<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Workflow Notes on Medium]]></title>
        <description><![CDATA[Stories by Workflow Notes on Medium]]></description>
        <link>https://medium.com/@workflow-notes?source=rss-ae857f7c0838------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*DIBCjPIJcEKWJNAAwExtJw.jpeg</url>
            <title>Stories by Workflow Notes on Medium</title>
            <link>https://medium.com/@workflow-notes?source=rss-ae857f7c0838------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 04:21:29 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@workflow-notes/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Turn a Policy or Handbook Section into a What This Means for Me Cheatsheet]]></title>
            <link>https://medium.com/@workflow-notes/turn-a-policy-or-handbook-section-into-a-what-this-means-for-me-cheatsheet-fa6b173d2515?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/fa6b173d2515</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[documentation]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Wed, 18 Nov 2026 01:25:20 GMT</pubDate>
            <atom:updated>2026-11-18T01:25:20.476Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste a handbook section → a personal cheatsheet: must-do, must-not, exceptions, and who to ask — no invented rules.</h4><p>Wednesday 2:10 p.m. A teammate pastes a short handbook section into the chat:</p><blockquote><em>“I’m a project lead. What does this mean for me — must-do, must-not, exceptions, and who to ask. Don’t invent a penalty or an owner if the section doesn’t name one.”</em></blockquote><p>That is not a five-bullet FAQ digest of a long document, and not a handoff checklist for a process you already know by heart. Ask a chatbot to “explain the policy,” and you often get a fluent card: a penalty nobody wrote, an “ask HR” owner nobody named, and an extra exception that was never in the excerpt.</p><p>What you need is a <strong>personal “what this means for me” cheatsheet</strong>: must-do, must-not, exceptions, and who to ask — with blanks left as MISSING or UNASSIGNED when the excerpt is silent. Below is a twelve-minute workflow. Paste one policy or handbook section. Leave unnamed who-to-ask as UNASSIGNED, unstated dates as NO DATE, and unstated penalties, forms, or rules as MISSING. Free-tier chatbot is enough. The model structures; you refuse invented rules, exceptions, penalties, and owners.</p><h3>The scenario we’ll use</h3><p>An illustrative handbook excerpt, incomplete on purpose — like most “what does this mean for me” pastes mid-afternoon.</p><blockquote><strong><em>Handbook section (illustrative): Gifts and hospitality</em></strong></blockquote><blockquote><em>Staff may not accept a gift or hospitality from a vendor, customer, or partner with a value above $80, except a branded item of nominal value (such as a pen or notebook) or a meal during a scheduled work meeting where the business purpose is documented.</em></blockquote><blockquote><em>If you are offered a gift above $80, you must decline it, or if declining is impractical, you must report it within 5 business days.</em></blockquote><blockquote><em>Do not solicit gifts.</em></blockquote><blockquote><em>This section does not apply to gifts from coworkers for birthdays or farewells.</em></blockquote><blockquote><strong><em>What the paste does NOT contain:</em></strong><em> who receives the report, or which form; a named person or role to ask (</em><em>UNASSIGNED); a penalty; a definition of &quot;impractical&quot;; a dollar cap for &quot;nominal value&quot; beyond the pen and notebook examples; a last-updated or effective date (</em><em>NO DATE); whether $80 is before or after tax (</em><em>MISSING).</em></blockquote><h3>Step 1: Pull must-do, must-not, and exceptions before you paste — don’t “explain the policy”</h3><p>Do not ask the model to “summarize this policy,” “make a five-question FAQ,” or “write the steps to file a report.” Without a gap pass, “report it within 5 business days” becomes “email HR on the gift form,” “nominal value” becomes “under $20,” and a penalty appears from nowhere.</p><p>Item in the excerptStill openMust not accept a gift or hospitality from a vendor, customer, or partner above $80Whether $80 is before or after tax: MISSINGException: branded item of nominal value (examples: pen, notebook)A dollar cap for “nominal” beyond those examples: MISSINGException: a meal during a scheduled work meeting, business purpose documented”Scheduled” / “documented”: only what the sentence saysIf offered a gift above $80: decline, or if declining is impractical, report within 5 business days”Impractical”: MISSING. Who receives the report / which form: MISSINGDo not solicit gifts — This section does not apply to gifts from coworkers for birthdays or farewellsOther coworker occasions: not stated — do not extend the exceptionWho to ask / penalty / effective dateUNASSIGNED / MISSING / NO DATE</p><p>If a rule, exception, penalty, form, or owner is not in the paste, do not write it as fact. <strong>`MISSING` / `UNASSIGNED` / `NO DATE` stay visible.</strong></p><h3>Step 2: Run the excerpt → personal cheatsheet prompt</h3><p><strong>Prompt — Policy or handbook section → “what this means for me” cheatsheet</strong></p><pre>You are turning a pasted company policy or handbook<br>section into a personal &quot;what this means for me&quot;<br>cheatsheet.</pre><pre>Hard rules:<br>- Use only facts in the pasted excerpt. Do not invent<br>  policy rules, exceptions, penalties, forms, dollar<br>  caps, or an escalation owner.<br>- If who-to-ask is not stated, write UNASSIGNED. Do<br>  not invent a role such as HR, Legal, or &quot;your<br>  manager.&quot;<br>- Unstated dates: NO DATE. Unstated penalties, forms,<br>  definitions, or caps: MISSING.<br>- Do not extend an exception beyond the cases the<br>  excerpt names.<br>- This is a personal cheatsheet, not a five-bullet FAQ<br>  digest of a long document, and not a how-to checklist<br>  for handing off a process. Do not output five Q&amp;As<br>  with source hints, or a numbered procedure someone<br>  else can follow.</pre><pre>Return markdown:<br>## Must do<br>## Must not<br>## Exceptions (only as stated)<br>## Who to ask (UNASSIGNED if unstated)<br>## Still MISSING / UNASSIGNED / NO DATE<br>## What not to write</pre><pre>Pasted policy or handbook section:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a short handbook section needs a personal readout.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong> for the sample above.</p><p><strong>Must do:</strong> If offered a gift above $80 from a vendor, customer, or partner, decline it. If declining is impractical, report it within 5 business days. For a meal under the scheduled-meeting exception, document the business purpose.</p><p><strong>Must not:</strong> Do not accept a gift or hospitality from a vendor, customer, or partner above $80, unless an exception in the excerpt applies. Do not solicit gifts.</p><p><strong>Exceptions (only as stated):</strong> A branded item of nominal value (examples: a pen or notebook). A meal during a scheduled work meeting where the business purpose is documented. This section does not apply to gifts from coworkers for birthdays or farewells. Do not extend “nominal value” to a dollar cap, or “coworker gifts” to other occasions.</p><p><strong>Who to ask:</strong> UNASSIGNED — the excerpt names no person or role.</p><p><strong>Still MISSING / UNASSIGNED / NO DATE:</strong> who receives the report, and which form; what “impractical” means; a dollar cap for “nominal value” beyond pen and notebook; whether $80 is before or after tax; a penalty; effective date (NO DATE); who to ask (UNASSIGNED).</p><p><strong>What not to write:</strong> “Ask HR or your manager.” “Violations may lead to disciplinary action.” “Nominal means under $20.” “Report it on the gift declaration form.” “Coworker wedding gifts are also fine.” A five-bullet FAQ. A numbered “how to file a gift report” checklist.</p><h3>Step 3: Human pass — five checks before you use the cheatsheet</h3><p>Do not ask the model to “make this more complete” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every rule, exception, penalty, form, date, and owner is in the paste or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>Who-to-ask stays `UNASSIGNED` when the section names no one</strong> — do not invent HR, Legal, or “your manager.”</li><li><strong>Exceptions are not extended</strong> — “pen or notebook” is not a $20 cap. “Birthdays or farewells” is not every coworker occasion.</li><li><strong>No invented penalty or process</strong> — a reporting window is not a form, a workflow, or a disciplinary line.</li><li><strong>This is a personal cheatsheet, not a five-bullet FAQ and not a handoff checklist</strong> — if the output is five Q&amp;As or a numbered procedure for someone else, re-run.</li></ol><p>If a line only works by inventing a rule, an owner, or a penalty, delete it.</p><h3>Step 4: How to use the cheatsheet when a real case shows up</h3><p>Keep the cheatsheet next to the excerpt, with the gap list visible. If you need to report a gift, the excerpt says “report it within 5 business days” — it does not say to whom. Write MISSING; do not guess a mailbox. Until someone names an owner or a form, who-to-ask stays UNASSIGNED. The excerpt does not give project-lead-specific extras; do not invent a stricter or looser rule for leads.</p><p>A cheatsheet that already “has an owner,” “has a penalty,” and “defines nominal as $20” is unauthorized policy.</p><h3>How this differs from “just ask AI to explain the policy” — and from a FAQ or a process checklist</h3><p>A soft “explain this policy” returns a fluent card with a penalty, an owner, and a tidier exception the excerpt never wrote. Two nearby workflows are the wrong artifact:</p><ul><li><strong>A long doc into a five-bullet FAQ</strong> starts from a long policy, SOP, or manual and returns five Q&amp;As with source hints, plus “who else to ask.” This page is not a team FAQ digest. It is a personal must-do / must-not / exceptions card from one excerpt.