SPOSABLE — YOUR FIRST SIX USEFUL AI WORKFLOWS Use fictional sample data first. Use an AI tool you are permitted to use. Do not share confidential information. Clarify → Gather context → Create → Check → Improve → Reuse Times are practice estimates, not guaranteed completion times. Worked examples follow each exercise; try before reading. ============================================================ 1. Turn messy notes into a clear brief — approximately 10 minutes LEARNING GOAL Separate facts, proposals and unknowns, then create a short brief with a clear decision request. IMAGINE A colleague hands you scattered notes. You need a brief another person can act on, not a prettier pile of words. You’ll turn a fictional onboarding idea into a decision-ready first draft. Your result: a brief under 180 words and a reusable brief-making template. SAMPLE INPUT FICTIONAL NOTES — HARBOR WORKSPACE N1. New members sometimes ask where to find the booking guide. We have not counted how often. N2. Idea: try a welcome sheet with a guide link. Not approved yet. N3. Available print budget for a trial: 60 euros. N4. No launch date agreed. N5. Jo offered to draft the sheet, subject to manager approval. N6. Proposed trial: two weeks at the front desk. N7. We do not yet know who will collect feedback. N8. Manager needs to decide whether to approve a small trial, not a permanent rollout. TRY Minute 0–2: read the sample. In your own words, name the decision the manager needs to make. Notice that a suggestion is not an approved plan. Minute 2–5: open a fresh AI conversation. Copy the starter prompt, followed by the sample input, into one message. Read the draft before asking for changes. Minute 5–8: check every name, number, date and commitment against the numbered notes. Ask for one targeted revision, such as making a proposed date explicitly unknown. Minute 8–10: edit the decision request yourself and save the final brief in your document app. Use the checklist before marking the lab complete. STARTER PROMPT Turn the SOURCE NOTES below into a decision brief of at most 180 words for the workspace manager. Use headings: Purpose, Known facts, Proposed trial, Open questions, Decision requested. Separate facts from suggestions. Include note IDs beside factual claims. Do not invent evidence, dates, owners, approvals or expected savings. Preserve conditions such as “subject to approval.” Flag missing information instead of filling it in. SOURCE NOTES: [Paste the sample input here.] CREATE Compare your draft with the source, not with how confident the writing sounds. If it says “Jo will launch next week,” correct it: the offer is conditional and no date is agreed. Ask: “Revise only the unsupported commitments. Keep the decision request to one sentence.” Then make the final edits yourself. Your saved brief should be understandable without the chat history. UNDERSTAND Fluent writing can hide a change in meaning. The useful skill is giving AI a purpose and checking whether it preserved the status of each claim. A proposal, an approval and a completed action are different things. Before you reuse this workflow, identify the facts it must not silently change. The worked example is one defensible answer, not the only wording that can pass. CHECK YOUR RESULT The decision request is explicit and asks about a trial, not a permanent rollout. The 60-euro budget and two-week proposal match the source. No launch date, approval, measured frequency or guaranteed saving is invented. Jo’s offer remains conditional, and the missing feedback owner is visible. The brief is under 180 words and understandable without the original chat. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: Turn these fresh notes into a brief: “Team wants a shared FAQ; no tool chosen; Lee can suggest questions; we have 45 minutes for a trial discussion; approval not requested yet.” Write your own instruction, then apply the same evidence and commitment checks to the new facts; do not reuse the original sample’s budget or timing. Compare the statuses of “wants,” “can” and “approved.” Save one sentence explaining what you corrected. REUSABLE TEMPLATE WORKFLOW: CLEAR BRIEF Use when: scattered notes need a decision or next step. Inputs: numbered source notes; audience; desired decision; length limit. Method: label facts, proposals and unknowns; draft headings; check each specific claim against its source; revise; save a standalone brief. Never invent: owners, dates, approvals or evidence. Done when: the reader can name the decision and the remaining unknowns. My preferred headings/tone: __________ My most common failure to check: __________ WORKED EXAMPLE — TRY FIRST Purpose: Decide whether to test a welcome sheet at Harbor Workspace. Known facts: Some new members ask for the booking guide; frequency has not been measured (N1). A 60-euro print budget is available for a trial (N3). Proposed trial: Put a sheet with the guide link at the front desk for two weeks (N2, N6). Jo has offered to draft it, subject to approval (N5). Open questions: Who collects feedback? When would the trial start? What observation would justify continuing? No date or feedback owner is agreed (N4, N7). Decision requested: Approve or decline the two-week trial, and if approved, name a feedback owner and agree a start date (N8). Why it works: it preserves uncertainty and asks for a decision without presenting the trial as already approved. REMIX Change the audience to a colleague who will run the trial. What extra information do they need? Keep unknowns visible. Share the brief through your normal channel only after reviewing it. No real member information is needed for this exercise. ============================================================ 2. Find the story in customer feedback — approximately 15 minutes LEARNING GOAL Group qualitative feedback into traceable themes, verify counts and distinguish a pattern from a representative conclusion. IMAGINE A pile of feedback can make the loudest comment feel like the whole story. You’ll practise on eight fictional comments and create a short insight brief with supporting evidence. Your result: a theme summary and one small experiment to investigate, not a claim about all customers. SAMPLE INPUT FICTIONAL MEMBER COMMENTS — HARBOR WORKSPACE C1. I like the space, but finding the booking guide took too long. C2. A picture showing how to book a room would help. C3. The front-desk team answered my question quickly. C4. I could not find the first step for booking a room. C5. The welcome email arrived a day late. C6. The front-desk team was friendly. C7. A short getting-started guide would help me book a room. C8. The workspace was clean when I arrived. TRY Minute 0–3: read the sample before involving AI. Mark the comment IDs that concern finding or following booking instructions. Minute 3–7: submit the prompt with the sample. Ask for categories and comment IDs, not only an executive summary. Minute 7–11: recount the proposed groups manually. Check whether any comment is omitted, duplicated without explanation, or interpreted too strongly. Minute 11–15: write a three-sentence brief with an observation, evidence and a proposed test. Save it with the source comments so someone else can check it. STARTER PROMPT Group the numbered fictional comments below into a small number of themes. For each theme give: a neutral label, exact supporting comment IDs, number of distinct comments, and one short exact quote. A comment can appear in more than one theme, but explain overlap and do not count it twice within one theme. Separate observations from interpretations. Do not invent comments, sentiment or population-level conclusions. Suggest one small test for the clearest problem, with uncertainty stated. COMMENTS: [Paste the sample input here.] CREATE If AI puts C3 under “confusing instructions,” challenge that classification: C3 describes a quick answer, not the cause of the question. Ask it to revise using explicit evidence only. Write three sentences: what this sample suggests, which comments support it, and what you would test. Do not describe this as a survey of all members. UNDERSTAND Theme names involve judgement; exact wording can differ. Counts and cited IDs must still be reproducible. Four of these eight selected comments concern booking guidance, but the sample is too small and its selection too unclear to estimate a population percentage. An AI summary is useful when it makes the evidence easier to inspect, not when it hides it behind a confident conclusion. CHECK YOUR RESULT Booking guidance includes C1, C2, C4 and C7; any different grouping is explicitly justified. Every count equals the number of distinct supporting IDs. Quoted words match the source, and any overlap is explained. The brief does not generalise this sample to all members. The proposed next step is a small test tied to the evidence, not a guaranteed fix. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: Classify these new notes: A “Opening hours are hard to find”; B “I need weekend hours”; C “The staff are kind”; D “Where are opening hours listed?” Decide whether A/B/D belong together or in separate findability and availability themes. Explain your choice, count the IDs and suggest what you would need to learn next. Different defensible groupings can pass. REUSABLE TEMPLATE WORKFLOW: EVIDENCE-BACKED THEMES Use when: a small set of comments needs an inspectable summary. Inputs: anonymised, numbered comments and the question you want to explore. Output: theme / IDs / distinct count / exact quote / uncertainty / possible test. Review: recount IDs, verify quotes, inspect ambiguous classifications, explain sample limitations. Done when: another person can reproduce each count and trace the recommendation to evidence. My category rules: __________ My next research question: __________ WORKED EXAMPLE — TRY FIRST One valid grouping: • Booking guidance: C1, C2, C4, C7 — 4 comments. Quote: “A picture showing how to book a room would help.” • Helpful front desk: C3, C6 — 2 comments. Quote: “The front-desk team was friendly.” • Welcome-email timing: C5 — 1 comment. • Cleanliness: C8 — 1 comment. Three-sentence brief: Four comments in this small sample concern finding or following room-booking guidance. C1, C2, C4 and C7 support that theme, including a request for a picture of the booking process. We could test a clearer guide with a few new members and observe whether they can find the first step; this sample does not establish how common the problem is across all members. Why it works: the recommendation is a test, and the supporting IDs make the interpretation inspectable. REMIX Apply the method to five non-sensitive meeting questions. Ask for “open questions” rather than “customer themes.” Before using real feedback in an external tool, remove identifiers and check that your organisation permits that use. ============================================================ 3. Turn a meeting into decisions and next steps — approximately 10 minutes LEARNING GOAL Distinguish agreed decisions, confirmed actions and suggestions; preserve missing owners and dates rather than inventing them. IMAGINE A polished recap can accidentally turn “someone should” into a commitment. You’ll make a concise follow-up from a fictional meeting. Your result: a decision log, an action list and a draft message that people can verify before you send it. SAMPLE INPUT FICTIONAL MEETING NOTES — WELCOME-SHEET TRIAL M1. Manager: “Let’s approve a two-week front-desk trial, with printing up to 60 euros.” M2. Jo: “I will prepare a first draft by Thursday.” M3. Manager: “We have not agreed the start date yet.” M4. Sam: “A QR code might be useful.” M5. Jo: “I can ask about a QR code, but I am not committing to adding it.” M6. Manager: “Someone needs to collect feedback. We’ll decide who at our next check-in.” M7. No date for that check-in appears in these notes. TRY Minute 0–2: underline the explicit approval and the one confirmed delivery commitment in the sample. Minute 2–5: submit the prompt and sample in a new conversation. Read the action list against the actual words spoken. Minute 5–8: find any invented owner, calendar date or agreement. Ask for one precise correction, then edit the follow-up yourself. Minute 8–10: save the reviewed log and draft message. This lab does not send anything or create calendar events. STARTER PROMPT Create a meeting follow-up from the notes below. Separate: 1) confirmed decisions, 2) confirmed actions with owner, stated due date and source ID, 3) suggestions not agreed, 4) unresolved questions. Use “not agreed” where an owner or date is missing. Preserve relative dates exactly; do not convert “Thursday” without knowing the meeting date. Do not treat “might,” “can” or “someone should” as commitments. End with a draft message under 100 words asking participants to confirm or correct the summary. Do not send it. NOTES: [Paste the sample input here.] CREATE Challenge any line that assigns feedback collection to Sam or promises a QR code. Neither is supported. Ask: “Show the source phrase that supports each confirmed action; move anything unsupported to open questions.” Then remove internal note IDs from the outward-facing message if they would confuse recipients, while keeping them in your review copy. UNDERSTAND A meeting recap is an account of decisions, not an opportunity to make new ones. Words of possibility and words of commitment must stay distinct. Dates also require context. A model can produce a plausible date for Thursday without knowing which week the meeting happened. Missing information should remain missing until someone confirms it. CHECK YOUR RESULT Only the trial and print limit are listed as confirmed decisions. Jo’s draft is the only confirmed delivery action, with Thursday left relative. The QR code remains a suggestion, not a promised deliverable. Start date, feedback owner and check-in date are visibly unresolved. The follow-up invites correction and is saved as a draft, not sent automatically. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: Summarise this new exchange: “Pat: I could review it. Morgan: I’ll email the outline tomorrow. Team: Let’s postpone launch until we know the venue.” What is actually committed? Preserve “tomorrow” unless a meeting date is provided. Write one clarification question for Pat. REUSABLE TEMPLATE WORKFLOW: DECISIONS, ACTIONS, UNKNOWNS Inputs: permitted meeting notes, date/context if known, intended recipients. Extract: decisions / confirmed actions / unagreed suggestions / open questions. For every action: owner, stated due date, supporting phrase. Review: test commitment wording; never infer consent, owners or calendar dates. Output: reviewable log plus a short draft follow-up inviting corrections. Human step: confirm sensitive details before sending. My team’s preferred format: __________ WORKED EXAMPLE — TRY FIRST Decision: Approve a two-week front-desk trial with a printing limit of 60 euros (M1). Confirmed action: Jo — prepare first draft — due Thursday, calendar date unconfirmed (M2). Suggestion: Consider a QR code; including one is not agreed (M4–M5). Open questions: Trial start date; feedback collector; next check-in date (M3, M6–M7). Draft message: “Thanks, everyone. We agreed to a two-week front-desk trial with printing up to 60 euros. Jo will prepare a first draft by Thursday; please confirm the calendar date. The start date, feedback owner and next check-in date are still open. A QR code was suggested but not agreed. Please correct anything I’ve missed or misrepresented.” Why it works: it makes the unknowns actionable without assigning them to people who did not agree. REMIX Use the method after a personal planning conversation or volunteer meeting. Compare the recap with the other person’s recollection before treating it as a shared plan. ============================================================ 4. Make a recommendation you can defend — approximately 15 minutes LEARNING GOAL Compare options against explicit constraints, verify simple calculations and make a conditional recommendation with traceable evidence. IMAGINE AI can make every option sound like the obvious winner. You’ll give it clear criteria and three fictional venue factsheets, then review the recommendation yourself. Your result: a short comparison and a defensible decision memo, with an important unknown kept visible. SAMPLE INPUT FICTIONAL FACTSHEETS — NOT REAL VENUES OR CURRENT QUOTES Need: a two-hour workshop for 18 people; room-hire budget no more than 200 euros; step-free access required; prefer a whiteboard. Compare room hire only; no information on taxes, deposits, extras or availability is provided. A. Alder Room: 24 seats; 80 euros/hour; step-free access confirmed; whiteboard included. B. Birch Room: 20 seats; 65 euros/hour; step-free access not stated; whiteboard included. C. Cedar Room: 16 seats; 70 euros/hour; step-free access confirmed; no whiteboard. Factsheet IDs are A, B and C. There are no source URLs because these are fictional training inputs. TRY Minute 0–3: identify the hard requirements. Calculate each two-hour room-hire cost yourself. Minute 3–7: submit the prompt and factsheets. Ask the AI to distinguish known failures from unknown information. Minute 7–11: verify the arithmetic and inspect whether it treated “not stated” as either yes or no. Neither is justified. Minute 11–15: write a recommendation under 150 words with the condition that would change your choice. Save it with the comparison. STARTER PROMPT Compare the three fictional room options against the stated requirements. Show the two-hour room-hire calculation for each. Label each hard requirement as met, not met or unknown, with its factsheet ID. Treat missing access or availability information as unknown. Do not invent prices, fees, URLs or current availability. Recommend the best-supported provisional option, identify what to confirm before booking and explain what new evidence could change the choice. Do not book anything. REQUIREMENTS AND FACTSHEETS: [Paste the sample input here.] CREATE If AI recommends Birch solely because it is cheapest, ask what evidence supports the access requirement. If it excludes Birch as inaccessible, ask where the source says that. Ask for a revised recommendation that is conditional rather than overconfident. Edit the final paragraph yourself so it names the decisive criteria and the unanswered questions. UNDERSTAND A good recommendation is a reasoned