, and regulated or privileged material.
3. Select the Academy path. Map the learner to Apply AI at Work, Build with AI, Lead AI Adoption, Teach and Learn with AI, or a carefully scoped combination.
4. Create role-specific practice tasks. Tie exercises to real but permitted work, such as a reviewed workflow, a non-sensitive test plan, a roadmap draft, a lesson activity, or a study schedule.
5. Add checkpoints. Require human review before external messages, submissions, grading, deployment, payment, publication, permission changes, or other consequential actions.
6. Capture evidence. Separate participation evidence, badge or assessment evidence, application examples, quality signals, and business or learning outcomes.
7. Reassess and revise. Treat the learning plan as a living document because OpenAI says Academy courses will continue changing as models, products, and guidance evolve.

The Champion deployment guide’s rollout stages—Activate, Engage sponsors, Launch, Reinforce and measure, and Share—are useful for organizations using these prompts at scale. A team can activate by defining audience groups and approved learning materials, engage sponsors by clarifying why each path matters, launch with role-specific starting points, reinforce through office hours and manager check-ins, and share an 8–12 week deployment summary. That summary should include participation and completion data where available, but it should also include applied examples and limitations in the data.

For individual learners, the same logic can be smaller. A student can use a prompt to create a two-week study plan with permitted materials and a self-check rubric. A developer can create a one-sprint learning plan for evaluation design using synthetic test cases. A manager can create a coaching plan for a team using approved workflows and office-hour topics. A teacher can create a lesson-planning checklist that preserves student privacy and academic-integrity rules. The workflow scales because it asks the same questions: What is permitted, who reviews, what evidence matters, and what remains uncertain?

How the prompt format will work in the remaining sections

The remaining sections will provide exactly 25 sequential prompts, each with the same five labels: Purpose, Copy-paste prompt, Required inputs, Expected output, and Verification checkpoint. The consistent format matters because role-based learning plans become easier to audit when every prompt states its intended use, required inputs, expected artifact, and human review step.

Every copy-paste prompt will instruct ChatGPT to use only redacted, synthetic, public, or organization-approved inputs; avoid passwords, tokens, private student or employee records, medical, legal, or financial records, confidential source code, protected assessment answers, and other sensitive data; treat Academy course completion or a badge as learning evidence rather than a license or authorization; keep final decisions with the authorized human; identify uncertainty and missing inputs; follow data, accessibility, academic-integrity, security, and review policies; and include a human verification checkpoint tied to a real permitted task.

The prompts will not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions. Where a prompt touches a consequential domain, it will route the output toward planning, review, documentation, or questions for an authorized professional rather than a final action. That design is deliberate: the goal is to make ChatGPT useful for learning while preserving the boundaries that real organizations and schools need.

Use the prompts as templates, not as universal policy. Before adopting them across a company, school, or public-sector organization, review them against local rules, bargaining obligations, accessibility requirements, data-protection laws, procurement terms, information-security controls, student policies, and professional duties. A well-written prompt can reduce ambiguity, but it cannot replace institutional governance or the judgment of qualified people responsible for the work.

Prompts 1–9: Build the baseline, choose the right path, and turn learning into verified practice

These first nine prompts establish the operating foundation for a role-based AI learning plan: what a learner already knows, which OpenAI Academy pathway best fits the role, which practice tasks are permitted, how employees should apply skills to real workflows, when agent delegation requires checkpoints, how developers should plan Codex practice, how API builders should evaluate systems, how leaders should select an adoption initiative, and who owns governance. OpenAI’s Academy expansion names four role-based paths—Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI—and the prompts below use those names as learning-path labels rather than as promises of access, reporting, productivity improvement, or readiness for consequential work.

Each copy-paste prompt contains the same mandatory contract: use only redacted, synthetic, public, or organization-approved inputs; avoid sensitive data; treat Academy completion and badges as learning evidence only; keep final decisions with authorized humans; surface uncertainty and missing inputs; follow organizational policy; and include a human verification checkpoint tied to a permitted task. That repetition is intentional: learners often reuse individual prompts outside the original program context, so every prompt must carry its own privacy, badge-boundary, uncertainty, policy, and human-decision controls.

Prompt 1: Baseline skills inventory for a role-based AI learning plan

Purpose

Use this prompt before assigning courses or practice work. It helps a learner or manager inventory current skills, risk awareness, role constraints, accessibility needs, and available practice opportunities. The result should distinguish confidence from evidence: a learner may feel fluent with ChatGPT, but still lack documented ability to write clear instructions, provide context, review responses, build reusable workflows, or set human checkpoints—the specific work habits OpenAI associates with the Apply AI at Work path.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping me create a role-based AI learning baseline aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create a baseline skills inventory for the learner described below. Organize the inventory by:
1. Current role and work context
2. Prior AI experience
3. Prompting skills: instructions, context, constraints, examples, and response review
4. Workflow skills: repeatable processes, documentation, escalation, and human checkpoints
5. Technical skills if relevant: Codex, API design, retrieval, evaluations, agents, production operations
6. Leadership skills if relevant: business value, prioritization, ownership, governance, roadmap
7. Education skills if relevant: learning objectives, permitted materials, source use, academic integrity, final human responsibility
8. Risk awareness: privacy, security, IP, accessibility, compliance, and overreliance
9. Evidence currently available
10. Recommended learning path or combination of paths

Learner role:
[ROLE]

Work or study context:
[CONTEXT]

Approved materials that may be discussed:
[APPROVED MATERIALS ONLY]

Known constraints:
[POLICIES, ACCESSIBILITY NEEDS, TIME AVAILABLE, TOOLS AVAILABLE]

Return:
- A baseline table with skill area, current evidence, confidence level, gaps, and recommended next action.
- Three permitted practice tasks that do not require sensitive data.
- A human verification checkpoint for one real permitted task.

Required inputs

• Learner role, such as analyst, software engineer, product manager, school administrator, instructor, student, executive sponsor, or operations lead.
• Approved context that can be safely shared, such as public job responsibilities, sanitized workflow descriptions, or synthetic examples.
• Known policies, including data-handling, academic-integrity, accessibility, security, and review requirements.
• Time available for learning and whether practice must occur inside a workplace, classroom, lab, or personal study setting.

Expected output

The expected output is a structured baseline table, not a final curriculum. It should show which skills are already evidenced, which are self-reported, and which need observed practice. A strong response will avoid ranking the learner by title alone; for example, a senior developer may still need evaluation-design practice, while a nontechnical operations manager may already have strong review and escalation habits.

Verification checkpoint

Ask the learner’s manager, instructor, mentor, or authorized reviewer to inspect one low-risk permitted task and confirm whether the learner’s baseline accurately reflects observed behavior. The reviewer should check the task instructions, the materials used, the AI-generated output, the learner’s edits, and any uncertainty log before approving the next learning step.

Prompt 2: Role-to-path mapping across the four OpenAI Academy pathways

Purpose

This prompt maps a person or cohort to one or more Academy pathways without assuming that everyone needs the same curriculum. OpenAI describes Academy as role-specific practice rather than a single universal course sequence. In an organization, a customer-support specialist may start with Apply AI at Work, a platform engineer may need Build with AI, a department head may need Lead AI Adoption, and a teacher or student may need Teach and Learn with AI. Mixed roles may need a primary path plus a small number of secondary modules.

This article examines how OpenAI Codex is becoming a unified desktop work operating system and how that shift changes developer workflows in 2026. The Codex Becomes OpenAI’s Desktop Work OS: How the Unified Workspace Changes Developer Workflows in 2026 article is a focused companion for Codex Developer Workflows because it directly matches the Codex developer workflow topic and gives developer-focused context beyond the current role-based learning prompt pack.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping map learners to OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Map the learner or cohort below to a primary Academy pathway and, if justified, one secondary pathway. Use these descriptions:
- Apply AI at Work: knowledge-worker practice with clear instructions, context, response review, reusable workflows, agent delegation, checkpoints, and human review.
- Build with AI: developer and technical-team practice with Codex or the OpenAI API, including planning, implementation, review, quality, solution design, evaluations, agents, retrieval, and production operations.
- Lead AI Adoption: leadership practice connecting business value, priorities, ownership, governance, strategy, and roadmap.
- Teach and Learn with AI: educator and student practice using permitted materials, learning objectives, source material, assignment requirements, and final human responsibility.