</li><li><strong>A process you know by heart into a checklist someone else can follow</strong> starts from a spoken dump of a handoff or setup and returns numbered steps for another person. This page is not a how-to you are handing off. It is what one policy section means for you.</li></ul><p>A useful cheatsheet is not a shorter handbook. It is an honest personal readout, with a gap list you can defend.</p><h3>Common failure modes the prompt should catch</h3><p>“Ask HR or your manager”; “violations may lead to disciplinary action”; “nominal means under $20”; a gift-declaration form the excerpt never named; extending coworker gifts to weddings or holidays; five FAQ bullets with source hints; a numbered “how to file the report” checklist.</p><p>If any appear, delete the line. Re-run with the gap table pinned at the top.</p><h3>Five mistakes that fake a “ready” cheatsheet</h3><p><strong>1. Inventing who to ask so the card looks usable.</strong> If the section names no person or role, the line is UNASSIGNED. A guessed &quot;ask HR&quot; will be treated as policy.</p><p><strong>2. Inventing a penalty, a form, or a filing path.</strong> “Report it within 5 business days” is not a form name, a mailbox, or a disciplinary outcome.</p><p><strong>3. Tightening a vague word into a rule.</strong> “Nominal value” is illustrated with a pen or notebook, not a $20 cap unless the excerpt says so. “Impractical” stays MISSING as a definition.</p><p><strong>4. Extending an exception.</strong> Birthday and farewell gifts from coworkers are carved out. Other coworker occasions are not.</p><p><strong>5. Handing back a five-bullet FAQ or a process checklist.</strong> Those are the right tools for a long-doc digest or a handoff, not for a personal “what this means for me” card.</p><h3>Make twelve minutes sacred before you treat the card as policy</h3><p>Keep the prompt in snippets. When a handbook section lands, run gap table → prompt → human pass before you accept a gift, file a report, or tell a teammate “the rule is.” The model is a structure engine, not a witness to an owner, a penalty, or an exception the excerpt never had. When someone says “just explain the policy,” you already have must-do, must-not, exceptions as stated, and a visible gap list — MISSING still showing.</p><p>— -</p><p>*Next handbook paste: will you use a fluent card that invents an owner and a penalty — or a cheatsheet whose blanks are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa6b173d2515" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn Two Vendor Quotes into a One-Page Buy Recommendation]]></title>
            <link>https://medium.com/@workflow-notes/turn-two-vendor-quotes-into-a-one-page-buy-recommendation-0010c2862732?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/0010c2862732</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[procurement]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Wed, 18 Nov 2026 01:20:58 GMT</pubDate>
            <atom:updated>2026-11-18T01:23:34.461Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste two vendor quotes (price, terms, scope) → a one-page buy rec: comparison, risks, which to recommend and why.</h4><p>Wednesday 11:05 a.m. A teammate drops two short quotes into the chat:</p><blockquote><em>“Can you turn these into a one-page buy recommendation before the 1 p.m. review? Comparison, risks, which to recommend and why. Don’t invent a discount, an SLA, or a delivery date if the quotes don’t have them. We need one-day facilitation. Headcount is 18. Venue is already booked by us.”</em></blockquote><p>That is not a competitor feature table, and not a request to pick a “better” vendor from reputation. Ask a chatbot to “recommend one,” and you often get a memo: a discount nobody offered, a “market rate,” a delivery date, and a quality claim the quotes never made.</p><p>What you need is a <strong>one-page procurement recommendation</strong>: comparison, risks from gaps only, and a recommendation grounded in the paste, with unimplied preferences labeled as hypotheses. Below is a twelve-minute workflow. Paste the two quotes. Leave unnamed owners as UNASSIGNED, unstated dates as NO DATE, and a one-sided term as MISSING. Free-tier chatbot is enough. The model structures; you refuse invented prices, discounts, totals, SLAs, dates, and reputation.</p><h3>The scenario we’ll use</h3><p>Two illustrative vendor quotes, short on purpose — like most comparison pastes an hour before a review.</p><blockquote><strong><em>Requester note (illustrative):</em></strong><em> One-day offsite facilitation. Headcount is 18. Venue already booked.</em></blockquote><blockquote><strong><em>Quote A — Northline Studio (illustrative):</em></strong><em> One-day facilitation, up to 20 people. Fee $2,400; materials $180; travel $220; total $2,800. Venue not included. Payment: 50% on booking, 50% within 14 days after the session. Cancellation: full refund if cancelled 10 or more days before; no refund inside 10 days. Session date not stated. Follow-up note not stated.</em></blockquote><blockquote><strong><em>Quote B — Harbor &amp; Co. (illustrative):</em></strong><em> One-day facilitation, up to 16 people. Fee $2,100; total $2,100. Printed workbooks included. Travel included if within the metro area; out-of-metro not quoted. Venue not included. Payment: 100% on booking. Cancellation: 50% refund if cancelled 7 or more days before; no refund inside 7 days. Follow-up summary included within 5 business days. Session date not stated.</em></blockquote><blockquote><strong><em>What the paste does NOT contain:</em></strong><em> a session date (</em><em>NO DATE on both); a discount; a live-session SLA; out-of-metro travel on B (</em><em>MISSING); a follow-up on A (</em><em>MISSING); reputation or quality scores; a named owner; a &quot;market rate.&quot;</em></blockquote><h3>Step 1: Line up the two quotes — don’t “pick a winner” yet</h3><p>Do not ask the model to “recommend the better vendor.” Without a gap pass, B’s lower total becomes “better value,” A’s capacity disappears, and a delivery date appears from nowhere.</p><p>ItemQuote AQuote BStill openScopeUp to 20Up to 16Needed: 18. B does not cover 18 as quotedTotal$2,800 ($2,400 + $180 + $220)$2,100 (workbooks + metro travel included)No discount. B out-of-metro travel: MISSING. Venue in metro?: MISSINGVenueNot includedNot includedAlready bookedPayment50% ($1,400) on booking; 50% ($1,400) within 14 days after100% on booking ($2,100) — CancellationFull refund at 10+ days; none inside 1050% refund at 7+ days; none inside 7Session date: NO DATE — window cannot be dated yetFollow-up / live SLAFollow-up MISSING; SLA MISSINGSummary within 5 business days; SLA MISSING — Decision ownerUNASSIGNEDUNASSIGNED —</p><p>If a price, discount, total, SLA, delivery date, or reputation claim is not in the paste, do not write it as fact. <strong>A one-sided term is `MISSING` on the gap side. `UNASSIGNED` / `NO DATE` stay visible.</strong></p><h3>Step 2: Run the two quotes → buy recommendation prompt</h3><p><strong>Prompt — Two vendor quotes → one-page buy recommendation</strong></p><pre>You are turning two pasted vendor quotes (price, terms, scope)<br>and any requester note into a one-page procurement recommendation.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent prices,<br>  discounts, totals, SLAs, delivery dates, vendor<br>  reputation, better-quality, or market-rate claims.<br>- If a term exists in one quote but not the other, mark<br>  it MISSING on the gap side. Do not fill the gap.<br>- Unnamed owners: UNASSIGNED. Missing session or<br>  delivery dates: NO DATE.<br>- Arithmetic: restate a total only as the quote states<br>  it, or as a shown sum of line items in the same quote.<br>  If a quoted total does not match, flag it; do not<br>  &quot;fix&quot; it.<br>- Recommendation and risks must be grounded only in<br>  differences and gaps in the pasted quotes.<br>- Any preference not strictly implied by the quotes is<br>  a hypothesis, labeled as such.<br>- This is a buy decision between two vendor quotes, not<br>  a product-vs-competitor feature matrix. Do not output<br>  a Feature / Them / Us / Priority table.</pre><pre>Return markdown:<br>## Comparison (price, terms, scope)<br>## Gaps (MISSING / UNASSIGNED / NO DATE)<br>## Risks (from differences and gaps only)<br>## Recommendation (grounded in the paste; hypotheses labeled)<br>## What not to write</pre><pre>Pasted quotes and requester note:<br>[paste]</pre><p>Save it as a snippet. Same prompt each time two quotes arrive before a review.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong>:</p><p><strong>Comparison:</strong> A covers 18 as quoted (up to 20). B’s cap is 16, so B does not cover the pasted headcount. A total $2,800 = $2,400 + $180 + $220. B total $2,100, workbooks and metro travel included. Difference $700; no discount in either quote. Payment, cancellation, venue, and follow-up as in the gap table. Session date NO DATE on both.</p><p><strong>Gaps:</strong> A follow-up MISSING. B out-of-metro travel MISSING. Venue in metro MISSING. Live-session SLA MISSING on both. Decision owner UNASSIGNED. No discount or market-rate figure.</p><p><strong>Risks:</strong> B’s $2,100 is not a comparable option for 18 people. If the venue is out of metro, B’s travel is still MISSING; A&#39;s $220 is already in the total. B takes 100% on booking and offers only a 50% refund with a 7-day window, versus A&#39;s full refund at 10 days — term differences, not quality scores.</p><p><strong>Recommendation:</strong> Quote A is the only quote whose stated capacity covers 18. Recommend A as quoted, gaps still visible. Hypothesis, not a fact: if Harbor &amp; Co. later confirms in writing that they can facilitate 18, and states price and terms for that scope, re-run. Until then, do not treat B’s $2,100 as a valid option for 18, or assume the same price.</p><p><strong>What not to write:</strong> “Harbor is better value / higher quality / the market rate.” “Apply a 10% discount.” “The offsite is on 5 December.” “B will make an exception for 18 at $2,100.”</p><h3>Step 3: Human pass — five checks before you send the one-pager</h3><p>Do not ask the model to “make this more decisive” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every price, term, date, and owner is in the paste or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>Arithmetic</strong> — each restated total matches the quote, or a shown sum of that quote’s line items: $2,400 + $180 + $220 = $2,800; 50% of $2,800 = $1,400. Do not invent a discount or “fix” a mismatch.