Turn each expert’s know-how into a course.

What the AI will do

• Investigates the learning need, helps identify the expert, and offers relevant areas of expertise for you to choose from.
• Draws out real judgments through an expert interview, a shareable AI session, or research into a public figure’s work, then writes the lessons and practice cases.
• Offers different simulated learner types to test the course in separate agent sessions where available, revises the material, and prepares a real learner trial.

What you can provide

• A capability people need, an expert you have in mind, or a rough work problem; the AI helps establish what is missing.
• Whatever useful material is already available: project documents, examples, correspondence, transcripts, or public sources. The expert can be you, someone you can work with, or a public figure you do not know.

Where you'll take part

• Choose the expertise to teach and how to work with an accessible expert: together, through an interview you conduct, or through a self-contained artifact they use on their desktop.
• You or the accessible expert correct the account of their expertise and the course. A separate commissioning person reviews it when there is one; for a public figure, you review the source-based course’s relevance.
• Choose the types of simulated learners you want, or skip them. Where possible, arrange a real learner’s attempt and return the responses and human assessment for revision.

What you'll get

A practical course with worked examples, learner practice, separate assessment guidance, and a clear account of its sources, revisions and actual testing. Any outstanding expert review or real learner trial remains explicit; a public-source course does not imply the public figure participated.

Activity context

Experts often notice things they never think to explain. This activity helps uncover those judgments and turn them into a course others can use, beginning with the expertise your learners need and the person who can teach it. That person can be you, an accessible expert, or a public figure whose work the AI can research.

For the user

Start with a rough learning need, an expert’s name, or a work problem. The AI investigates relevant available context, offers areas of expertise to choose from, and helps select an interview or public-research route. You choose the focus and simulated learner types, review the course with the people involved, and return actual learner feedback when a trial is possible. The AI prepares the questions, any shareable handoff, the course, and its revisions.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Help me turn expertise into a practical course that teaches people how to make difficult judgments, not just repeat an expert’s conclusions. Establish the capability the learners need and whose expertise could teach it, then help us choose the route below. You own the investigation, interviewing support, course writing, checks, and revisions. Keep my contribution focused on choices, missing knowledge, and actual review.

Identify the need and the expert

Start with whatever I have supplied, including a rough idea, an expert’s name, a work problem, or existing material. Inspect and try relevant connected context before asking me to reconstruct it. Search project plans, work products, training material, emails and chats for decisions people struggle with, who handles them well, and what learners already know. Use meeting transcripts to find explanations of difficult judgments; calendar invitations and project roles can identify the people and work to investigate. Search correspondence about those people’s contributions and learners’ recurring questions to distinguish documented expertise and learning needs from reputation or your inference.

Keep this first investigation short and directed toward a useful choice. Establish what learners should be able to do afterward, who they are and their existing expertise, and who the expert is. Consequential examples include translating enterprise policy into operating decisions, judging whether data is fit for a major investment decision, and planning a business-critical IT migration. These are examples, not the required subjects. If the record leaves a necessary choice unresolved, ask one focused question at a time and wait.

Offer the routes that fit what we know:

• I am the expert. Interview me directly and help turn my judgments into teachable material.
• I have access to the expert, such as someone I work with. They can sit with me and answer your questions, I can interview them with your help, or you can create a self-contained artifact for them to use in their own AI conversation on their desktop.
• The expert is a public figure I do not know. Research their work and the evidence about it to create a source-based course; their direct participation is not assumed.

Confirm the route and the actual expert rather than selecting one on my behalf. If I know the need but not the expert, use the available records or public professional material to propose candidates with the evidence for each, then let me choose. If I know the expert but not the subject, use their relevant work to help identify it.

If connections fail, inspect available access or export options and help with the smallest specific user action needed. Accept pasted material, links, recordings you can actually inspect, or rough notes. With no accessible internal record, use a focused interview to establish the need and learner baseline; mark its evidence limits. For a public figure, accessible source material is necessary for attributing expertise or teaching to that person. Ask for usable source text if browsing fails rather than filling the gap from an imagined interview.

Choose the expertise to teach

Using this first evidence pass, promptly present a short list of relevant areas of expertise. For each, name the capability, give a concrete decision it would help a learner make, and show the expert or source evidence supporting it. Separate evidence from a promising but unconfirmed possibility.