Learner or cohort description:
[ROLE OR COHORT]

Primary responsibilities:
[RESPONSIBILITIES]

AI tasks they are allowed to practice:
[PERMITTED PRACTICE TASKS]

Tasks they are not allowed to automate or decide:
[PROHIBITED OR CONSEQUENTIALLY CONTROLLED TASKS]

Return:
- Recommended primary path and rationale.
- Optional secondary path and rationale.
- Skills that are in scope.
- Skills that are out of scope for this learning plan.
- A 2-week starter plan using only permitted practice.
- A human verification checkpoint for path selection.

Required inputs

• A role or cohort description with enough detail to distinguish business, technical, leadership, educator, and student responsibilities.
• A list of permitted practice tasks, preferably approved by a manager, instructor, program owner, or administrator.
• A list of tasks that remain off limits, such as grading, legal advice, employment decisions, publishing, deployment, credentialing, payments, or security changes.

Expected output

The output should recommend a primary pathway and explain the mapping in operational terms. For example, a software engineer building internal tools should not be sent only to a general workplace prompting path if their actual learning need is evaluation design, retrieval behavior, code review, and production operations. Conversely, a nontechnical team lead should not be routed into Build with AI merely because their team uses an API product.

Verification checkpoint

The program owner should compare the recommended path against the learner’s authorized work. If the plan includes technical building, production operations, classroom activities, or leadership governance, the reviewer should verify that the learner has the right supervision and that the prompt did not transform a learning pathway into an access decision or job qualification.

Prompt 3: Permitted practice-task selection for safe learning exercises

Purpose

This prompt converts a job, class, or project context into safe practice tasks. It is especially useful because Academy-style learning is practical: learners should practice with work-like materials, but not with materials they are not permitted to disclose or process. The prompt forces a separation between real tasks, redacted variants, synthetic substitutes, public examples, and tasks that are too sensitive for the learning environment.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping select permitted AI practice tasks for a role-based learning plan.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Review the possible practice tasks below and classify each as:
A. Safe to use as written
B. Safe only if redacted
C. Use a synthetic substitute
D. Use a public substitute
E. Not appropriate for this learning plan

For each task, explain the reason and suggest a safer version when needed.

Role or course context:
[CONTEXT]

Possible practice tasks:
[TASK LIST]

Policies or restrictions:
[POLICIES]

Available public or synthetic materials:
[MATERIALS]

Return:
- A classification table.
- A revised practice-task list for the learner.
- Data that must be removed before practice.
- Reviewers who should approve each task type.
- A human verification checkpoint before the learner uses a real task.

Required inputs

• A candidate list of work, study, teaching, development, or leadership tasks.
• The policies that determine what can be used, including workplace confidentiality, student privacy, academic-integrity rules, source-code restrictions, data classification, and client or customer obligations.
• Any approved public datasets, synthetic cases, sample documents, or sanitized templates.

Expected output

The expected output is a classification table that helps learners avoid accidental disclosure. A good result will not simply say “redact sensitive data”; it should identify categories to remove, such as names, account details, unpublished code, assessment answers, personnel records, client facts, security details, and contract language. It should also identify tasks that remain inappropriate even after redaction because the decision itself is consequential.

Verification checkpoint

Before learners use real work or classroom material, an authorized reviewer should approve the selected practice task and the redaction method. The learner should retain a short record of what was changed, why the remaining material is permitted, and what final human review is required.

Prompt 4: Employee workflow practice plan for Apply AI at Work

Purpose

This prompt creates a workplace practice plan for employees who need to improve everyday AI use. It focuses on clear instructions, context, output checking, reusable workflows, and human review—the practices OpenAI associates with Apply AI at Work. The plan is appropriate for knowledge work such as summarizing public materials, drafting internal outlines from approved sources, comparing options, preparing meeting notes from non-sensitive notes, or turning a recurring checklist into a reusable prompt.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping an employee design an Apply AI at Work practice plan.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create a 10-business-day employee practice plan for Apply AI at Work. The plan must help the learner practice:
1. Writing clear instructions
2. Providing appropriate context
3. Asking for structured output
4. Reviewing responses for accuracy and completeness
5. Turning one repeated activity into a reusable workflow
6. Deciding when to escalate to a human reviewer
7. Documenting what changed after AI assistance

Employee role:
[ROLE]

Permitted work-like task:
[TASK]

Approved materials:
[APPROVED MATERIALS]

Manager or reviewer:
[REVIEWER ROLE]

Constraints:
[TIME, ACCESSIBILITY, TOOLS, POLICIES]

Return:
- A day-by-day practice plan.
- A reusable prompt template for the chosen task.
- A response-review checklist.
- A small evidence log template.
- A human verification checkpoint with reviewer questions.

Required inputs

• A role and one recurring task that can be practiced safely.
• Approved source material, such as sanitized notes, public documentation, internal non-sensitive templates, or synthetic cases.
• The reviewer role responsible for confirming quality and policy compliance.

Expected output

The plan should produce a sequence of short practice sessions rather than a vague “learn prompting” assignment. It should include a reusable prompt template, a response-review checklist, and an evidence log showing the original task goal, what the model produced, what the learner changed, what was uncertain, and what the reviewer approved or rejected.

Verification checkpoint

The manager or reviewer should inspect one completed workflow artifact and confirm that the learner did not paste sensitive material, did not accept the output uncritically, documented uncertainty, and made the final work decision themselves or escalated it to the correct human authority.

Prompt 5: Agent-delegation checkpoints and human review boundaries

Purpose

This prompt is for learners practicing agent-style delegation: asking AI to break down work, perform multi-step assistance, or prepare a draft plan. OpenAI’s description of Apply AI at Work includes agent delegation, checkpoints, and human review, which means the learning objective is not “let the AI run unattended.” The objective is to define what may be delegated, what must be reviewed, and where the agent must stop before external or consequential action.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping design a safe agent-delegation practice exercise for AI learning.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create an agent-delegation checklist for the task below. The checklist must define:
1. What the AI may do independently as drafting, analysis, organization, or preparation
2. What the AI may not do
3. What information may be used
4. What information must never be used
5. Stop points where the AI must ask for human review
6. Quality checks before any output is reused
7. Escalation rules for uncertainty, missing evidence, conflicting instructions, or policy concerns
8. A final human decision step

Permitted task:
[TASK]

Approved inputs:
[APPROVED INPUTS]

External actions or consequential decisions that are prohibited:
[PROHIBITED ACTIONS]

Reviewer:
[REVIEWER ROLE]

Return:
- A delegation boundary table.
- A step-by-step checkpoint plan.
- A short learner-facing instruction block.
- A reviewer-facing approval checklist.
- A human verification checkpoint before any external message, submission, publication, system change, payment, booking, or deployment.

Required inputs

• A task that can be decomposed safely, such as preparing a draft outline, organizing public research notes, generating test-case ideas from a synthetic specification, or creating a meeting agenda from approved content.
• Approved input categories and prohibited input categories.
• External actions or consequential decisions that must not be automated.

Expected output

The output should define clear stop points. For example, the AI may draft an internal summary from approved notes, but it must not send the email, publish the page, change permissions, approve a purchase, modify production code, or represent the organization externally. The best responses will include escalation triggers for uncertainty, missing evidence, conflicting instructions, suspected sensitive data, and policy ambiguity.

Verification checkpoint

A human reviewer should walk through the delegation boundary before the learner uses it. The reviewer should confirm that every external message, submission, purchase, booking, deployment, permission change, legal commitment, grade, employment action, or other consequential operation requires explicit human approval.