link between criteria and evidence. It does not need to erase uncertainty. AI can organise a comparison, but you remain responsible for checking numbers, sources and whether your criteria actually matter. These factsheets are invented for practice. For a real purchase or booking, consult current primary sources and confirm total costs and availability. A completed training memo is not permission to act. CHECK YOUR RESULT The two-hour room-hire costs are A 160, B 130 and C 140 euros. Cedar is identified as too small for 18 people. Birch’s step-free access remains unknown, not assumed present or absent. The recommendation cites the decisive facts and a condition that could change the choice. Availability and extras are not invented, and no booking is made. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: The group is now 15 people, the room-hire budget is 150 euros and a whiteboard is optional. Reassess all three options without changing their factsheets. Explain why Cedar may now be the best-supported provisional option and why Birch still needs an access check. REUSABLE TEMPLATE WORKFLOW: DEFENSIBLE RECOMMENDATION Decision: __________ Hard constraints: __________ Preferences: __________ Options and dated primary evidence: __________ Method: calculate comparable totals; mark each constraint met / not met / unknown; compare; recommend conditionally. Review: reproduce calculations; verify source freshness; distinguish absent evidence from evidence of absence. Output: comparison + recommendation + what would change my mind + checks before acting. My final decision remains a human decision. WORKED EXAMPLE — TRY FIRST Alder: 2 × 80 = 160 euros; 24 seats meets capacity; step-free access confirmed; whiteboard included (A). Birch: 2 × 65 = 130 euros; 20 seats meets capacity; access unknown; whiteboard included (B). Cedar: 2 × 70 = 140 euros; 16 seats fails the 18-person requirement; access confirmed; no whiteboard (C). Recommendation: Alder is the best-supported provisional choice because it meets the stated room-hire budget, capacity and access requirements and includes a whiteboard. Before booking, confirm availability and any additional charges. Birch costs 30 euros less for room hire and could become preferable if step-free access and total costs are confirmed. Cedar does not fit 18 people on the supplied facts. Why it works: an unknown is not turned into a fact, and the recommendation states what could change it. REMIX Apply the comparison method to a low-stakes personal choice using current sources you verify. Start with criteria rather than asking the AI for “the best” option. ============================================================ 5. Turn messy notes into a clear presentation — approximately 15 minutes LEARNING GOAL Build a concise, audience-specific storyline that distinguishes observed evidence, interpretation and a proposed next decision. IMAGINE You have a small trial result and a meeting with a manager. You’ll use AI to create a five-slide storyline, then make one real slide yourself. Your result: a five-slide outline and a drafted slide that asks for a decision without overselling the evidence. SAMPLE INPUT FICTIONAL TRIAL NOTES — HARBOR WORKSPACE P1. A welcome-sheet trial ran at the front desk for two weeks. P2. Six participating new members tried finding the booking guide using the sheet; five found it without asking staff, one asked for help. P3. The six participants were a convenience sample; there was no baseline or comparison group. P4. Two participants said the link text was too small. P5. Printing cost 18 euros. Staff time was not measured. P6. Proposal: make the link text larger and run another small trial. Not approved yet. P7. Audience: the manager deciding whether to approve a second trial. Do not ask for permanent rollout. TRY Minute 0–3: write the decision you want from the manager. Circle what was measured and what was not. Minute 3–7: submit the prompt and notes. Look at the headlines without reading the bullets: do they form a sensible story? Minute 7–11: audit the claims. The trial did not establish a causal improvement or staff-time saving. Request one revision where needed. Minute 11–15: copy the outline into your presentation tool or a document. Build one slide with a takeaway headline, no more than three bullets and a speaker note that preserves the caveat. A full designed deck is an optional extension. STARTER PROMPT Create a five-slide outline for the audience and decision in the notes. Each slide needs a clear takeaway headline, no more than three short bullets and one sentence of speaker notes. Label source