Invite me to select one, two, several, or none of these areas, combine them, or propose another. Wait for my selection before deep extraction or course drafting. If I choose none, use my response to revise the shortlist. For multiple areas, propose a manageable sequence with separate learning outcomes and clear boundaries; confirm material scope changes. Do not quietly expand a short course into a complete academy.

Extract the difficult judgments

For an accessible expert, use the selected participation route. The expert’s own words and corrections remain distinct from mine when we are different people.

• Together in this conversation: identify who is answering, then interview the expert one question at a time.
• I interview the expert: prepare a concise, tailored interview guide from the actual cases and evidence, with opening questions and useful follow-up probes. Help me capture the conversation in a usable form. Read the resulting transcript, recording, or notes yourself, identify the judgments still unexplained, and prepare only the necessary follow-ups. Do the synthesis and course drafting yourself.
• The expert works on their own desktop: create a complete downloadable or copyable handoff for a fresh AI session. Include the agreed learners, outcomes, selected expertise, relevant source excerpts or accessible references, known decisions, open questions, speaker roles, and the full interview-to-course method in these instructions. Write the actual tailored prompt, not a reference to this conversation or an instruction to ask me for a prompt. Give it directions to inspect its own access, interview one question at a time, wait for expert corrections, draft and revise the course, and keep assessment material separate. Include the later commissioner-review, selected simulated-learner, and actual-trial stages, their pauses, and return instructions. The first return package is the expert-corrected course plus evidence and questions; I bring the commissioning response back to that session, or return the package here for you to continue. Make the selected handoff route explicit, keep a concise progress record with the package, and do not duplicate work across sessions. Hold the dependent stages while an actual response is pending.

In every accessible-expert route, start with a real consequential case the expert handled. Use relevant work and transcripts to make the questions concrete. Explore the evidence they used, options they considered, tradeoffs, people involved, result, and how they judged it. Ask what they noticed that a less experienced person might miss. Follow their reasoning at the moments where alternatives diverged. Seek a contrasting case where the same approach failed or had to change, and identify the conditions that explain the difference. Documents support this interview; they do not replace eliciting tacit judgment. Summarize the account with clear separation between expert statements and your inferences. Wait for the expert’s corrections or explicit confirmation before writing lessons.

For a public figure, search their authored books or articles, talks, interviews, demonstrations, and available transcripts, then independent reporting, analysis, and criticism about their work. Build a dated source record with links and useful page references or timestamps. Follow concrete cases through evidence, choices, tradeoffs, and outcomes; compare cases that expose limits or changes in their approach. Distinguish the expert’s stated reasoning from observers’ accounts and your reconstruction. Check important claims against the underlying material, resolve conflicting versions where possible, and show where the record cannot reveal a private judgment. Do not invent first-person testimony, impersonate an interview, or imply endorsement. Present the resulting expertise map and its limits to me for scope and interpretation corrections before drafting. My agreement does not make an inference the expert’s own view. This route needs source verification, not approval from an inaccessible public figure.

For a large body of work, divide bounded source groups among agents where useful and available. Give each sources, selected questions, required evidence and a completion criterion; retain a coverage record and reconcile findings against originals. Work in waves if needed. With no agents, use manageable passes or fresh sessions with explicit source-and-decision handoffs; disclose shared-context review. Do not treat an uninspected corpus as completed research.

Write and review the actual course

Write usable lessons for the selected capabilities, including a worked case that reveals the reasoning, new practice scenarios requiring transfer, and separate assessment guidance. Explain unfamiliar terms without simplifying away difficult decisions. Connect each lesson and assessment to the intended capability. Preserve uncertainty and explain where different choices can be defensible. Label invented scenarios as fictional; do not invent internal practices. For specialized, changing, or disputed outside facts, consult reliable current sources and link them, or identify exactly what remains unchecked.

The learner-facing materials must stand on their own. Keep model answers, scoring criteria that reveal solutions, and reviewer rationale in a separate assessor document. Assess the quality of evidence use, judgment and tradeoffs, not merely recall or agreement with one preferred answer.

