# AI Teachers — Free workshop: Delegate your daily work to AI Updated 2026-10-25 · Free written training workbook Use invented or permitted data. Your AI tool may require a separate subscription. Download this workbook and save answers locally; this course site saves only completion checkmarks. ## The outcome A practical, tool-neutral introduction to Claude Cowork, ChatGPT Work and AI agents. No paid AI account is required to complete the fictional exercise. ## What you will learn - Choose a realistic, low-risk task - Write a clear delegation brief - Run and review one workflow - Choose whether an agent is justified ## Lesson 1: 00–10 min · Pick the right work (10 min) The best first delegation is frequent, bounded, and easy for you to evaluate. ### Follow these steps 1. Write down three tasks you repeat every week. 2. Score each 1–3 for repetition, stable inputs, easy review, and low error cost. 3. Choose the safest task; do not select sensitive client communications or irreversible transactions. 4. Record how long it takes you today, including revisions. ### Copyable prompt Compare these three work tasks for a beginner AI workflow: meeting action summary, weekly status report, and replying to a customer complaint. Score repeatability, input clarity, approval needs, failure cost, and review effort. Recommend a low-risk first pilot and explain why. ### Deliverable A chosen task, success measure, baseline minutes and clear human owner. ### Quality gate The result is observable and a human can approve it before use. ### Watch out Avoid claiming any time savings before measuring the complete process. My result: What I verified: What I will change: --- ## Lesson 2: 10–25 min · Build the brief (15 min) Agents work more reliably when the input, output, quality rule, and permissions are explicit. ### Follow these steps 1. Use invented project notes or a document you have permission to share. 2. Name the audience, objective, source of truth, required sections, length and deadline. 3. List forbidden actions and what the AI must do when a fact is missing. 4. Require a draft first, with sources and open questions. ### Copyable prompt Act as an operations assistant. Use only the following fictional notes: [PASTE NOTES]. Produce a one-page action summary with decisions, owner, due date, blocker and source quote for each item. Do not invent owners or dates. Mark gaps as NEEDS CONFIRMATION. Draft only; do not send messages or change any systems. ### Deliverable A reusable delegation brief with a review/approval gate. ### Quality gate All factual claims trace back to the supplied notes. ### Watch out A polished answer can still be wrong; factual evidence matters more than formatting. My result: What I verified: What I will change: --- ## Lesson 3: 25–40 min · Do a first run (15 min) Compare the result to the source material, not to what sounds plausible. ### Follow these steps 1. Open an approved AI tool you already use, or work through the example on paper. 2. Paste your brief and permitted fictional notes; ask for an output. 3. Compare the summary, owner and dates to the notes line by line. 4. Request one correction and save the revised version. ### Copyable prompt Review the draft above against the source notes. For each action show the supporting source text or mark UNVERIFIED. List contradictions, missing decisions and the one question a human needs to answer. Do not silently fix uncertain facts. ### Deliverable A first draft and a separate reviewer pass. ### Quality gate At least three source-to-output comparisons; unknowns visibly labeled. ### Watch out Do not grant live posting, file deletion, spending or sending permissions in this first exercise. My result: What I verified: What I will change: --- ## Lesson 4: 40–50 min · Turn the task into an agent-ready workflow (10 min) Only after a successful pilot should you consider connectors, scheduling or specialist agents. ### Follow these steps 1. Name the trigger: manual request, a scheduled review, or an approved event. 2. Map where inputs come from and where reviewed results go. 3. Identify what the agent can do independently versus what needs approval. 4. Write two failure cases: incomplete notes and contradictory dates. ### Copyable prompt Turn my tested action-summary brief into an agent workflow specification. Include trigger, authorized input, data restrictions, draft output, approval step, two failure tests, logging, stop rule, and rollback. Keep all external actions in proposal mode until I approve them. ### Deliverable One simple workflow map, approval boundary, and two test cases. ### Quality gate The draft stays useful even without a connector or premium feature. ### Watch out An AI prompt does not technically enforce permissions; settings and application controls must. My result: What I verified: What I will change: --- ## Lesson 5: 50–60 min · Keep, improve or stop (10 min) Good AI adoption measures accepted work, total effort and failure handling. ### Follow these steps 1. Add your setup, correction and review minutes. 2. Check whether output quality meets the rule you wrote. 3. Decide keep, revise or stop for this task. 4. Download the workbook and choose the next specialist course. ### Copyable prompt Evaluate this pilot using my baseline minutes, AI setup time, review time, errors and accepted output. Present keep, improve or stop, with one next experiment and an explicit list of unknowns. Avoid projecting savings from a single run. ### Deliverable Completed workbook and one measurable next experiment. ### Quality gate An accountable human owns the result and the next step. ### Watch out Never interpret a workshop completion checkmark as certification or a professional qualification. My result: What I verified: What I will change: --- ## Final review - [ ] Every material fact has a permitted source. - [ ] Missing data is visible. - [ ] External actions still require authorized approvals. - [ ] I have tested at least one failure case. - [ ] I know who owns the result and how to stop the workflow. Official reference links: - ChatGPT Work and Codex overview: https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex - Claude Cowork overview: https://support.claude.com/en/articles/15520349-use-claude-cowork-on-web-desktop-and-mobile AI Teachers · https://aiteachers.cc/academy/delegate-daily-work/