IDs for factual claims in the notes. Cover: decision context, what we tried, what we observed, limitations and adjustment, decision requested. Preserve the small sample and lack of baseline. Do not claim causal improvement, time saved, universal success or approval. End with a specific request for a second trial, not permanent rollout. SOURCE NOTES: [Paste the sample input here.] CREATE If the headline says “Our new sheet cuts support time,” replace it with an observation you can support. A safer version is “Five of six participants found the guide without asking staff.” Ask: “Shorten the headlines while keeping their caveats accurate.” Choose the version you can say aloud naturally. Create at least one slide or document section; do not stop with an unsaved chat answer. UNDERSTAND Presentations often distort evidence through compression. An accurate paragraph can become an overstated headline. Review the headlines, charts and speaker notes separately. This sample tells you what happened in a small trial, not what would happen for all members. Matching the story to the decision matters more than decorative polish. CHECK YOUR RESULT The outline has five slides with one main point each and at most three bullets. Five of six is reported as a small-sample observation, not a causal improvement. The lack of baseline and the unmeasured staff time stay visible. The request is a second trial, not a claimed approval or permanent rollout. At least one slide or document section has been created and read aloud. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: Reframe the same evidence for front-desk volunteers who will help run a future trial. Draft three headlines and one instruction they would need confirmed. Do not turn the proposal into an approved start date. Explain why this audience needs a different story from the manager. REUSABLE TEMPLATE WORKFLOW: EVIDENCE TO STORYLINE Audience: __________ Decision or understanding needed: __________ Verified source facts and IDs: __________ Story: context → activity → observations → limits → requested next step. For each slide: one defensible headline, up to three bullets, one speaker note. Review: check headline claims, sample limits, chart arithmetic and unapproved commitments. Done when: a reader can identify both the requested decision and what is still uncertain. WORKED EXAMPLE — TRY FIRST 1. Decide whether to refine the welcome sheet • Small front-desk trial completed (P1). • Today’s decision: a second trial, not rollout (P6–P7). Speaker note: We are asking to learn more, not declaring success. 2. We tested the sheet with six new members • Two-week trial (P1). • Convenience sample; no baseline (P2–P3). Speaker note: The sample is useful for exploration, not a population estimate. 3. Five participants found the guide unaided • Five of six found it without staff help; one needed help (P2). • Two reported small link text (P4). Speaker note: These are observations, not proof the sheet caused an improvement. 4. Make the link easier to see, then test again • Larger text is proposed (P6). • Printing cost 18 euros; staff time unmeasured (P5). Speaker note: We cannot claim cost or time savings. 5. Approve a second small trial? • Confirm the revised design and evaluation question. • Agree an owner and start date before proceeding. Speaker note: These next steps are requests, not previously agreed commitments. Why it works: every claim keeps its scale, and the conclusion matches the decision. REMIX Try the same material as a 100-word update. What can you cut without losing the caveat? If the caveat disappears, the shorter version is not yet ready to share. ============================================================ 6. Give one repetitive task a head start — approximately 15 minutes LEARNING GOAL Create a reusable workflow with input requirements, review checks and a human handoff; test it on a second case and a missing-information case. IMAGINE The final lab turns a useful one-off prompt into a small system you can use again. You’ll build a weekly-update workflow, test it twice and write down when it should stop and ask for help. Your result: a reusable workflow card and two reviewed sample outputs. It remains human-assisted, not an unattended automation. SAMPLE INPUT FICTIONAL WEEKLY UPDATE INPUT A Done: welcome-sheet draft written. Next: ask the manager to review it. Blocker: feedback owner not agreed. Dates: none confirmed. Audience: teammates. FICTIONAL WEEKLY UPDATE INPUT B Done: the manager reviewed the draft and requested larger link text. Next: Jo agreed to revise the text by Friday; calendar date not