In the accessible-expert route, have the expert check the actual lessons, worked cases, scenarios and assessment guidance. Wait for corrections or explicit confirmation; resolve consequential gaps through further interviewing and revise. If a different person commissioned the course, prepare a concise review package for them: the expert-corrected course, separate assessment guidance, intended outcomes, and unresolved decisions. Wait for their actual response or explicit acceptance, then address it and have the expert check changes to expert judgment. If I am both expert and commissioner, make that explicit and combine those reviews rather than asking me to pretend to be two people.

In the public-source route, verify course claims and attributed reasoning against the cited sources, using a separate source reviewer where available. Give that reviewer the actual course and evidence, not a defense of your intentions. Resolve material findings and have me review relevance and use for the intended learners. If I am commissioning for someone else, obtain that person’s actual response as above. No stage may require the public figure’s participation.

Let me choose simulated learners

Offer a small set of useful simulated learner types adapted to the actual learner evidence, then wait for my choice of one, several, alternatives, or none. Possible types include someone new to this domain, a capable colleague transferring from another function, an experienced skeptic who challenges assumptions, someone who knows the terminology but applies rules too mechanically, and a time-pressed practitioner. Explain what each would help test. Derive their knowledge, work context, concerns and likely misconceptions from training records, work examples, transcripts or other learner evidence where available. Label added characteristics as simulation assumptions; do not present a fictional profile as a real person’s response.

For each selected type, use a separate fresh-context subagent if available. Give it only its role brief, the learner-facing lessons, and a new test scenario. Withhold the answer key, assessment guidance, author discussion, other learners’ answers, and intended reviewer conclusions. Ask it to attempt the test first, showing its choices and reasons, then report confusing passages, missing knowledge and questions. Save the attempt before assessment or feedback is supplied. If a brief already assumes mastery of something being tested, correct that mismatch before running it.

Have a separate assessor, where available, compare the completed attempts with the assessment guidance and intended learning outcomes. Require specific evidence from each attempt and the actual lessons. Reconcile disagreements and distinguish weak teaching, ambiguous tests, missing prerequisites, and plausible alternative reasoning. Revise the course and tests accordingly; have the accessible expert check changes to expert judgment, or source-check them in the public route. Where a material issue remains, test the revision with a fresh learner context and a new case. Avoid claiming that agreement among agents proves the course works.

If subagents are unavailable, offer self-contained briefs for separate fresh sessions that I can run, or a clearly labeled less-independent review in the current session. Do not quietly give the answer key to a supposed learner. Keep any user-run simulation pending until its actual output returns; if I decline simulated learners, proceed without claiming this review occurred.

Try it with a real learner and finish

Simulated learners help refine the materials; they do not fulfill the real learner trial. Prepare a practical trial kit: learner-facing lessons and a new case, a separate assessor guide to withhold until the attempt, and a short observation guide covering where the learner gets stuck, choices and reasons, evidence noticed or missed, and demonstration of the intended capability. Ask me to arrange an available intended learner’s actual attempt and return their own answers and observations.

With an accessible expert, that expert assesses the learner’s reasoning; you help interpret the evidence and revise. In the public-source route, establish who can assess the reasoning: an accessible person with relevant expertise where possible. If only my review and your source-based assessment are available, state that limit and keep expert validation unperformed. A public figure is never presumed to have assessed the learner.

Wait for the actual attempt and assessment. Use them to revise lessons, examples or tests, explain substantive changes, and obtain the relevant human accuracy check. If the learner has not demonstrated the capability to the assessor’s satisfaction, prepare a focused revision and another new case, and await the next actual attempt. If a trial, review, or qualified assessment is unavailable, deliver the useful work completed with the exact pending stage rather than inventing evidence or withholding the whole course.

Deliver the course, learner practice material, separate assessment guidance, source and expertise record, and a brief review account: which human and simulated reviews actually happened, what the learner demonstrated, revisions made, and remaining gaps. In the public route, describe it as a source-based course about the expert’s work. Prepare it in a usable format and help place it where the intended learners can access it within my authorization. Distinguish a drafted or reviewed course from one tested with a real learner, and a real learner trial from validation by the original expert.

Done when

You have a usable course for the selected capability, learner practice material and separate assessment guidance, with the relevant human reviews and revisions completed. Full learner-tested completion includes a real learner reasoning through a new case to the appropriate assessor’s satisfaction and the revised course being available to its intended learners. If access, expert review or a real trial is missing, you receive the completed material with that stage explicitly pending. Public-source work remains identified as such, and simulated learners do not count as a real trial.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.