Prompt 6: Codex developer learning plan for Build with AI

Purpose

This prompt helps developers use the Build with AI pathway as a structured learning plan for software-development practice. OpenAI describes Build with AI as covering Codex or OpenAI API work, including planning, implementation, review and quality, solution design, evaluations, agents, retrieval, and production operations. For Codex learners, the safe default is to practice on synthetic repositories, toy projects, training branches, or explicitly approved code—not on confidential source code or production systems without authorization.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping a developer create a Build with AI learning plan focused on Codex-style software-development practice.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create a developer learning plan for practicing with Codex or Codex-like AI coding assistance. Cover:
1. Problem framing and development planning
2. Breaking a task into implementation steps
3. Asking for code suggestions using only approved or synthetic code
4. Reviewing generated code for correctness, security, maintainability, and tests
5. Writing or improving tests
6. Documenting assumptions and limitations
7. Deciding what must not be delegated
8. Preparing a pull-request-style review package for a human reviewer

Developer background:
[BACKGROUND]

Approved repository or synthetic project:
[APPROVED PROJECT DESCRIPTION]

Languages or frameworks:
[LANGUAGES/FRAMEWORKS]

Security and review requirements:
[REQUIREMENTS]

Return:
- A 3-week learning plan.
- A weekly practice task using only approved materials.
- A code-review checklist.
- A test and evaluation checklist.
- A human verification checkpoint before merge, deployment, credential handling, permission changes, or production operations.

Required inputs

• The developer’s background and the technologies they are allowed to discuss.
• An approved repository description, synthetic project, public learning project, or sanitized task specification.
• Security, testing, code review, and deployment policies that govern the practice environment.

Expected output

The expected plan should separate learning from deployment. It may include task breakdown, test creation, code explanation, refactoring suggestions, and pull-request preparation, but it must preserve human review before merge or release. A strong plan also includes a review checklist for generated code: dependency risks, secrets handling, authorization checks, error handling, logging, privacy, tests, maintainability, and alignment with project conventions.

Verification checkpoint

A qualified developer or engineering reviewer should inspect the practice branch, test output, review notes, and AI interaction summary before any merge, deployment, or production operation. Academy participation or a badge must not replace the organization’s code-review, security-review, or release-management process.

Prompt 7: API builder evaluation plan for Build with AI

Purpose

This prompt is for builders using the OpenAI API or designing AI-enabled systems who need an evaluation plan before release. OpenAI’s Academy materials place evaluations, agents, retrieval, and production operations inside Build with AI, but course learning alone does not establish that a system is safe, reliable, compliant, or useful in a local environment. The prompt asks for test tasks, success criteria, failure categories, review gates, and evidence collection without inventing benchmarks or assuming production readiness.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping an API builder create a Build with AI evaluation plan for an AI-enabled application.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create an evaluation plan for the AI-enabled application below. Do not claim the system is production-ready. Define what evidence would be needed before any release decision.

Application description:
[APPLICATION]

Allowed data for evaluation:
[REDACTED, SYNTHETIC, PUBLIC, OR APPROVED DATA]

Users and use cases:
[USERS AND USE CASES]

Known risks:
[RISKS]

Policies and review requirements:
[POLICIES]

Return:
- Evaluation objectives.
- Test-case categories, including normal, edge, refusal/insufficient-evidence, safety, privacy, security, accessibility, and misuse cases where relevant.
- Success criteria and failure taxonomy.
- Human-review plan.
- Logging or evidence artifacts to retain, avoiding sensitive content where possible.
- Release-gate questions for authorized humans.
- A human verification checkpoint before deployment, external access, publication, payments, security changes, or consequential decisions.

Required inputs

• A concise application description, including the user group and intended tasks.
• Allowed evaluation data that is public, synthetic, redacted, or organization-approved.
• Known risks, such as hallucinated facts, unsafe tool use, privacy exposure, biased output, stale retrieval, overconfident answers, or weak refusal behavior.
• Policies and review gates that apply before deployment.

Expected output

The output should be an evaluation plan with measurable categories, not a release recommendation. It should define what will be tested, what counts as acceptable, what failures require redesign, and who reviews the evidence. It should include refusal or insufficient-evidence behavior because a useful AI application must know when not to answer, not just when to produce fluent text.

Verification checkpoint

An authorized product, engineering, security, legal, privacy, accessibility, or compliance reviewer—depending on the application—should approve the evaluation plan before testing expands beyond approved data. A human release authority should decide whether evidence is sufficient for any deployment step.

Prompt 8: Leadership initiative selection for Lead AI Adoption

Purpose

This prompt helps leaders select one AI adoption initiative for a learning roadmap. OpenAI describes Lead AI Adoption as connecting business value, priorities, ownership, governance, strategy, and a roadmap. The prompt prevents a common failure mode: launching broad AI training without choosing a specific, reviewable initiative and without defining what evidence would show that the initiative changed work quality, cycle time, risk, or employee experience.

This enterprise strategy analysis examines OpenAI’s reported shift from chat use toward multi-step Codex agents and the organizational adoption implications, giving leaders a concrete context for role-specific learning plans. The OpenAI’s Shift from Chat to Agents: How 97.9% Internal Codex Adoption Is Reshaping Enterprise AI Strategy article is a focused companion for AI Leadership Plans because it is a direct enterprise leadership and adoption target, unlike the draft university-student prompt article.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping a leader select one AI adoption initiative for a Lead AI Adoption learning plan.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Help select one AI adoption initiative from the candidate list below. Evaluate each candidate using:
1. Business or mission value
2. Learner readiness
3. Data sensitivity
4. Review burden
5. Accessibility and inclusion considerations
6. Technical complexity
7. Governance requirements
8. Measurement feasibility
9. Risk of overclaiming causation from usage metrics or course completion
10. Suitability for a limited pilot

Organization or team context:
[CONTEXT]

Candidate initiatives:
[INITIATIVES]

Known policies and constraints:
[POLICIES]

Available sponsors and owners:
[SPONSORS/OWNERS]

Return:
- A comparison table.
- One recommended initiative and rationale.
- Initiatives to defer and why.
- A draft scope statement.
- Initial success measures that do not assume causation.
- Governance and review needs.
- A human verification checkpoint for sponsor approval before launch.

Required inputs

• A short description of the organization, department, school, or team context.
• A list of candidate initiatives, such as employee workflow support, developer enablement, educator planning support, student study skills, knowledge-base assistance, or leadership planning.
• Known constraints, including sensitive data, regulated decisions, union or workforce obligations, accessibility requirements, academic policies, security controls, and reporting availability.
• Potential sponsors and operational owners.

Expected output

The expected output should recommend one constrained initiative, not a sweeping transformation program. It should identify why certain attractive candidates should be deferred, especially if they involve high-risk data, unclear ownership, weak review capacity, or consequential decisions. It should also avoid causal overclaims: usage increases, course completion, or badge counts may indicate participation, but they do not prove that learning caused business outcomes.

Verification checkpoint

The executive sponsor or authorized program owner should approve the initiative scope before launch. That approval should confirm the business objective, permitted data, review gates, accessibility commitments, measurement plan, and the fact that learning evidence is not a substitute for governance or operational controls.

Prompt 9: Governance ownership map for a role-based AI learning program

Purpose

This prompt creates the ownership map that should surround any role-based AI learning program. The Academy Champion deployment guide describes rollout activities such as activation, sponsor engagement, launch, reinforcement and measurement, and sharing results. Those activities require named owners; otherwise, completion signals can be mistaken for capability, practice examples can be collected without review, and learners can be left uncertain about which materials, tools, or tasks are permitted.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping create a governance ownership map for a role-based AI learning program aligned with OpenAI Academy pathways.

Mandatory contract:
- Use only redacted, synthetic, public, or organization-approved inputs.
- Do not request, infer, store, or expose passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, regulated personal data, privileged material, or other sensitive information.
- Treat OpenAI Academy course completion or a badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, authorization to access systems, or proof that a person or system is competent for consequential work.
- Keep final decisions with the authorized human reviewer. Do not automate hiring, firing, promotion, grading, discipline, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.
- Identify uncertainty, missing inputs, assumptions, and risks. If information is missing, ask concise follow-up questions or label the output as provisional.
- Follow the organization’s data-handling, accessibility, academic-integrity, security, and review policies.
- Include a human verification checkpoint tied to a real permitted task.