supplied. Blocker: start date still not agreed. Audience: teammates. EDGE CASE INPUT C “same as last week, nearly ready” — no other detail provided. TRY Minute 0–3: name the recurring task and the fields it needs. Read inputs A, B and C. Predict why C is too vague for a factual status report. Minute 3–7: use the prompt with input A only. Review the draft against the input. Minute 7–11: use the same prompt in a fresh chat with input B. Check that it does not carry over outdated facts from A. Minute 11–15: test input C and save a workflow card that includes a stop rule, a review checklist and the human action required before sending. STARTER PROMPT Create a weekly update from this run’s input only. Do not use facts from earlier conversations. Use headings Done, Next and Needs attention, under 120 words total. Keep completion, proposals and commitments distinct. Do not invent dates, approvals, progress or people. If the input is too vague to support an update, ask up to three clarification questions instead of drafting. Preserve relative dates without guessing their calendar date. End with a separate “Review before sending” list of unresolved details. Do not send messages or change any external system. THIS RUN’S INPUT: [Paste input A, B or C, one at a time.] CREATE Save your improved instruction together with its required input fields, rather than saving the prompt alone. Add a stop rule: do not produce a factual report when evidence is missing. Compare A and B. A says a review is next; B says it happened. A reusable workflow must respond to the actual input, not a remembered template story. Keep both reviewed outputs as test cases for future edits. UNDERSTAND A reusable workflow contains more than wording: it defines inputs, output, checks, exceptions and responsibility. Changing the model or prompt can change behaviour, so repeat the test cases after a change. Generating a draft is different from sending it. You remain responsible for fact-checking and approving the communication. Do not connect automatic sending simply because two examples worked. CHECK YOUR RESULT The saved card defines the task, required inputs, output format and audience. Inputs A and B produce distinct updates that preserve completion status. Input C triggers useful clarification rather than invented facts. The workflow includes a stop rule and specific fact/date/owner checks. A human reviews and sends the output; no unattended action is implied. INDEPENDENT PRACTICE WITHOUT THE STARTER PROMPT: Design a reusable handover workflow for a colleague taking over a project. Define required inputs, the output headings, two review checks and one stop condition. Test it using a fictional case and then a case missing an owner. Save the two test outputs with the workflow. REUSABLE TEMPLATE MY REUSABLE WORKFLOW CARD Task and intended audience: __________ Use it when: __________ Do not use it when: __________ Required inputs and allowed sources: __________ Steps / prompt: __________ Expected output and length: __________ Facts, calculations, dates and permissions I must check: __________ Stop and ask when: __________ Human approval / next action: __________ Test case 1 and expected behaviour: __________ Test case 2 / missing-input case: __________ Last checked date and tool used: __________ WORKED EXAMPLE — TRY FIRST Input A: Done: The welcome-sheet draft is written. Next: Ask the manager to review it. Needs attention: A feedback owner and dates are not yet agreed. Review before sending: Confirm whether a review request has already been made; don’t say it has without evidence. Input B: Done: The manager reviewed the draft and requested larger link text. Next: Jo will revise the text by Friday; the calendar date needs confirmation. Needs attention: The trial start date is not agreed. Review before sending: Confirm which Friday before relying on the deadline. Input C: A useful response is clarification, not a confident update. For example: “What changed this week? What is actually complete? What is the next agreed action?” Why it works: the same workflow changes with the evidence and refuses to manufacture missing progress. REMIX Choose a real low-stakes task that recurs this week, using only data you are allowed to share with your chosen tool. Try the workflow independently. Record what you corrected and whether you would use it again. That independent reuse is more meaningful than collecting a completion badge. FINAL INDEPENDENT EXERCISE https://sposable.com/knowledge-work/capstone/ Keep your outputs and a record of corrections. A completion mark records your activity, not independent certification.