</li><li><strong>Gaps stay gaps</strong> — a term on one side only is MISSING on the other. No invented SLA or delivery date.</li><li><strong>The recommendation is grounded</strong> — any extra preference is a labeled hypothesis. No quality, reputation, or market-rate claims.</li><li><strong>This is a buy recommendation, not a competitor feature table</strong> — if the output is a Feature / Them / Us / Priority matrix, re-run.</li></ol><p>If a line only works by inventing a discount, a date, or a “better” vendor, delete it.</p><h3>Step 4: How to use the recommendation in the actual review</h3><p>Send the one-pager with the gap list visible. If someone asks “can Harbor take 18?” and the quote says 16, say the quote does not cover 18; do not write an exception. If a revised quote arrives, re-run. The owner stays UNASSIGNED unless the paste named one. The session date stays NO DATE until a quote or booking note states it.</p><p>A buy recommendation that already “has a discount,” “has a delivery date,” and “picks the higher-quality vendor” is an unauthorized vendor story.</p><h3>How this differs from “just ask AI to recommend one” — and from a feature comparison</h3><p>A soft “recommend one” returns a fluent memo with a value judgment, a quality claim, and a date nobody quoted. One nearby workflow is the wrong artifact:</p><ul><li><strong>Competitor page notes into a one-page feature comparison</strong> starts from competitor-page scraps and returns a Feature / Them / Us / Priority table. This page is a procurement decision between two vendor quotes — price, terms, scope, risks, and which quote to buy — not a feature matrix.</li></ul><p>A useful buy recommendation is a comparison, a risk list from the gaps, and a recommendation that only goes as far as the quotes.</p><h3>Common failure modes the prompt should catch</h3><p>“Harbor is better value”; “Northline is higher quality”; a market rate; a 10% discount; “the session is on 5 December”; an SLA nobody quoted; “B can take 18 at the same price”; a Feature / Them / Us / Priority table; reputation or star ratings.</p><p>If any appear, delete the line. Re-run with the gap table pinned at the top.</p><h3>Five mistakes that fake a “ready” buy recommendation</h3><p><strong>1. Treating the lower total as “better value” when scope does not match.</strong> A $2,100 quote for 16 people is not a comparable option for a pasted headcount of 18.</p><p><strong>2. Inventing a discount, a cleaned-up total, an SLA, or a delivery date.</strong> MISSING and NO DATE are usable. A made-up 5 December is a commitment.</p><p><strong>3. Filling the gap side instead of leaving it `MISSING`.</strong> A’s missing follow-up is not “no follow-up.” B’s missing out-of-metro travel is not “free everywhere.”</p><p><strong>4. Adding a quality, market-rate, or reputation claim the quotes never made.</strong> “More professional” and “the going rate” are not in the paste.</p><p><strong>5. Handing back a competitor feature table.</strong> Feature / Them / Us / Priority is for product-versus-competitor notes, not for two vendor quotes.</p><h3>Make twelve minutes sacred before you recommend a vendor</h3><p>Keep the prompt in snippets. When two quotes land before a review, run gap table → prompt → human pass (including the arithmetic) before you recommend. The model is a comparison engine, not a witness to a discount, an SLA, or a reputation the quotes never had. When someone says “just tell us which vendor to pick,” you already have a comparison, a risk list, and a recommendation that only goes as far as the quotes — MISSING still showing.</p><p>— -</p><p>*Next pair of quotes: will you send a fluent memo that invents a discount and a “better” vendor — or a one-page buy recommendation whose gaps are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0010c2862732" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn a Chaotic Brainstorm into a Prioritized Next-Step Backlog]]></title>
            <link>https://medium.com/@workflow-notes/turn-a-chaotic-brainstorm-into-a-prioritized-next-step-backlog-f408baa90552?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/f408baa90552</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 01:28:18 GMT</pubDate>
            <atom:updated>2026-11-17T01:28:18.439Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste stickies, messy workshop talk, or a voice note → grouped themes and a backlog tagged do / don’t / needs clarification.</h4><p>Tuesday 4:40 p.m. The huddle ends. A teammate dumps stickies plus a few voice-note lines:</p><blockquote><em>“Don’t leave this as a wall of stickies. Turn it into a backlog: group the themes, then next steps tagged do / don’t / needs clarification. Don’t invent an owner or a date, and don’t turn half-ideas into committed work.”</em></blockquote><p>That is not a customer-feedback theme memo, and not a personal today plan. Ask a chatbot to “organize our brainstorm” or “turn this into tasks” and you often get a fluent plan: an owner nobody named, a Friday deadline, effort points, and half-ideas promoted into committed work.</p><p>What you need is a <strong>prioritized next-step backlog</strong>: grouped themes, then a list tagged do / do not / needs clarification, with guesses labeled as hypotheses and blanks left as MISSING. Below is a twelve-minute workflow. Paste the stickies, messy talk, or short voice transcript. Leave unnamed owners as UNASSIGNED and unstated dates as NO DATE. Free-tier chatbot is enough. The model groups; you refuse invented owners, dates, effort, and any rewrite that turns an unclear idea into committed work.</p><h3>The scenario we’ll use</h3><p>An illustrative 20-minute huddle dump, messy on purpose — like most sticky walls after someone says “getting a design review takes too many pings.”</p><blockquote><strong><em>Stickies and talk (illustrative):</em></strong><em><br>- “too many pings for a design review”<br>- “maybe a short request form?”<br>- “what if we had design office hours”<br>- “Casey mentioned a shared board last month? or was that a different team? not sure”<br>- “don’t add a new tool if we already have a board”<br>- “a template for what to attach”<br>- “who would triage requests? nobody said”<br>- “I don’t want another weekly design meeting”<br>- “could try office hours with one squad first”<br>- “auto-assign a designer?? that feels like a different problem”</em></blockquote><blockquote><strong><em>What the paste does NOT contain:</em></strong><em> agreement that any item is committed work; an owner for triage or office hours; a calendar date (“last month” is not a date for this work); effort; a priority rank; a tool name beyond “a shared board”; a policy.</em></blockquote><h3>Step 1: Separate themes, refusals, and half-ideas before you paste</h3><p>Do not ask the model to “turn this into tasks” or “make us a plan for the week.” Without a gap pass, “could try office hours” becomes Casey’s Friday job, “maybe a form” becomes committed work, and “don’t add a new tool” disappears.</p><p>Item in the dumpStill openPain: too many pings for a design reviewCount and time window: MISSINGIdeas: request form; office hours; a template for attachmentsHedged or unfinished. No agreement. Effort: MISSING”Could try office hours with one squad first”A possible trial, not a decision. Owner: UNASSIGNED. Date: NO DATECasey mentioned a shared board last month — “not sure” if it was a different teamWhether a board already exists: MISSING. Casey is a source, not an ownerRefusals: don’t add a new tool if a board already exists; no new weekly design meetingDo-not items, not tasks to startUnclear: who triages; auto-assign a designer (“maybe the wrong problem”)Triage owner: UNASSIGNED. Priority and effort: MISSING</p><p>If an owner, date, effort, or rank is not in the paste, do not write it as fact. <strong>MISSING / UNASSIGNED / NO DATE stay visible. Proposed is not committed.</strong></p><h3>Step 2: Run the dump → next-step backlog prompt</h3><p><strong>Prompt — Ideation dump → next-step backlog</strong></p><pre>You are turning pasted sticky notes, messy workshop talk, or a<br>short voice transcript into grouped themes and a next-step backlog.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent owners, dates,<br>  effort, points, priority ranks, tool names, or policy.<br>- This is an ideation dump, not customer/user feedback and<br>  not a personal to-do list. Do not output five feedback<br>  themes with frequency guesses and evidence quotes, or a<br>  today plan with three must-dos and a tomorrow pile.<br>- Do not elevate unclear, hedged, or unfinished ideas into<br>  committed work. No agreement to start an item means it<br>  is not a commitment.<br>- DO: only next steps the paste clearly agrees to start. If<br>  none are agreed, leave DO empty or list only proposed<br>  trials labeled &quot;proposed, not committed.&quot;<br>- DO NOT: items the paste rejects or warns against.<br>- NEEDS CLARIFICATION: incomplete, contradicted, or<br>  ownerless ideas. Unnamed owners: UNASSIGNED. Missing<br>  dates: NO DATE. &quot;Last month&quot; is not a work date.<br>- Any priority order is a hypothesis unless the paste<br>  already ranked the items. Label it as a hypothesis.</pre><pre>Return markdown:<br>## Themes (grouped from the paste; not commitments)<br>## Do (agreed steps only; else proposed, not committed)<br>## Do not<br>## Needs clarification<br>## Priority (hypothesis unless already ranked)<br>## Still MISSING / UNASSIGNED / NO DATE<br>## What not to write</pre><pre>Pasted stickies, talk, or voice transcript:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a sticky wall or voice note becomes a backlog.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong> for the sample.</p><p><strong>Themes (grouped; not commitments):</strong> (1) request quality — form, attachment template; (2) access and routing — office hours, shared board, triage; (3) what not to add — no new tool if a board already exists, no new weekly design meeting; (4) maybe the wrong problem — auto-assign a designer.