Task:
Create a governance ownership map for the learning program below. Include owners for:
1. Pathway selection
2. Permitted practice-task approval
3. Data-handling and privacy review
4. Security review
5. Accessibility support
6. Academic-integrity review if learners are students or educators
7. Manager or instructor coaching
8. Office hours or application sessions
9. Evidence collection
10. Badge and completion interpretation
11. Measurement and causal-caution review
12. Escalation for policy uncertainty
13. Final approval for external, consequential, production, grading, employment, legal, financial, medical, payment, publication, or deployment actions

Program context:
[CONTEXT]

Learner groups:
[GROUPS]

Available owners:
[OWNER ROLES]

Policies:
[POLICIES]

Return:
- A RACI-style table with Responsible, Accountable, Consulted, and Informed roles.
- Decision rules for when learners must stop and ask for review.
- Evidence artifacts to collect without unnecessary sensitive data.
- Badge-boundary language for program communications.
- A human verification checkpoint for governance approval before launch.

Required inputs

• The program context, such as a department rollout, developer enablement effort, classroom adoption plan, leadership cohort, or cross-functional pilot.
• Learner groups and the Academy pathways likely to apply to each group.
• Available owner roles, including sponsors, managers, instructors, IT, security, privacy, legal, HR, accessibility, academic-integrity leads, and technical reviewers where relevant.
• Policies governing data use, tools, review, accessibility, records, assessment, and external communication.

Expected output

The output should be a practical ownership table that makes responsibilities visible before the program starts. It should include language explaining that completion and badges are learning evidence, not authorization for production access, regulated decisions, grading, employment action, publication, deployment, or safety claims. It should also identify evidence artifacts that can be collected safely, such as redacted practice logs, reviewer checklists, office-hour themes, and before/after workflow examples where permitted.

Verification checkpoint

The sponsor and governance owners should review the RACI map before launch and confirm that every learner knows where to ask questions, which materials are permitted, who reviews practice work, how accessibility needs are handled, and who has final authority over consequential actions. If no accountable owner exists for a risk area, the program should pause that activity until ownership is assigned.

Prompts 10–18: Teach, study, coach, and prove applied learning without crossing policy boundaries

Prompts 10–18 shift from organizational setup into classroom, student, manager, and support workflows. OpenAI describes Teach and Learn with AI as a path for educators and students that should be reviewed against learning objectives, source material, and assignment requirements, with educators and students retaining responsibility for the final result. That boundary matters: these prompts help design plans, practice tasks, feedback routines, accessibility adaptations, office hours, and evidence rubrics, but they do not grade students, approve employment decisions, certify competence, authorize production deployment, or replace a qualified human reviewer.

The Academy Champion deployment guide also emphasizes support structures such as sponsor reinforcement, manager involvement, office hours or application sessions, and evidence of applied work. The prompts below convert those ideas into copy-paste workflows that require permitted inputs, human verification, and concrete application evidence. Use them with redacted, synthetic, public, or organization-approved materials only, especially in education settings where private student records, protected assessment answers, disability documentation, disciplinary records, and family information must not be pasted into a general prompt.

This guide presents a structured prompting framework that organizes prompts into role, context, instructions, examples, input, and output schema for consistent AI behavior. The The Structured Prompting Prompting Framework: Complete Guide for 2026 article is a focused companion for Educator Prompting because educators creating reliable learning-plan prompts need a clear prompt structure, making this framework useful for improving instructional prompt design.

Prompt 10: Educator lesson-planning assistant using permitted materials

Purpose

This prompt helps an educator build a lesson plan aligned with learning objectives while staying inside permitted-material rules. It is designed for the Teach and Learn with AI path and assumes that the teacher, department, school, or institution remains responsible for the final lesson, classroom suitability, copyright compliance, accessibility, and academic-integrity rules.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping me draft an educator-reviewed lesson plan for the OpenAI Academy Teach and Learn with AI pathway.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, disability records, disciplinary records, unpublished student work without permission, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, student grade, safety certification, authorization to teach a regulated subject, or proof that the lesson is compliant.

Keep final decisions with the authorized human educator or institution. Do not automate grading, disciplinary, hiring, firing, promotion, credential, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, curriculum, copyright, and review policies. Include a human verification checkpoint tied to a real permitted teaching task.

Task: Create a lesson plan using the inputs below. Align the plan to the learning objectives and permitted materials. Flag any places where the educator must verify accuracy, appropriateness, accessibility, source permissions, or alignment with school policy.

Inputs:
- Course or subject:
- Learner level:
- Lesson duration:
- Learning objectives:
- Permitted source materials:
- Materials that must not be used:
- Required standards or curriculum constraints:
- Accessibility considerations, stated generically:
- Academic-integrity expectations:
- Desired classroom activity:
- Human reviewer:
- Date needed:

Required inputs

• Course or subject, learner level, lesson length, and learning objectives.
• A list of permitted materials, such as public readings, approved slides, approved textbook chapters, or teacher-created notes.
• Institutional constraints, including academic-integrity rules, accessibility obligations, and any required curriculum standards.
• A named human reviewer by role, such as course lead, department chair, instructional designer, or teacher of record.

Expected output

The output should be a structured lesson plan with objectives, materials, opening activity, guided practice, independent practice, formative checks, closing reflection, and teacher-review flags. It should also list what the AI could not verify, such as whether a reading is licensed for classroom use or whether a proposed activity fits a specific learner accommodation.

Verification checkpoint

Before using the lesson, the authorized educator reviews all content against the permitted source materials, confirms that no protected assessment answers or private student details were included, checks accessibility and academic-integrity fit, and records one permitted classroom task where the plan will be tested and revised.

Prompt 11: Assessment review without exposing protected answers or grading decisions

Purpose

This prompt helps educators review the design quality of an assessment without asking ChatGPT to grade individual students or expose answer keys that should remain protected. It is useful for checking alignment, clarity, cognitive level, accessibility, and fairness while keeping final grading and assessment approval with the authorized educator.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping me review an assessment design, not grade students or reveal protected answers.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, disability records, student identifiers, disciplinary records, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, credentialing decision, or proof that an assessment is valid.

Keep final decisions with the authorized human educator or institution. Do not automate grading, disciplinary action, credentialing, hiring, firing, promotion, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, assessment, and review policies. Include a human verification checkpoint tied to a real permitted assessment-review task.

Task: Review the assessment design below for alignment, clarity, accessibility, academic-integrity risk, and review needs. Do not solve protected assessment items or assign grades. If sample items are included, treat them as drafts for design critique only.

Inputs:
- Course or subject:
- Learner level:
- Learning objectives:
- Assessment type:
- Draft assessment instructions:
- Sample non-protected item types or public examples:
- Rubric draft, if permitted:
- Known accessibility requirements, stated generically:
- Academic-integrity concerns:
- Human reviewer:
- Deadline:

Required inputs

• Learning objectives and assessment purpose.
• Draft instructions or sample non-protected item types, not secure exam content.
• A rubric draft if policy permits sharing it.
• Accessibility and academic-integrity constraints expressed without private student details.

Expected output

The output should identify alignment gaps, ambiguous instructions, likely accessibility barriers, overreliance on recall or unsupported item types, and points requiring human review. A strong response should separate design advice from prohibited grading or credentialing decisions.

Verification checkpoint

The educator verifies that no protected answer key, private student data, or secure assessment content was used, then reviews the recommendations against institutional assessment policy before revising the assessment.

Prompt 12: Student study plan from approved course materials

Purpose

This prompt helps students create a study plan from approved materials without outsourcing learning, violating academic-integrity rules, or asking ChatGPT to complete graded work. It is aligned with Teach and Learn with AI because it emphasizes review against source material, assignment requirements, and final student responsibility.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping me create a study plan from materials I am permitted to use.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private instructor materials, classmates’ private work, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, admission decision, credential, or proof that I have mastered the subject.

Keep final decisions and final submitted work with the authorized human student and educator. Do not automate grading, disciplinary, credential, admission, hiring, firing, promotion, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow my school’s or organization’s data, accessibility, academic-integrity, security, citation, collaboration, and review policies. Include a human verification checkpoint tied to a real permitted study task.

Task: Build a study plan that helps me learn, practice, self-check, and ask better questions without completing graded work for me.

Inputs:
- Course or subject:
- Upcoming topic or exam scope:
- Approved materials:
- Assignment or exam rules:
- Time available:
- Concepts I find difficult:
- Preferred study format:
- Accessibility needs, stated generally:
- What help is allowed:
- Teacher, tutor, or study-group verification option:

Required inputs

• Approved study sources, such as course notes, public readings, assigned textbook sections, or instructor-approved resources.
• Rules for the assignment, exam, or course collaboration policy.
• Available study time and topics of difficulty.
• A permitted human verification option, such as office hours, a tutor, a teaching assistant, or a study group.

Expected output

The output should include a schedule, concept map, practice routine, self-check questions, source-review steps, and a list of questions to bring to a teacher or tutor. It should avoid producing final answers for graded assignments unless the user confirms that such help is allowed and the content is not protected.

Verification checkpoint

The student compares the plan against the syllabus or assignment rules, removes anything that would violate academic integrity, and brings at least one unresolved concept question to an authorized teacher, tutor, or study group.

Prompt 13: Group-project role plan with human accountability

Purpose

This prompt helps students, educators, and workplace learners divide group-project responsibilities without letting ChatGPT assign grades, make personnel judgments, or replace a team agreement. It supports safer collaboration by making roles, evidence, review points, and permitted AI uses explicit.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping a group create a project role plan with clear responsibilities and human accountability.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private peer evaluations, disciplinary records, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, credential, team placement mandate, or proof of competence.

Keep final decisions with the authorized humans: the students, educator, manager, or project owner. Do not automate grading, hiring, firing, promotion, discipline, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, collaboration, authorship, and review policies. Include a human verification checkpoint tied to a real permitted group-project task.

Task: Create a fair project role plan that identifies responsibilities, collaboration norms, permitted AI use, review checkpoints, and evidence each member can provide.

Inputs:
- Project goal:
- Course, team, or program context:
- Permitted materials and tools:
- Prohibited materials or AI uses:
- Team roles needed:
- Team member preferences, stated without private records:
- Timeline:
- Required deliverables:
- Review authority:
- Accessibility or participation considerations, stated generally:

Required inputs

• Project goal, deliverables, timeline, and review authority.
• Permitted and prohibited AI uses, including whether AI may support brainstorming, drafting, summarizing, coding, editing, or practice.
• Role needs such as researcher, drafter, reviewer, presenter, tester, coordinator, or documentation owner.
• General accessibility or participation considerations without private medical or disability records.

Expected output

The output should include a role matrix, communication plan, review cadence, evidence expectations, and escalation path for unresolved disagreements. It should also recommend that the group document AI assistance according to course or organizational policy.

Verification checkpoint

The team reviews the role plan with the instructor, manager, or project owner where required, confirms that permitted AI-use rules are understood, and records one shared deliverable that will receive human review before submission or publication.

Prompt 14: Career-preparation plan without employment-decision automation

Purpose

This prompt helps learners translate AI-course participation and practice artifacts into a career-development plan while avoiding unsupported claims about badges, employability, or professional authorization. It is appropriate for students, career centers, managers, and knowledge workers who want to describe learning evidence responsibly.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping me create a career-preparation plan based on learning goals, permitted practice work, and evidence I can verify.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private performance reviews, immigration records, salary records, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, promotion decision, hiring qualification, grade, safety certification, regulated credential, or proof of job readiness.

Keep final decisions with the authorized human learner, career advisor, manager, recruiter, or employer. Do not automate hiring, firing, promotion, discipline, admissions, grading, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, résumé, portfolio, and review policies. Include a human verification checkpoint tied to a real permitted career-preparation task.

Task: Build a career-preparation plan that maps my learning goals to practice artifacts, interview preparation, portfolio evidence, and honest language about course completion or badges.

Inputs:
- Target role or career direction:
- Current skills:
- OpenAI Academy pathway or course area:
- Permitted practice artifacts:
- Skills I want to demonstrate:
- Constraints on sharing work:
- Portfolio or résumé rules:
- Interview or presentation context:
- Human reviewer:
- Timeline:

Required inputs

• Target role or career direction and current skills.
• Permitted practice artifacts, such as redacted workflow summaries, public code samples, approved teaching materials, or synthetic case studies.
• Constraints on sharing employer, school, client, or team work.
• A human reviewer such as a career advisor, manager, mentor, or instructor.

Expected output

The output should include a skills map, practice-artifact plan, résumé language that avoids exaggeration, interview practice questions, and a verification checklist. It should explain that a badge or completion record can support a learning narrative but does not prove authorization, professional competence, or job eligibility.

Verification checkpoint

The learner asks a qualified career advisor, manager, instructor, or mentor to review one résumé bullet or portfolio entry for accuracy, confidentiality, and appropriate claims before sharing it externally.

Prompt 15: Accessibility adaptation for role-based AI learning

Purpose

This prompt supports inclusive learning-plan design without asking the user to disclose disability records or medical details. It is suitable for training leads, educators, managers, and learners who need flexible formats, pacing, and participation options while following institutional accessibility policies.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping adapt a role-based AI learning plan for accessibility, inclusion, and practical participation.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, disability documentation, accommodation letters, diagnosis details, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, accommodation approval, legal compliance determination, or proof of competence.

Keep final decisions with the authorized human learner, educator, manager, accessibility office, HR function, or institution. Do not automate grading, hiring, firing, promotion, discipline, accommodation determinations, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, privacy, accommodation, and review policies. Include a human verification checkpoint tied to a real permitted learning task.

Task: Adapt the learning plan below so learners have practical ways to participate, practice, review, and demonstrate applied learning without requiring disclosure of private medical or disability information.

Inputs:
- Audience or role:
- Academy pathway or course area:
- Learning objective:
- Current activity format:
- Known barriers, stated generally:
- Permitted alternative formats:
- Required deadlines or pacing:
- Review policy:
- Human accessibility or program contact:
- Evidence of learning needed:

Required inputs

• Audience, learning objective, pathway, and current activity format.
• General barriers, such as time-zone constraints, screen fatigue, language complexity, inaccessible documents, or limited quiet workspace.
• Permitted alternatives, such as transcripts, captions, written reflection, smaller practice tasks, asynchronous participation, or manager-supported time blocks.
• The authorized accessibility, HR, instructor, or program contact for formal decisions.

Expected output

The output should recommend accessible formats, pacing options, participation choices, and evidence alternatives while flagging anything that requires formal review. It should avoid determining whether a person is entitled to an accommodation or making legal compliance claims.

Verification checkpoint

The program owner checks the adaptation with the authorized accessibility or policy contact, then tests one permitted learning task with the revised format before using it broadly.

Prompt 16: Manager coaching plan for applying Academy learning at work

Purpose

This prompt helps managers coach employees who are applying Academy learning to real workflows. It reflects the Champion deployment guide’s emphasis on manager reinforcement and application evidence, while keeping performance management, promotion, discipline, and production authorization outside the prompt’s decision scope.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping a manager create a coaching plan for employees applying role-based AI learning at work.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private performance files, compensation data, disciplinary records, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, promotion decision, performance rating, grade, safety certification, system-access authorization, or proof of productivity improvement.

Keep final decisions with the authorized human manager, employee, governance owner, security reviewer, legal reviewer, or business owner. Do not automate hiring, firing, promotion, discipline, compensation, grading, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, HR, privacy, acceptable-use, and review policies. Include a human verification checkpoint tied to a real permitted work task.

Task: Create a coaching plan that helps employees apply learning safely, choose permitted practice tasks, document evidence, and escalate work requiring review.

Inputs:
- Team function:
- Employee role group, not individual private details:
- Relevant Academy pathway:
- Business workflow to improve:
- Permitted materials:
- Prohibited materials:
- Review or approval requirements:
- Risks to manage:
- Coaching cadence:
- Evidence of applied learning needed:

Required inputs

• Team function, role group, and relevant Academy path, such as Apply AI at Work, Build with AI, or Lead AI Adoption.
• Permitted work materials and prohibited data categories.
• Review requirements for customer communication, code deployment, legal review, security changes, publication, or other consequential actions.
• Evidence the manager wants to see, such as a redacted workflow map, revised checklist, evaluation notes, or reviewed draft.

Expected output

The output should include coaching questions, a practice-task menu, a review cadence, escalation rules, and a non-punitive evidence plan. It should remind the manager that completion or badge status can show participation but must not be used by itself as a performance rating or employment decision.

Verification checkpoint

The manager and employee select one permitted workflow, define the human reviewer, and document what evidence will show learning application without exposing confidential or private records.