</p><p><strong>Do:</strong> No agreed next step in the paste. Proposed, not committed: try office hours with one squad first — still UNASSIGNED, date NO DATE, effort MISSING.</p><p><strong>Do not:</strong> Don’t add a new tool if a board already exists. Don’t add another weekly design meeting.</p><p><strong>Needs clarification:</strong> Who would triage (UNASSIGNED); whether a shared board already exists (Casey&#39;s mention is uncertain; Casey is not the owner); form vs template — which, if any; auto-assign a designer (maybe the wrong problem).</p><p><strong>Priority (hypothesis, not a ranking in the paste):</strong> Hypothesis — clarify whether a board already exists, then consider the one-squad office-hours trial, before any new form.</p><p><strong>Still MISSING / UNASSIGNED / NO DATE:</strong> agreement to start any item; triage and office-hours owners (UNASSIGNED); dates (NO DATE); effort; a named existing board; a priority rank; policy.</p><p><strong>What not to write:</strong> “Casey owns office hours and will start Friday.” “The form is a P0.” “Two-point effort.” “We decided to auto-assign designers.” “Five feedback themes from users.” “Today’s three must-dos.”</p><h3>Step 3: Human pass — five checks before you send the backlog</h3><p>Do not ask the model to “make this more decisive” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every owner, date, effort figure, and rank is in the paste or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>Unclear ideas are not committed work</strong> — “maybe,” “what if,” “could try,” and “not sure” stay proposed or under clarification.</li><li><strong>Do-not items stay do-not</strong> — a refusal is not rewritten as a task to start.</li><li><strong>Priority is a hypothesis unless the paste ranked the items.</strong></li><li><strong>This is not a feedback-theme memo or a today plan</strong> — if the output is five feedback themes with frequency guesses, or three must-dos plus tomorrow, it failed. Re-run.</li></ol><p>If a line only works by inventing an owner or treating a half-idea as decided, delete it.</p><h3>Step 4: How to use the backlog after the huddle</h3><p>Send the backlog to the group with the gap list visible. If someone later agrees to try office hours, you update the Do line yourself. If they ask “who owns triage?” and the paste named no one, say UNASSIGNED. Casey mentioned a board; that does not make Casey the owner.</p><p>A backlog that already “has owners,” “starts Friday,” and “commits the form” is not a backlog. It is unauthorized work.</p><h3>How this differs from “just ask AI to organize the brainstorm” — and from five themes or a today plan</h3><p>A soft “organize this brainstorm” returns a fluent plan in which every sticky is already a task, plus an owner and a date nobody gave. Two nearby workflows are the wrong artifact here:</p><ul><li><strong>Raw feedback into five themes</strong> starts from customer or user replies and returns themes, evidence quotes, frequency guesses, and suggested next steps. This dump is internal ideation, not feedback quotes.</li><li><strong>An overflowing to-do list into a today plan</strong> starts from a personal task pile and returns three must-dos, deferrals, and tomorrow. This page is not a personal day plan. It is a shared backlog tagged do / do not / needs clarification.</li></ul><p>A useful backlog is not a neater sticky wall, and not a set of decided tasks. It is grouped themes, honest tags, and a gap list that keeps half-ideas from becoming work.</p><h3>Common failure modes the prompt should catch</h3><p>“Casey owns office hours”; “the form starts Friday”; invented effort or a P0; “we decided to auto-assign designers”; “maybe” rewritten as agreed work; a refusal turned into a build task; five feedback themes with frequency guesses; a today plan with three must-dos and tomorrow.</p><p>If any appear, delete the line. Re-run with the gap table pinned at the top of the paste.</p><h3>Five mistakes that fake a “ready” backlog</h3><p><strong>1. Elevating a half-idea into committed work.</strong> “Maybe,” “what if,” “could try,” and “not sure” are not agreements.</p><p><strong>2. Inventing an owner, a date, effort, or a priority rank so the list looks official.</strong> UNASSIGNED, NO DATE, and MISSING are usable. A made-up Friday is a commitment.</p><p><strong>3. Treating a mention as an assignment.</strong> “Casey mentioned a shared board” names a source, not an owner.</p><p><strong>4. Rewriting a refusal as a task.</strong> “Don’t add a new tool” and “I don’t want another weekly meeting” belong in Do not, not in Do.</p><p><strong>5. Handing back a feedback-theme memo or a personal today plan.</strong> Those are the right tools for user replies or an overflowing personal list, not for this dump.</p><h3>Make twelve minutes sacred before stickies become “work”</h3><p>Keep the prompt in snippets. When a sticky wall or messy voice note lands after a huddle, run gap table → prompt → human pass before anyone starts a form or meeting. The model is a grouping engine, not a witness to owners, dates, or decisions the dump never had.</p><p>When someone says “just turn the stickies into tasks,” you already have grouped themes, a do / do not / needs-clarification list, and a visible gap list — MISSING still showing until a human fills it.</p><p>— -</p><p>*Next sticky wall: will you send a fluent plan that invents an owner and a Friday — or a backlog whose half-ideas are still waiting on clarification?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f408baa90552" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn Error Messages or Screenshots into a Bug Report Someone Can Act On]]></title>
            <link>https://medium.com/@workflow-notes/turn-error-messages-or-screenshots-into-a-bug-report-someone-can-act-on-6f6c7b7d0932?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/6f6c7b7d0932</guid>
            <category><![CDATA[work]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 01:24:40 GMT</pubDate>
            <atom:updated>2026-11-17T01:26:24.737Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste the error text or what you saw on screen → a bug report: repro steps, expected vs actual, severity, and what’s still missing.</h4><p>Tuesday 11:20 a.m. A teammate drops a screenshot into the chat:</p><blockquote><em>“Can you file this? The export failed. I’ll send the screenshot. Write a bug report engineering can act on — repro steps, expected vs actual, severity, and what we’re still missing. Don’t invent a browser or a severity if I didn’t give them.”</em></blockquote><p>That is not a request for a diagnosis, and it is not a request for the model to “see” the image. If you only attach the screenshot and ask a chatbot to “write the ticket,” you often get Chrome 120, an 11:04 a.m. timestamp, a user ID, four repro steps the paste never contained, and a P1 nobody assigned.</p><p>What you need is a <strong>handoff-ready bug report</strong>: visible text and a short description of what was on screen, repro steps only as the paste supports them, expected vs actual without invented rules, severity as a labeled hypothesis or MISSING, and a gap list. Below is a twelve-minute workflow. Type out the error text, the error code, and a short description of the screen (or alt text). Leave unnamed owners as UNASSIGNED, unstated times as NO DATE, and unstated browsers, versions, and IDs as MISSING. Free-tier chatbot is enough. The model structures; you refuse invented steps, browsers, versions, timestamps, user IDs, and severity.</p><h3>The scenario we’ll use</h3><p>An illustrative failed export, incomplete on purpose — like most screenshot handoffs that arrive with a red banner and almost no environment detail.</p><blockquote><strong><em>What was on screen (illustrative):</em></strong><em><br>- I was on the Time Log page and clicked Export.<br>- A red banner appeared: “Export failed. Error code TL-4402. Please try again later.”<br>- The table still showed this week’s rows. No download started.<br>- I clicked Export once more and got the same banner.</em></blockquote><blockquote><strong><em>What the paste does NOT contain:</em></strong><em> browser, OS, or app version; a clock time or calendar date (NO DATE); a user ID or account name; how I reached the Time Log page; file format or file name; whether anyone else is affected; expected behavior stated as a sentence; a severity label.</em></blockquote><h3>Step 1: Type the visible text before you paste — don’t assume the image can be seen</h3><p>Do not ask the model to “write a ticket from this screenshot” and stop there. In this workflow, the model does not see an attached image unless you paste the words. Without a gap pass, “Export failed” becomes a Chrome repro, an 11:04 timestamp, and a P1.</p><p>Item in the pasteStill openAction: Time Log page, clicked ExportHow the page was reached: MISSINGRed banner text and code TL-4402 — Table still showed this week’s rows; no download; retry once, same bannerExpected behavior stated: MISSING. File format / name: MISSINGEnvironmentBrowser, OS, app version: MISSING. Timestamp: NO DATE. User ID: MISSING. Who else is affected: MISSINGSeverityNot labeled: MISSING (any severity is a hypothesis, not a fact)Owner to receive the reportUNASSIGNED</p><p>If a browser, version, timestamp, user ID, repro step, or severity is not in the paste, do not write it as fact. <strong>MISSING / UNASSIGNED / NO DATE stay visible.</strong></p><h3>Step 2: Run the error text → bug report prompt</h3><p><strong>Prompt — Error message or on-screen description → handoff-ready bug report</strong></p><pre>You are turning pasted error text, an error code, and a short<br>description of what was on screen into a bug report someone<br>can hand to engineering or support.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent repro steps,<br>  browsers, OS, app versions, timestamps, user IDs, account<br>  names, file sizes, or severity.<br>- This workflow cannot see an attached image. If a detail<br>  was only in a screenshot and was not typed out, write<br>  MISSING. Do not pretend you looked at the image.<br>- Repro steps only as far as the paste supports them. Do<br>  not add &quot;open Chrome,&quot; &quot;log in,&quot; or a URL the paste<br>  does not contain.