Prompt 17: Office-hours agenda for AI learning support and application sessions

Purpose

This prompt helps program champions, educators, or team leads run office hours or application sessions, a support pattern specifically consistent with the Champion deployment guide’s reinforcement stage. The session should help learners bring questions, troubleshoot safe practice tasks, and identify review needs, not bypass approvals or turn usage metrics into proof of causation.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping plan office hours or an application session for role-based AI learning.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private customer records, private performance files, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, grade, safety certification, deployment authorization, access approval, or proof that usage changes caused business outcomes.

Keep final decisions with authorized humans: instructors, managers, program owners, security teams, legal reviewers, business owners, or learners. Do not automate hiring, firing, promotion, discipline, grading, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, privacy, acceptable-use, and review policies. Include a human verification checkpoint tied to a real permitted office-hours task.

Task: Create a practical office-hours agenda that helps learners apply Academy learning to safe tasks, ask better questions, identify risks, and leave with a verified next step.

Inputs:
- Audience:
- Academy pathway or course area:
- Session length:
- Common learner questions:
- Permitted practice materials:
- Topics that must be escalated:
- Accessibility needs, stated generally:
- Facilitator role:
- Sponsor or manager reinforcement message:
- Evidence to collect after the session:

Required inputs

• Audience, pathway, session length, and facilitator role.
• Permitted practice materials and topics that must be escalated rather than handled live.
• Common questions or friction points from learners.
• Evidence to collect, such as anonymized themes, reviewed practice examples, or follow-up support needs.

Expected output

The output should include a timed agenda, facilitation prompts, safe-demo options, escalation language, accessibility checks, and a follow-up evidence template. It should separate participation signals from actual applied-learning evidence and warn against treating usage changes alone as proof that training caused an outcome.

Verification checkpoint

The facilitator confirms that examples used in the session are redacted, synthetic, public, or approved; records unresolved policy or review questions; and assigns one permitted follow-up task to a human owner.

Prompt 18: Application-evidence rubric for learning that changed real work

Purpose

This prompt creates a rubric for evaluating application evidence after learners complete Academy-related activities. It follows the Champion deployment guide’s distinction between participation or completion signals and examples of application. It also preserves the boundary that a badge is not a license, production approval, safety certification, employment decision, grade, or proof that a business metric changed because of training.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping design an application-evidence rubric for a role-based AI learning program.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or rely on passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, private customer records, private performance reviews, disciplinary records, or other sensitive data.

Treat any OpenAI Academy course completion or badge as learning evidence only. Do not treat it as a professional license, production approval, employment decision, promotion decision, grade, safety certification, access authorization, causal proof of performance improvement, or proof that a system is safe.

Keep final decisions with authorized humans: educators, managers, program owners, security reviewers, legal reviewers, business owners, or learners. Do not automate hiring, firing, promotion, discipline, grading, credentialing, publication, payment, medical, legal, tax, financial, deployment, or other consequential decisions.

Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, privacy, acceptable-use, measurement, and review policies. Include a human verification checkpoint tied to a real permitted application-evidence task.

Task: Build a rubric that distinguishes participation, completion, badge evidence, practice quality, human review, applied workflow change, and measured outcome evidence. Make clear that usage metrics alone do not prove causation.

Inputs:
- Program audience:
- Relevant Academy pathway or course area:
- Learning objectives:
- Permitted practice tasks:
- Evidence learners can submit:
- Evidence that must not be submitted:
- Reviewers:
- Scoring or maturity levels desired:
- Accessibility considerations:
- Measurement cautions or confounding factors:

Required inputs

• Audience, pathway, and learning objectives.
• Permitted evidence types, such as redacted before-and-after workflow notes, approved prompts, reviewed drafts, test plans, lesson reflections, or support tickets showing process improvement.
• Evidence that must not be submitted, including private records, confidential code, protected answers, regulated data, or customer identifiers.
• Reviewer roles and any desired maturity levels.

Expected output

The output should be a rubric that separates access, participation, completion, badge evidence, applied practice, reviewer approval, quality improvement, and outcome claims. It should include cautions about missing baseline data, confounding changes, self-reporting limits, and the fact that reporting availability can depend on organizational arrangements and account context.

Verification checkpoint

The program owner tests the rubric on one redacted or synthetic evidence sample, asks an authorized reviewer to confirm that the rubric does not create employment, grading, credentialing, publication, deployment, or safety-approval decisions, and revises the rubric before broad use.

Prompt 19: Launch communications for a role-based Academy learning program

Purpose

Use this prompt to draft practical launch communications that explain who the program is for, which OpenAI Academy pathway each audience should start with, what participation means, and where human review remains required. OpenAI’s Academy deployment guide recommends audience-specific starting points, direct course links, support contacts, a launch window, protected learning time, and sponsor or manager reinforcement. This prompt turns those recommendations into messages for employees, developers, leaders, educators, and students without promising course availability, reporting access, badges, or productivity gains.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping draft launch communications for a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Draft a launch communication package for the following audience groups:
- Knowledge workers
- Developers or technical teams
- Leaders or sponsors
- Educators
- Students

For each group, provide:
1. A short announcement paragraph.
2. The recommended starting pathway and why it fits the role.
3. A permitted practice-task suggestion.
4. A reminder about sensitive data and human review.
5. A support channel or office-hours suggestion.
6. A verification checkpoint where a human reviews one real permitted work, teaching, study, or technical artifact.

Avoid claiming that course completion proves competence, that badge status authorizes consequential work, or that usage changes prove causation.

Required inputs

• Audience groups included in the launch.
• Approved Academy pathway starting points for each group.
• Program dates or launch window, if already approved.
• Support contacts, office-hours schedule, or help channel.
• Local policy reminders for data handling, accessibility, academic integrity, security, and review.

Expected output

The output should be a communication pack with role-specific announcements, short rationale statements, practice-task examples, sensitive-data warnings, support instructions, and a human verification checkpoint for each audience. The tone should be operational rather than promotional: participants should understand what to do next, what not to upload, and what evidence they may produce after completing a course or learning activity.

Verification checkpoint

An authorized program owner should confirm that every audience segment has the correct pathway, approved support contact, accurate launch timing, and a permitted practice task. Legal, HR, education, security, or compliance reviewers should review any message that could be interpreted as changing employment expectations, academic requirements, access rights, or production authorization.

Prompt 20: Participation analysis without overclaiming learning impact

Purpose

Use this prompt to analyze participation, completion, and support attendance while keeping the interpretation conservative. The Academy deployment guide distinguishes participation and completion signals from examples of applied work, and it warns that a change in workspace usage alone is not proof that courses caused the change. This prompt helps a team summarize learning-program engagement without turning attendance into competence or badge status into authorization.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping analyze participation in a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Analyze the following program participation information:
- Invited audience groups:
- Number invited by group:
- Number started by group:
- Number completed by group:
- Office-hours or application-session attendance:
- Known missing data:
- Known reporting limitations:
- Voluntary feedback themes:
- Examples of applied learning, if approved for sharing:

Produce:
1. A participation summary by audience group.
2. A completion summary that avoids treating completion as proof of capability.
3. A list of reporting gaps and self-report limitations.
4. A cautious interpretation of what participation may suggest.
5. A separate section titled "What this data cannot prove."
6. Recommended next actions for manager follow-up, office hours, accessibility support, and evidence collection.

Required inputs

• Aggregated counts for invitations, starts, completions, and support attendance.
• Approved segmentation categories, such as department, role family, or cohort.
• Known reporting limits, including accounts not covered by available reporting.
• Voluntary feedback themes that have been redacted and approved for analysis.
• Approved examples of applied work, if available.

Expected output

The output should separate access, start rate, completion, office-hours participation, and applied-work examples. It should explicitly say that completion is evidence of participation in learning activity and not a license, safety certification, production approval, employment qualification, or causal proof of improved performance. The analysis should identify missing data instead of smoothing over incomplete or self-reported information.

Verification checkpoint

A program reviewer should compare the summary against the underlying aggregated records and confirm that no private student or employee record is exposed. If the analysis will be shared with managers, executives, instructors, or sponsors, a human owner should ensure it cannot be used as a hidden employment, grading, or disciplinary decision system.