<br>- If expected behavior is not stated, write MISSING. Do<br>  not invent a product rule or size limit.<br>- Unnamed owners: UNASSIGNED. Missing dates or times:<br>  NO DATE. Hedged banner text (&quot;may be,&quot; &quot;try again<br>  later&quot;) is not a confirmed root cause.<br>- Severity: if the paste has no label, write MISSING, or<br>  one short hypothesis clearly labeled as a hypothesis,<br>  not as an assigned priority.</pre><pre>Return markdown:<br>## Summary (from the paste only)<br>## Repro steps (only steps in the paste)<br>## Expected vs actual (expected is MISSING if unstated)<br>## Severity (hypothesis labeled as such, or MISSING)<br>## Environment<br>## Still MISSING / UNASSIGNED / NO DATE<br>## What not to write</pre><pre>Pasted error text and on-screen description:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a screenshot or red banner arrives with almost no environment detail.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong> for the sample above.</p><p><strong>Summary:</strong> On the Time Log page, clicking Export showed a red banner: “Export failed. Error code TL-4402. Please try again later.” The table still showed this week’s rows. No download started. A second click produced the same banner.</p><p><strong>Repro steps (only steps in the paste):</strong> (1) On the Time Log page, click Export. (2) See the red banner with error code TL-4402. (3) Click Export once more; same banner. Do not add a browser, a login step, or a URL.</p><p><strong>Expected vs actual:</strong> Actual: export fails with TL-4402; no download; table still shows this week’s rows; retry once, same result. Expected: MISSING — no required outcome or file format is stated.</p><p><strong>Severity:</strong> MISSING as an assigned label. Hypothesis only (not a priority): this blocks this person&#39;s export on this page; the paste does not say whether anyone else is affected.</p><p><strong>Environment:</strong> Browser MISSING. OS MISSING. App version MISSING. Timestamp NO DATE. User ID MISSING.</p><p><strong>Still MISSING / UNASSIGNED / NO DATE:</strong> how the Time Log page was reached; expected behavior; file format and name; browser / OS / version; timestamp; user ID; whether others are affected; assigned severity; receiving owner (UNASSIGNED).</p><p><strong>What not to write:</strong> “Repro: open Chrome, log in, go to /time-log.” “Occurred at 11:04 a.m.” “User 88421.” “P1 — export is down for the team.” “Root cause is a timeout.”</p><h3>Step 3: Human pass — five checks before you hand it over</h3><p>Do not ask the model to “make this more complete” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every repro step, browser, version, timestamp, user ID, and severity is in the paste or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>The image was not treated as read</strong> — if a banner was only in a screenshot and not typed out, it should not appear as if the model saw it. Type it in, or leave it MISSING.</li><li><strong>No invented expected behavior or root cause</strong> — “try again later” is banner text, not a diagnosis.</li><li><strong>Severity stays a labeled hypothesis or `MISSING`</strong> — not an assigned P0/P1.</li><li><strong>This is a handoff report, not a fix plan</strong> — if the output is a patch list, a blame line, or a “ready” ticket with a filled environment, it failed. Re-run.</li></ol><p>If a line only works by inventing a browser, a step, or a severity, delete it.</p><h3>Step 4: How to use the report in the actual handoff</h3><p>Send the report to engineering or support with the gap list visible. If they ask “what browser?” and the paste has none, say MISSING; do not guess Chrome. If later you learn a file format or a timestamp, you add it. Until then, the receiving owner stays UNASSIGNED unless the paste named one.</p><p>A bug report that already “has the browser,” “has the user ID,” and “is a P1” is not a bug report. It is an unauthorized incident.</p><h3>How this differs from “just ask AI to write the ticket”</h3><p>A soft “write a ticket from this screenshot” returns a fluent form with an environment, a severity, and a four-step repro the paste never contained. This workflow is different on both ends: the input is <strong>typed error text, an error code, and a short description of what was on screen</strong> — not an image the model is assumed to see — and the output is a <strong>handoff-ready bug report</strong> whose blanks stay visible. It is not a status report, a decision log, or a meeting follow-up.</p><p>A useful bug report is not a complete ticket. It is an honest account of what appeared, the steps actually taken, and a gap list you can defend.</p><h3>Common failure modes the prompt should catch</h3><p>“Open Chrome and log in”; a timestamp or user ID that was never typed; “P1” written as assigned; expected behavior invented from a product rule; “root cause is a timeout”; a patch or a blamed team; a walkthrough of a screenshot the model was not given as text.</p><p>If any of those appear, delete the line. Re-run with the gap table pinned at the top of the paste.</p><h3>Five mistakes that fake a “ready” bug report</h3><p><strong>1. Assuming the model can see the screenshot.</strong> In this workflow it cannot, unless you type the visible text, the error code, and a short description or alt text.</p><p><strong>2. Inventing repro steps so the ticket looks standard.</strong> A login step, a URL, or a browser that was not in the paste is a false lead.</p><p><strong>3. Filling environment or identity to look thorough.</strong> A guessed browser, version, timestamp, or user ID will be treated as fact.</p><p><strong>4. Promoting a hedge or a hunch into root cause or expected behavior.</strong> “Please try again later” is not a diagnosis. If expected was not stated, it is MISSING.</p><p><strong>5. Assigning severity so the ticket looks urgent.</strong> If nobody labeled it, write MISSING or a hypothesis labeled as such.</p><h3>Make twelve minutes sacred before you file the ticket</h3><p>Keep the prompt in snippets. When a screenshot or a red banner lands in the chat, type the visible text first, then run gap table → prompt → human pass before you hand it over. The model is a structure engine, not a witness to a browser, a timestamp, or a severity the paste never had.</p><p>When someone says “just file it from the screenshot,” you already have repro steps that stop where the paste stops, expected vs actual with blanks, and a visible gap list — MISSING still showing until a human fills it.</p><p>— -</p><p>*Next failed export: will you hand over a fluent ticket that invents a browser and a P1 — or a bug report whose blanks are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6f6c7b7d0932" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn Scope-Creep Messages into a One-Page Change Request]]></title>
            <link>https://medium.com/@workflow-notes/turn-scope-creep-messages-into-a-one-page-change-request-4d66d8c7b29a?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d66d8c7b29a</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 01:46:19 GMT</pubDate>
            <atom:updated>2026-11-16T01:46:19.340Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste a “can you also…” thread → a one-page change request: proposed adds and cuts, impact only as known, and what they must confirm.</h4><p>Thursday 3:10 p.m. The work has already started. A teammate pings you:</p><blockquote><em>“Don’t just absorb the extras. Turn this thread into a one-page change request: what’s already agreed, what’s being asked on top, what we’d have to cut, the impact we can support, and what they must confirm before anyone starts the add-ons.”</em></blockquote><p>That is not a kickoff agenda, and it is not acceptance criteria on a brand-new brief. It is a <strong>change-control page</strong> after work is already under way, when mid-stream “can you also…?” messages start to rewrite the job. Ask a chatbot to “update the scope” and you often get extras treated as approved, a three-day estimate nobody gave, a Friday date built from “maybe next week,” and an owner assigned to a legal review that still has no name.</p><p>What you need is a <strong>one-page change request</strong>: agreed scope, proposed additions still pending confirmation, cuts (or MISSING if nobody named any), impact only as the messages support it, and what the other side must confirm. Below is a twelve-minute workflow. Paste the thread. Leave unnamed owners as UNASSIGNED, unstated dates as NO DATE, and unstated hours, costs, and rules as MISSING. Free-tier chatbot is enough. The model structures; you refuse to promote a &quot;can you also&quot; into approved scope, and you refuse invented effort, dates, and policy.</p><h3>The scenario we’ll use</h3><p>An illustrative mid-stream thread, incomplete on purpose — like most “while you’re at it” piles after work has started.</p><blockquote><strong><em>Already-agreed scope, from a week-old message in the same thread (illustrative):</em></strong><em> Riley: “Let’s refresh the pricing-page headline and the three plan cards. Copy and layout only — no new payment flow. That’s the scope I’m working from.”</em></blockquote><blockquote><strong><em>This week’s add-on messages (illustrative):</em></strong><em><br>- Morgan: “can you also add a comparison table vs the old plans while you’re in there?”<br>- Riley: “I can look once we confirm it’s in scope — I haven’t started a table.”<br>- Sam: “while you’re at it, can we put a ‘talk to sales’ form on the same page? Legal still needs to review the form copy — I don’t have the legal text yet.”<br>- Morgan: “also the FAQ accordion at the bottom? shouldn’t take long”<br>- Sam: “can we move the launch? maybe next week? or keep the original? not sure”</em></blockquote><h3>Step 1: Separate agreed scope from proposed extras before you paste</h3><p>Do not ask the model to “update the plan to include everything in the thread.” Without a gap pass, “shouldn’t take long” becomes two hours, “maybe next week” becomes a launch date, and three unconfirmed asks become the new official scope.