Prompt 21: Usage-versus-causation review for AI learning programs

Purpose

Use this prompt when leaders notice changes in ChatGPT, Codex, API, or workspace usage after a learning launch and want to understand what can and cannot be inferred. OpenAI’s Academy deployment guidance cautions that usage changes alone do not prove that training caused a change. This prompt helps teams separate correlation, plausible contribution, confounding factors, and evidence still needed before making business-value claims.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping review whether changes in AI usage can be connected to a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Review the following information:
- Learning-program timeline:
- Audience groups:
- Participation and completion summary:
- Usage metrics available:
- Product, policy, staffing, project, or seasonal changes during the same period:
- Examples of applied work:
- Baseline measures, if any:
- Quality or review outcomes, if any:
- Missing or unreliable data:

Produce:
1. A timeline that places learning activity beside usage changes and other relevant events.
2. A correlation summary that avoids causal language unless evidence supports it.
3. A confounder list.
4. Evidence that would strengthen or weaken a causal claim.
5. A recommended follow-up measurement plan using before/after comparison, holdout tasks, manager review, or quality rubrics where feasible.
6. A plain-language executive summary with cautious wording.

Required inputs

• Aggregated usage measures and their date ranges.
• Learning-program timeline, including launch, office hours, reinforcement, and reassessment dates.
• Baseline, before/after, or cohort comparison data where available.
• Known confounders such as new tools, policy changes, staffing changes, deadlines, or seasonality.
• Approved examples of applied work and review outcomes.

Expected output

The output should include cautious language such as “usage increased during the program window,” “the learning program may have contributed,” or “the available data does not establish causation.” It should recommend additional evidence, such as quality reviews, task-level before/after comparisons, manager observations, learner artifacts, and repeated measurement after reinforcement. It should not infer productivity improvement from message volume, course completion, badge status, or tool activity alone.

Verification checkpoint

A human analytics owner should check the calculations, time windows, cohort definitions, and confounder list before the summary is used in executive reporting. If the report might influence funding, staffing, performance evaluation, grades, deployment authorization, or policy changes, it must be reviewed by the responsible governance, HR, academic, legal, security, or finance authority.

Prompt 22: Learner feedback synthesis for course, support, and accessibility improvements

Purpose

Use this prompt to turn learner feedback into actionable improvements while protecting privacy and avoiding unsupported conclusions about capability. Feedback can reveal unclear instructions, inaccessible materials, inadequate practice time, missing manager support, or policy confusion. It should not be treated as a substitute for verified work quality, secure implementation review, educator judgment, or local assessment standards.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping synthesize learner feedback for a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Synthesize the following redacted learner feedback:
- Audience group:
- Pathway:
- What learners found useful:
- What learners found confusing:
- Barriers to participation:
- Accessibility concerns:
- Policy or data-handling questions:
- Requests for additional support:
- Examples of applied learning, if approved:

Produce:
1. A theme summary grouped by role or pathway.
2. A severity ranking for barriers that affect safe, accessible, or effective participation.
3. Recommended changes to communications, office hours, manager coaching, practice tasks, or accessibility support.
4. Policy questions that require human owner review.
5. Feedback that cannot be interpreted because data is missing, anecdotal, or too sparse.
6. A follow-up survey or interview question set that avoids collecting sensitive personal records.

Required inputs

• Redacted learner feedback and voluntary comments.
• Audience group and pathway metadata approved for analysis.
• Accessibility, academic-integrity, security, and data-handling concerns raised during the program.
• Support requests from office hours or application sessions.
• Known limits in response rate or representativeness.

Expected output

The output should identify practical improvements, such as clearer launch instructions, more protected learning time, additional office-hours sessions, better examples for a specific role, or accessibility changes. It should separate high-severity blockers from preferences and should flag policy questions for human owners rather than inventing new rules.

Verification checkpoint

A program lead should confirm that feedback summaries do not expose identifiable learner information and that proposed changes are feasible under local policy. Accessibility-related changes should be reviewed by the appropriate accessibility or learner-support owner before publication or implementation.

This article offers 25 ChatGPT prompts for interactive learning and study, covering quizzes, flashcards, spaced repetition, adaptive study plans, and self-assessment. The 25 ChatGPT Prompts for Interactive Learning and Study: Quizzes, Flashcards, Spaced Repetition, and Adaptive Study Plans article is a focused companion for Student Study Prompts because it is an exact semantic fit for student study prompts and complements the current article’s student learning-plan section with practical study workflows.

Prompt 23: Reassessment plan after initial Academy learning and applied practice

Purpose

Use this prompt after a cohort has completed initial learning activities and attempted permitted practice tasks. Reassessment should compare current capability against the baseline, review real artifacts where allowed, and identify remaining risks. It should not convert badges or completions into authorization for production deployment, regulated decisions, grading, employment decisions, or independent work without role-appropriate review.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping design a reassessment plan for a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Create a reassessment plan using:
- Original baseline skill areas:
- Academy pathway or courses completed:
- Practice tasks attempted:
- Applied-work examples approved for review:
- Manager, instructor, or reviewer observations:
- Known policy or quality concerns:
- Accessibility or support needs:
- Time available for reassessment:

Produce:
1. A reassessment rubric tied to the original baseline.
2. A practical task or artifact review for each audience group.
3. Criteria for "needs more practice," "ready for supervised use," and "requires specialist review."
4. A list of topics that require refresher learning.
5. A human-review workflow for disputed or borderline cases.
6. A conservative summary template that avoids credential, employment, grading, or deployment claims.

Required inputs

• Original baseline skills or learning objectives.
• Course completion or badge evidence, if available and approved for use.
• Permitted artifacts from real work, study, teaching, or technical practice.
• Reviewer observations from managers, instructors, mentors, or technical leads.
• Policy concerns and accessibility needs observed during the program.

Expected output

The output should provide a reassessment rubric that is grounded in observable behavior: better instructions, appropriate context, source checking, human review, secure handling of inputs, evaluation plans, or responsible use of AI in study and teaching. It should use conservative readiness language such as “ready for supervised practice” or “requires specialist review,” not “certified,” “approved,” or “fully qualified.”

Verification checkpoint

An authorized reviewer should examine a real permitted artifact before marking a learner, team, or cohort as ready for a new level of supervised practice. Any decision involving employment, grades, access privileges, production deployment, regulated work, security posture, or legal obligations must follow the organization’s formal process and cannot be delegated to the prompt output.

Prompt 24: Badge-use policy for Academy course badges

Purpose

Use this prompt to draft a badge-use policy that explains what an Academy course badge may show and what it must not be used to decide. OpenAI says passing a course assessment earns an OpenAI Academy course badge, but the source findings require a strict boundary: a badge is learning evidence, not a professional license, industry certification, guarantee of competence, safety certification, production approval, or causal proof that outcomes improved.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping draft a badge-use policy for a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Draft a badge-use policy that covers:
1. What an Academy course badge may indicate.
2. What a badge must not be used to authorize or decide.
3. How managers, instructors, team leads, and learners may reference badges.
4. What additional evidence is required for supervised work, production access, publication, grading, or regulated tasks.
5. How to handle missing, disputed, expired, or unverifiable badge evidence.
6. A short FAQ for learners and managers.

Use plain language and include a strong warning that badge status must not be treated as a professional license, production approval, employment decision, grade, safety certification, or causal proof of improved performance.

Required inputs

• Organization policy on credentials, learning records, and personnel or student records.
• Approved uses for completion evidence in coaching, learning plans, or voluntary portfolios.
• Prohibited uses, especially employment, grade, access, deployment, regulated, or disciplinary decisions.
• Reviewer roles for technical, academic, security, legal, HR, or compliance signoff.
• Retention and privacy expectations for badge or completion records.

Expected output

The output should be a practical policy draft with allowed uses, prohibited uses, evidence requirements, escalation paths, and learner-facing FAQ language. It should make clear that badge evidence can support a learning discussion, reassessment, or portfolio review, but cannot replace review of actual work by authorized humans.

Verification checkpoint

HR, legal, education, security, privacy, or compliance owners should review the policy before it is published or enforced. A manager or instructor should verify a real permitted task separately from badge status before allowing any expanded responsibility, and formal authorization must remain outside the prompt output.

Prompt 25: 8–12 week role-based AI learning program summary

Purpose

Use this prompt to produce the final 8–12 week deployment summary recommended by the Academy deployment guide. The summary should combine launch activity, participation, support, applied examples, feedback, reassessment, risks, and next steps. It should remain evidence-based, explicitly separate usage from causation, and preserve the badge boundary.

Copy-paste prompt

Mandatory CE106 learning safety contract: Use only redacted, synthetic, public, or organization-approved inputs. Do not include passwords, tokens, private student records, private employee records, medical records, legal records, financial records, confidential source code, protected assessment answers, or other sensitive data. A course badge is learning evidence only, not a professional license and not a production approval, employment decision, grade, or safety certification. An authorized human retains responsibility for the final decision. Identify uncertainty and missing input. Follow the organization's data, accessibility, academic integrity, security, and review policy. Verification checkpoint: an authorized human reviews one approved task.

You are helping draft an 8–12 week summary for a role-based AI learning program aligned with OpenAI Academy pathways: Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI.

Use only redacted, synthetic, public, aggregated, or organization-approved inputs. Do not ask for or include passwords, tokens, private student or employee records, medical/legal/financial records, confidential source code, protected assessment answers, or other sensitive data. Treat Academy course completion or a badge as learning evidence, not a professional license, production approval, employment decision, grade, or safety certification. Keep final decisions with the authorized human. Identify uncertainty and missing inputs. Follow the organization’s data, accessibility, academic-integrity, security, and review policies. Include a human verification checkpoint tied to a real permitted task.

Do not automate hiring, firing, promotion, grading, disciplinary, medical, legal, tax, financial, payment, credential, publication, or deployment decisions.

Create an 8–12 week program summary using:
- Program goals:
- Audience groups:
- Pathways used:
- Launch communications completed:
- Participation and completion data:
- Office hours and reinforcement activities:
- Applied-work examples approved for sharing:
- Usage metrics, if available:
- Feedback themes:
- Reassessment findings:
- Badge-use policy status:
- Risks, incidents, or unresolved questions:
- Recommended next phase:

Produce:
1. An executive summary.
2. A role-by-role participation and support summary.
3. A section separating participation, applied evidence, usage metrics, and causal claims.
4. A feedback and accessibility improvement plan.
5. A risk table.
6. A reviewer workflow for the next phase.
7. A badge-boundary reminder.
8. A next-8-weeks action plan with owners and checkpoints.

Required inputs

• Program goals and launch window.
• Audience groups and OpenAI Academy pathways used.
• Aggregated participation, completion, support, and reassessment information.
• Approved applied-work examples and feedback themes.
• Usage data, if available, with known limitations and confounders.
• Risk, incident, accessibility, policy, and badge-use updates.

Expected output

The output should be an evidence-based deployment summary that leaders can use for governance and planning. It should explain what happened, what learners tried, what support worked, what remains uncertain, and what the next phase should change. It should not claim that Academy participation caused business outcomes unless the organization has appropriate evidence beyond timing and usage metrics.

Verification checkpoint

The program owner, sponsor, learning lead, and relevant security, privacy, accessibility, HR, legal, academic, or technical reviewers should approve the final summary before it is shared broadly. Any external publication, budget commitment, access expansion, policy change, code deployment, classroom assessment change, or public credential claim requires separate authorized human approval.

Final operating guidance for Prompts 19–25

Prompts 19–25 move from individual learning design into program operations: launch communications, participation analysis, usage interpretation, feedback synthesis, reassessment, badge policy, and the 8–12 week summary. The common failure mode at this stage is overinterpretation. A team may have real engagement, useful examples, and enthusiastic feedback, yet still lack evidence that the program caused a measurable business, academic, quality, or safety outcome. Keep the language precise: participation is participation; completion is completion; a badge is course-assessment evidence; applied artifacts show examples of changed work; usage metrics show activity; causal claims require stronger design and review.

The OpenAI Academy sources describe role-specific pathways rather than a universal curriculum. Apply AI at Work is oriented toward knowledge-worker workflows such as instructions, context, reviewing responses, reusable workflows, agent delegation, checkpoints, and human review. Build with AI is for developers using Codex or building with the OpenAI API and includes planning, implementation, review, quality, solution design, evaluations, agents, retrieval, and operations. Lead AI Adoption emphasizes business value, ownership, governance, strategy, and roadmaps. Teach and Learn with AI covers educator and student use where learners and educators remain responsible for final work. These boundaries should appear in program communications, assessment rubrics, and summaries.

Risk	Where it appears	Operational control	Human reviewer
Badge overreach	Managers, learners, or sponsors treat a course badge as authorization.	Publish a badge-use policy stating that badges are learning evidence only.	HR, legal, compliance, academic, or program owner.
Private data exposure	Learners paste employee records, student records, source code, credentials, or regulated data into prompts.	Require redacted, synthetic, public, or organization-approved inputs and provide safe practice tasks.	Security, privacy, data owner, or instructor.
Causal overclaim	Usage rises after launch and is described as proof of productivity or quality improvement.	Separate usage, participation, applied examples, quality review, and causal evidence.	Analytics owner, sponsor, governance lead.
Consequential automation	Prompt outputs are used for hiring, grading, deployment, payments, publication, or regulated decisions.	Require formal human approval and prohibit automated consequential decisions in every prompt.	Authorized decision owner and relevant policy function.
Accessibility gap	Learners cannot participate because formats, timing, or support are inaccessible.	Collect feedback, provide alternative formats where approved, and document accommodation paths.	Accessibility or learner-support owner.
Unreviewed technical use	Developers treat Build with AI learning as approval to ship code, agents, retrieval systems, or production changes.	Require code review, security review, evaluation, deployment gates, and rollback planning outside the course record.	Engineering lead, security lead, product owner.

Reviewer workflow for the final learning-program package

1. Program owner review: Confirm that the launch, participation, feedback, reassessment, and summary materials accurately represent the approved program scope and do not promise outcomes that were not measured.
2. Pathway owner review: Check that Apply AI at Work, Build with AI, Lead AI Adoption, and Teach and Learn with AI are assigned to appropriate audiences and that no learner group is pushed into unrelated material without a clear reason.
3. Privacy and security review: Verify that examples, survey comments, artifacts, and metrics are redacted, aggregated, or approved for use, and that no credentials, confidential source code, private records, or regulated data are exposed.
4. Accessibility and academic-integrity review: For schools, universities, training teams, and learner-support groups, confirm that participation routes are accessible and that student use respects assignment requirements, permitted materials, and final human responsibility.
5. Badge-boundary review: Confirm that every mention of completion or badges states the correct boundary: learning evidence only, not a license, safety certification, production approval, employment decision, grade, or proof of causation.
6. Analytics review: Validate participation numbers, reporting gaps, cohort definitions, time windows, and confounders before any usage or business-value statement is shown to sponsors.
7. Authorized-decision review: Route any action involving external communication, publication, spending, hiring, grading, access expansion, code deployment, security change, or legal commitment to the formal human approver.

A useful 8–12 week summary should include enough evidence to guide the next phase without pretending that the first phase answered every question. A strong summary names the pathways used, describes who participated, gives examples of permitted applied work, identifies support gaps, explains what usage data can and cannot show, and proposes a narrower next set of practices. A weak summary celebrates completion numbers without evidence of changed behavior, omits feedback from underrepresented groups, treats badges as credentials, or recommends production use without review.

For the next phase, use a staged operating rule. Continue broad learning access where policy allows, but require stricter evidence as the action becomes more consequential. A learner can experiment with a synthetic document after basic instruction; a team changing a customer workflow needs manager and policy review; a developer shipping an agent, retrieval system, or API integration needs engineering, security, evaluation, and deployment review; an educator changing assessment practice needs academic review; a leader claiming business impact needs an analytics design that addresses confounders and missing data.

Operational rule: OpenAI Academy learning can support capability development, but it does not replace governance, policy, review, evaluation, source verification, accessibility support, academic integrity, security controls, or authorized human judgment.

Useful Links

• OpenAI: Expanding OpenAI Academy with new learning paths
• OpenAI Academy course catalog
• OpenAI Academy Courses Champion Deployment Guide
• OpenAI Help Center: OpenAI Academy courses