</p><p>Item in the thread | Still open</p><p>Agreed scope: pricing-page headline + three plan cards; copy and layout only; no new payment flow | The only agreed scope in the paste</p><p>Proposed adds: comparison table (Morgan); “talk to sales” form (Sam); FAQ accordion (Morgan) | None confirmed. Riley has not started the table. Legal text: MISSING. Legal-review owner: UNASSIGNED. Hours: MISSING — do not adopt “shouldn’t take long”</p><p>Cuts or trade-offs if extras are accepted | None named: MISSING</p><p>Launch date; cost; policy; approval | Original date not in the paste (NO DATE). “Maybe next week”: not confirmed (NO DATE). Cost, policy, approval: MISSING</p><p>If an add has not been confirmed, it is not agreed scope. <strong>Proposed stays proposed. MISSING / UNASSIGNED / NO DATE stay visible on the page.</strong>Step 2: Run the thread → one-page change request prompt</p><p><strong>Prompt — Scope-creep thread → one-page change request</strong></p><pre>You are turning a pasted message thread into a one-page change<br>request. Work has already started. These are mid-stream add-on<br>requests, not a first brief and not a kickoff.</pre><pre>Hard rules:<br>- Use only facts in the thread. Do not invent owners, dates,<br>  hours, costs, approval status, or policy rules.<br>- Split &quot;already agreed scope&quot; from &quot;proposed additions.&quot; A<br>  &quot;can you also&quot; item stays pending confirmation. Do not<br>  promote it into approved or current scope.<br>- If nobody named a cut, write MISSING. Do not invent one.<br>- Impact only as the messages support it. Do not invent a<br>  day count, hour count, or cost. &quot;Shouldn&#39;t take long&quot; is<br>  not an estimate. Unnamed owners: UNASSIGNED. Unconfirmed<br>  dates: NO DATE. &quot;Maybe next week&quot; is not a date.<br>- This is a change-request page, not a done-when acceptance<br>  checklist and not a kickoff agenda.</pre><pre>Return markdown:<br>## Current agreed scope (from the paste only)<br>## Proposed additions (pending confirmation — not approved scope)<br>## Proposed cuts or trade-offs (only cuts the thread names; else MISSING)<br>## Impact (only as supported; hours, cost, unsupported dates: MISSING / NO DATE)<br>## What the other side must confirm<br>## Still MISSING / UNASSIGNED / NO DATE<br>## What not to write</pre><pre>Thread:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time add-on messages arrive after work has already started.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong> for the sample thread above.</p><p><strong>Current agreed scope:</strong> Refresh the pricing-page headline and the three plan cards. Copy and layout only. No new payment flow.</p><p><strong>Proposed additions (pending confirmation — not approved scope):</strong> (1) comparison table vs the old plans — Morgan; Riley has not started it and asked for confirmation first; (2) a “talk to sales” form — Sam; legal form text MISSING; legal-review owner UNASSIGNED; (3) FAQ accordion — Morgan; hours MISSING; &quot;shouldn&#39;t take long&quot; is not an estimate.</p><p><strong>Proposed cuts or trade-offs:</strong> MISSING — the thread names nothing to take out.</p><p><strong>Impact (only as supported):</strong> The form cannot be finished from this thread (legal copy MISSING; reviewer UNASSIGNED). Launch timing unconfirmed: original date not in the paste (NO DATE); &quot;maybe next week&quot; is not a confirmed date. Hours and cost: MISSING. The &quot;no new payment flow&quot; boundary still stands.</p><p><strong>What the other side must confirm:</strong> which adds are in; what comes out if any add is in, or that nothing comes out (MISSING); who owns legal review (UNASSIGNED) and when the legal text will exist; whether launch timing is changing; that no add is absorbed until confirmed.</p><p><strong>Still MISSING / UNASSIGNED / NO DATE:</strong> confirmation of each add; any cut; legal form text; legal-review owner; launch dates; hours; cost; policy; approval.</p><p><strong>What not to write:</strong> “Scope now includes the table, the form, and the FAQ.” “The FAQ will take two hours.” “Launch is next Friday.” “Sam owns legal review.” “Riley has already started the table.”</p><h3>Step 3: Human pass — five checks before you send the page</h3><p>Do not ask the model to “make this more decisive” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every owner, date, hour, cost, and approval is in the thread or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>Extras stay proposed</strong> — no “can you also” item is written as already-approved scope.</li><li><strong>No invented effort or date</strong> — “shouldn’t take long” and “maybe next week” are not turned into hours or a launch day.</li><li><strong>Cuts are not invented to look tidy</strong> — if the thread named none, the page says MISSING.</li><li><strong>This is not a kickoff agenda, and not an acceptance-criteria checklist</strong> — if the output is a meeting plan or a done-when list for a first brief, it failed. Re-run.</li></ol><p>If a line only works by treating an add-on as approved or by inventing a day count, delete it.</p><h3>Step 4: How to use the page in the actual thread</h3><p>Send the one-pager back and wait for confirmation before starting a table, a form, or an accordion. If someone confirms one add but not the others, update the page yourself — do not pretend all three were approved. If they ask “how long will the FAQ take?” and the thread has no hours, say MISSING; do not adopt &quot;shouldn&#39;t take long.&quot; Until confirmation, the agreed work is still the headline and the three plan cards.</p><p>A change request that already “includes the extras” and “launches next week” is not a change request. It is unauthorized scope.</p><h3>How this differs from “just ask AI to update the scope” — and from acceptance criteria or a kickoff agenda</h3><p>A soft “update the scope” returns a fluent plan in which every mid-stream ask is already in, plus an effort number and a date nobody confirmed. Two nearby workflows are the wrong artifact here:</p><ul><li><strong>Acceptance criteria</strong> turn a *first*, vague brief into a done-when checklist before work starts. This page assumes work has already started, and that a later thread is trying to change it.</li><li><strong>A kickoff agenda</strong> turns a messy brief into a timeboxed first meeting. This page is not a meeting agenda. It is mid-stream change control, and it asks for confirmation *before* extras are absorbed.</li></ul><p>A useful change request is not a new plan that quietly swallowed the thread. It is one page that keeps proposed adds proposed, keeps impact inside the messages, and lists what they must confirm.</p><h3>Common failure modes the prompt should catch</h3><p>The extras written as current scope; “the FAQ will take two hours”; “launch is next Friday” built from “maybe next week”; “Sam owns legal”; a cut you invented so the page looks balanced; “Riley already started the table”; a kickoff agenda or a done-when checklist for a first brief.</p><p>If any of those appear, delete the line. Re-run with the gap table and the agreed-scope sentence pinned at the top of the paste.</p><h3>Five mistakes that fake a “clean” change request</h3><p><strong>1. Promoting “can you also” into already-approved scope.</strong> An unconfirmed ask is a proposal, not current scope.</p><p><strong>2. Inventing hours, costs, or a launch date so impact looks complete.</strong> “Shouldn’t take long” is not an estimate. “Maybe next week” is not a date.</p><p><strong>3. Inventing a cut, an owner, or a policy to make the page look official.</strong> If the thread named none, do not invent them.</p><p><strong>4. Treating a hedge as confirmation.</strong> “I can look once we confirm” is not a start. “Or keep the original? not sure” is not a launch decision.</p><p><strong>5. Sending a kickoff agenda or an acceptance-criteria list instead.</strong> Those are the right tools for a first brief, not for mid-stream add-ons.</p><h3>Make twelve minutes sacred before you absorb the extras</h3><p>Keep the change-request prompt in snippets. When a “can you also…” thread lands on work already started, run gap table → prompt → human pass before you start the add-on. The model is a structure engine, not a witness to approvals or dates the thread never contained.</p><p>When someone says “just add it in while you’re there,” you already have one page of proposed adds, visible gaps, and a confirmation list — with MISSING still showing until a human fills it.</p><p>— -</p><p>*Next “can you also…?” thread: will you absorb the extras into a fluent new plan — or send back a one-page change request whose proposals are still waiting on confirmation?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d66d8c7b29a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn a Bloated Slide Deck into a 5-Minute Talk Track]]></title>
            <link>https://medium.com/@workflow-notes/turn-a-bloated-slide-deck-into-a-5-minute-talk-track-bd15a46bc1cc?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/bd15a46bc1cc</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[meetings]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 01:42:46 GMT</pubDate>
            <atom:updated>2026-11-16T01:42:46.102Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste a slide outline or notes → a 5-minute talk track: opening, one line per slide, close, and likely questions — no invented numbers.</h4><p>Wednesday 9:25 a.m. Your update is at 9:40, and the agenda says five minutes. A teammate drops a link:</p><blockquote><em>“This is the deck — 10 slides including the appendix. Give me a 5-minute talk track: an opening, one sentence on each slide I should actually show, a close, and the questions I should be ready for. Don’t invent a metric the notes don’t have, and don’t turn this into a written FAQ.”</em></blockquote><p>That is not a request for a document digest. It is a request for a <strong>spoken talk track</strong>. Ask a chatbot to “summarize this deck” or “turn these slides into an FAQ” and you often get a confident memo: a conversion rate the notes never stated, an owner nobody named, and five written Q&amp;As — none of which tells you what to say out loud, what to skip, or how to finish in five minutes.</p><p>What you need is a <strong>five-minute talk track</strong>: an opening, one sentence per slide you will actually show, a close, and *likely* questions labeled as hypotheses, with blanks left as MISSING. Below is a twelve-minute workflow. Paste the outline or speaker notes. Leave unnamed owners as UNASSIGNED and unstated dates as NO DATE. Free-tier chatbot is enough. The model paces; you refuse invented figures, owners, quotes, and any rewrite that turns the deck into a FAQ.</p><h3>The scenario we’ll use</h3><p>An illustrative weekly update deck, bloated on purpose — like most files that land just before a five-minute slot.</p><blockquote><strong><em>Deck title (illustrative):</em></strong><em> Customer Onboarding Pilot — Weekly Update</em></blockquote><blockquote><strong><em>Slide outline and speaker notes (illustrative):</em></strong><em><br>1. Title — Customer Onboarding Pilot — Weekly Update<br>2. Agenda — five bullets (background, this week, metrics, risks, next)<br>3. Purpose and scope — “reduce time-to-first-value for new accounts; email + in-app checklist only; phone support is not in this pilot”<br>4. What shipped — “checklist v2 is in staging; welcome-email subject lines drafted”<br>5. What did not ship / risks — “in-app tooltip still blocked on design review”<br>6. Metrics — title only; notes blank<br>7. Customer comment — one account said the checklist was “clearer than last month” (no other quote)<br>8. Next week — “hope to get the tooltip into staging” (no date)<br>9. Appendix — six screenshots, no commentary; competitor comparison, title only, no figures<br>10. Closing — “Questions”</em></blockquote><h3>Step 1: Mark what you can say aloud before you paste</h3><p>Do not ask the model to “summarize the deck” or “make me an FAQ.” Without a gap pass, the blank metrics slide becomes “conversion is up,” and “design review” becomes a named person with a Friday deadline.</p><p>Item in the notes | Still open</p><p>Purpose and scope: time-to-first-value; email + in-app checklist; phone support is out | No baseline or target figure</p><p>Shipped: checklist v2 in staging; welcome-email subject lines drafted | —</p><p>Blocker: tooltip still blocked on design review | Owner: UNASSIGNED. Date: NO DATE</p><p>Metrics slide | All figures: MISSING</p><p>One quoted comment: “clearer than last month” | No other quotes; no named account beyond “one account”</p><p>Next week: “hope to get the tooltip into staging”; appendix screenshots and competitor slide | Calendar date: NO DATE. Commentary and competitor figures: MISSING</p><p>If a figure, owner, date, or quote is not in the notes, do not say it aloud. <strong>MISSING / UNASSIGNED / NO DATE stay visible on the talk track.</strong>Step 2: Run the outline → 5-minute talk track prompt</p><p><strong>Prompt — Bloated slide outline or speaker notes → 5-minute talk track</strong></p><pre>You are turning a pasted slide outline or speaker notes into a<br>5-minute spoken talk track for an upcoming meeting.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent metrics, baselines,<br>  targets, owners, dates, quotes, or competitor figures.<br>- If a slide has a title but no figure, write MISSING. Do not<br>  invent a number so the line sounds finished.<br>- Unnamed owners: UNASSIGNED. Missing dates: NO DATE. &quot;Last<br>  month&quot; or &quot;hope to&quot; is not a calendar date.<br>- This is a spoken talk track, not a written FAQ. Do not output<br>  five written Q&amp;As instead.<br>- Opening and close are lines to say aloud. One spoken sentence<br>  per slide you recommend showing. PARK the rest (especially<br>  appendices with no notes). Pace for five minutes.<br>- Likely questions are hypotheses, not questions already asked,<br>  and not quotes.</pre><pre>Return markdown:<br>## Opening (spoken)<br>## Talk track (one spoken sentence per slide you will show; PARK the rest, with a reason)<br>## Close (spoken) — no invented decision, owner, or date<br>## Likely questions (hypotheses, not real audience questions)<br>   Three hypotheses, each with a response that flags MISSING / UNASSIGNED / NO DATE<br>## Still MISSING / UNASSIGNED / NO DATE<br>## What not to say aloud</pre><pre>Pasted outline or speaker notes:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a long deck arrives too close to a short slot.</p><h3>What a healthy output looks like (illustrative)</h3><p>An <strong>illustrative shape</strong> for the sample above. Remaining shown slides follow the same one-line pattern.</p><p><strong>Opening (spoken, illustrative):</strong> “Five minutes on the customer onboarding pilot — email and the in-app checklist only; phone support is not in this pilot. Checklist v2 is in staging; the tooltip is still blocked on design review. No figures in the metrics notes, and no owner or date for that review, so I won’t invent them.”</p><p><strong>Talk track (illustrative):</strong> PARK slides 1–2 (the agenda is already framed by the opening) and 9 (appendix: no commentary, no competitor figures). Show 3–8 with one spoken sentence each: purpose and scope, what shipped, the tooltip blocker (reviewer UNASSIGNED, date NO DATE), metrics with every figure MISSING, the one quoted comment, and next week as a hope, not a date. Close on slide 10; don&#39;t invent answers.</p><p><strong>Close (spoken, illustrative):</strong> “From the notes: checklist v2 in staging, tooltip still blocked on a design review with no owner or date. If you need a figure I don’t have, I’ll say it’s missing.”</p><p><strong>Likely questions (hypotheses, not real audience questions):</strong></p><ol><li>Hypothesis: someone asks for the conversion rate or baseline. Response: “That’s missing from the notes — I shouldn’t invent a figure in the room.”</li><li>Hypothesis: someone asks when the tooltip will ship, or who owns the design review. Response: “Owner is unassigned, and there’s no date. ‘Hope to get it into staging’ is not a date.”</li><li>Hypothesis: someone wants the appendix walked through. Response: “No figures and no commentary in the notes. I’d rather park it than invent a walkthrough.”</li></ol><p><strong>Still MISSING / UNASSIGNED / NO DATE (illustrative):</strong> metrics figures; baseline; design-review owner (UNASSIGNED); tooltip date (NO DATE); competitor figures; appendix commentary.</p><p><strong>What not to say aloud (illustrative):</strong> “Conversion is up.” “Alex owns the design review.” “We’ll ship Friday.” “Customers are loving it.” A read-through of the appendix.</p><h3>Step 3: Human pass — five checks before you walk in</h3><p>Do not ask the model to “make this punchier” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every figure, owner, date, and quote is in the notes or explicitly MISSING / UNASSIGNED / NO DATE.</li><li><strong>No invented number on a blank metrics slide</strong> — if the notes have no figure, the spoken line has no figure.</li><li><strong>Likely questions stay hypotheses</strong> — not questions already asked, and not quotes.</li><li><strong>Pacing is spoken, and five minutes</strong> — appendices with no notes are parked.</li><li><strong>This is not a FAQ</strong> — if the output is five written Q&amp;As and no talk track, it failed. Re-run.</li></ol><p>If a spoken line only works by inventing a number or an owner, delete it.</p><h3>Step 4: How to use the talk track in the actual slot</h3><p>Use the opening, then speak the one-line track for the slides you show. If someone asks a question that is not one of the three hypotheses, answer the actual question, still without inventing a figure. If they ask “what’s the number on slide 6?” and the notes are blank, say MISSING.</p><p>A talk track that already “has the number” and “names the reviewer” is not a talk track. It is an unauthorized briefing.</p><h3>How this differs from “just ask AI to summarize my slides” — and from a 5-bullet FAQ</h3><p>A soft “summarize this deck” returns a fluent memo, a guessed metric, and a written digest. A long-doc FAQ is the right tool when the input is a policy, an SOP, or a manual, and the output will be read, not spoken. This workflow is different on both ends: the input is a <strong>slide outline or speaker notes</strong>, and the output is a <strong>spoken five-minute track</strong> — pacing, what to say while the slide is on screen, what to park, and hypothesized questions you might hear live.</p><p>A useful talk track is not a shorter FAQ. It is an opening, one honest sentence per shown slide, a close, and a gap list you can defend out loud.</p><h3>Common failure modes the prompt should catch</h3><p>A conversion rate that was never in the notes; “Alex owns the design review”; “we’ll ship the tooltip on Friday”; hypothesized questions written as if already asked; a five-bullet FAQ in place of the talk track; a narrated appendix; a competitor claim with no figures.</p><p>If any of those appear, delete the line. Re-run with the gap table pinned at the top of the paste.</p><h3>Five mistakes that fake a “ready” five-minute talk</h3><p><strong>1. Inventing a metric so the metrics slide has something to say.</strong> A blank notes field is a gap, not permission to sound quantitative.</p><p><strong>2. Treating likely questions as questions already asked.</strong> Hypotheses are rehearsal tools, not evidence.</p><p><strong>3. Reading the appendix because the slides are there.</strong> In a five-minute slot, a screenshot pile with no commentary is parked, not narrated.</p><p><strong>4. Filling an owner or a date so the blocker sounds managed.</strong> UNASSIGNED and NO DATE are speakable. A made-up Friday is a commitment.</p><p><strong>5. Turning the talk track into a written FAQ.</strong> If the reader can only scan answers and cannot hear a paced opening, per-slide line, and close, you have the wrong artifact.</p><h3>Make twelve minutes sacred before the five-minute slot</h3><p>Keep the talk-track prompt in snippets. When a bloated deck arrives too close to a short slot, run gap table → prompt → human pass before you walk in. The model is a pacing engine, not a witness to figures or owners that were never in the notes.</p><p>When someone says “just walk us through the deck,” you already have an opening, one line per shown slide, a close, and a visible gap list — with MISSING still showing until a human fills it.</p><p>— -</p><p>*Next bloated deck: will you walk in with a fluent summary that invents a metric and a date — or a five-minute talk track whose blanks are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bd15a46bc1cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Rehearse a Difficult Conversation: Your Opening Line and Answers to Pushback]]></title>
            <link>https://medium.com/@workflow-notes/rehearse-a-difficult-conversation-your-opening-line-and-answers-to-pushback-8fa26a79ff49?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/8fa26a79ff49</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[communication]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 01:58:19 GMT</pubDate>
            <atom:updated>2026-11-13T01:58:19.617Z</atom:updated>
            <content:encoded><![CDATA[<p>Friday 4:40 p.m. Your one-on-one starts in twenty minutes. You know the ask is coming, and you want to decline — without inventing a policy, a workload number, or a line your manager never said. A teammate pings you:</p><blockquote><em>“Before you go into that conversation, give me an opening line, the three pushbacks you’re most likely to hear, and a response to each. Stay inside what actually happened. Don’t write me a script that puts words in their mouth.”</em></blockquote><p>That is not a request for a speech. It is a request for a <strong>rehearsal brief</strong>. Ask a chatbot to “help me say no” or “help me ask for a raise” and you often get a confident script: a handbook clause that was never cited, a 55-hour week you never documented, and a quote your manager never spoke.</p><p>What you need is a <strong>conversation rehearsal</strong>: an opening line that matches your stated goal, three *likely* pushbacks labeled as hypotheses, and a response to each — with unknowns left as MISSING and no manipulative tactics. Below is a twelve-minute workflow. Paste the background and your goal. The same prompt works before you ask for a raise, decline extra work, chase a payment or a number, or deliver bad news to your manager. Free-tier chatbot is enough. The model structures; you refuse invented policies, figures, and quotes the paste never contained.</p><h3>The scenario we’ll use</h3><p>An illustrative paste, incomplete on purpose — like most notes you have twenty minutes before a one-on-one. The worked example is declining extra work. Swap the background and goal if your conversation is a raise, a payment chase, or bad news; the hard rules stay the same.</p><blockquote><strong><em>Background (illustrative):</em></strong><em> Yesterday’s standup: my manager, Jordan, said “Can you take a first look at the onboarding deck for the new account?” From my own notes, I already own the weekly status for Project A, and the vendor review due Friday.</em></blockquote><blockquote><strong><em>Goal (illustrative):</em></strong><em> I do not want to take on the onboarding deck this cycle unless we name something I should stop doing.</em></blockquote><blockquote><strong><em>What I have not done or been given:</em></strong><em> I have not agreed to take the deck. I have not been given a written scope, a due date, an hours estimate, whose work this is inherited from, or what I may drop. Jordan did not call it a formal assignment, did not name a deadline, and did not say what I should deprioritize.</em></blockquote><blockquote><strong><em>What the paste does NOT contain</em></strong><em> is listed in the gap table below. Do not treat “first look” as a defined scope. Do not invent hours, a policy, or any Jordan quote beyond the one sentence in the paste.</em></blockquote><h3>Step 1: Mark facts, the goal, and blanks before you paste</h3><p>Do not ask the model to “write a script that will get them to yes.” Without a gap pass, “first look” becomes a two-hour task, and “the handbook says I can refuse” appears from nowhere.</p><ul><li><strong>Item in the paste</strong>: Jordan’s only quoted line: “Can you take a first look at the onboarding deck for the new account?” — <strong>Still open</strong>: No other Jordan quotes in the paste</li><li><strong>Item in the paste</strong>: Current work, from my notes: Project A weekly status; vendor review due Friday — <strong>Still open</strong>: Hours I work, and hours the deck would take: MISSING</li><li><strong>Item in the paste</strong>: Goal: decline the deck this cycle unless something comes off my plate — <strong>Still open</strong>: Trade-off item: not yet named</li><li><strong>Item in the paste</strong>: Scope, due date, hours estimate, whose work this is, what I may drop — <strong>Still open</strong>: MISSING</li><li><strong>Item in the paste</strong>: Formal assignment language; a named deadline from Jordan; a deprioritization — <strong>Still open</strong>: Not in the paste</li><li><strong>Item in the paste</strong>: Company policy on workload, “capacity,” or refusal; salary; performance rating — <strong>Still open</strong>: MISSING — do not invent</li></ul><p>If a policy, figure, date, or quote is not in the paste, it is not a fact. <strong>MISSING stays visible on the brief.</strong></p><h3>Step 2: Run the background + goal → rehearsal prompt</h3><p><strong>Prompt — Background and goal → opening line, 3 pushbacks, and responses</strong></p><pre>You are turning a pasted background and a stated goal into a short<br>rehearsal for a difficult workplace conversation.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent company policies,<br>  handbook clauses, salary bands, market data, performance ratings,<br>  hours worked, hours a task will take, due dates, or quotes the<br>  other person has not said in the paste.<br>- If a figure, date, policy, scope, or quote is absent, write<br>  MISSING. Do not fill it so the script sounds complete.<br>- The opening line must match the stated goal and must not add<br>  facts the paste does not contain.<br>- The three pushbacks are hypotheses about what the other person<br>  might say. Label them as hypotheses, not as quotes, and not as<br>  predictions presented as fact.<br>- Responses may not use guilt, false deadlines, flattery-as-a-trap,<br>  or other manipulative tactics. They may ask for a missing fact<br>  or a trade-off the paste already frames as the goal.<br>- Do not recommend a path the paste did not authorize (for example<br>  escalating to HR, threatening to resign, or naming a colleague<br>  to take the work) unless that path is already in the paste.</pre><pre>Return markdown with exactly these sections:</pre><pre>## Opening line<br>- One short opening, facts and goal only.</pre><pre>## Three likely pushbacks (hypotheses)<br>1. Hypothesis, not a quote.<br>2. Hypothesis, not a quote.<br>3. Hypothesis, not a quote.</pre><pre>## A response to each<br>- A response that stays inside the paste, flags MISSING facts,<br>  and does not manipulate.</pre><pre>## Still MISSING<br>- Bullet each absent policy, figure, date, scope, or quote<br>  the conversation may need.</pre><pre>## What not to say<br>- Lines that would invent a policy, a number, a quote, or a<br>  tactic the paste does not support.</pre><pre>Background:<br>[paste]</pre><pre>Goal:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a hard conversation is coming and your notes are incomplete — which they usually are.</p><h3>What a healthy output looks like (illustrative)</h3><p>Not a live one-on-one transcript — an <strong>illustrative shape</strong> for the sample paste above.</p><p><strong>Opening line (illustrative):</strong> “Jordan, I want to be clear before this becomes a formal assignment. I can keep Project A’s weekly status and the vendor review due Friday. I can’t also take the new-account onboarding deck this cycle unless we name something I should stop. I don’t have a due date, a scope, or an hours estimate for the deck — those are still missing.”</p><p><strong>Three likely pushbacks (hypotheses, not quotes):</strong></p><ol><li>Hypothesis: Jordan says it is only a first look and will not take long.</li><li>Hypothesis: Jordan says there is no one else who can do it.</li><li>Hypothesis: Jordan says this needs to be done before Friday, or that it is visible to leadership.</li></ol><p><strong>A response to each (illustrative):</strong></p><ol><li>“If it’s only a first look, can we write down what ‘done’ means and how much time you’re asking for? I don’t have either of those yet.” (Scope and hours are MISSING. Do not agree to an unbounded first look.)</li><li>“If there’s no one else, I still need a trade-off — Project A status or the Friday vendor review. I don’t have a name of someone who can take the deck, and I shouldn’t invent one.”</li><li>“A Friday deadline for the deck, and any note that leadership is watching, were not in yesterday’s standup line. If those are now part of the ask, I need them stated, and I still need something taken off my plate.”</li></ol><p><strong>Still MISSING (illustrative):</strong> written scope; due date for the deck; hours estimate; whose work this is; what may be dropped; any Jordan quote beyond the “first look” sentence; company workload policy; hours I work.</p><p><strong>What not to say (illustrative):</strong> “I already work 55 hours.” “The handbook says I can refuse.” “You told me this was optional.” “Everyone knows Q4 is overloaded.” A flattery line used as a trap. A guilt line. A fake counter-deadline.</p><p>Notice what stayed out: no invented policy, no invented hours, no extra Jordan quotes, no colleague volunteered on their behalf, no HR threat, no “just say no and walk out.”</p><h3>Step 3: Human pass — five checks before you walk in</h3><p>Do not ask the model to “make this more persuasive” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every policy, figure, date, and quote is in the paste or explicitly MISSING.</li><li><strong>No invented quotes</strong> — hypothesized pushbacks stay labeled as hypotheses. Jordan is not given lines the paste did not contain.</li><li><strong>No manipulative tactics</strong> — no guilt, no false urgency you invented, no flattery-as-a-trap, no fabricated deadline.</li><li><strong>The opening matches the goal</strong> — here, decline this cycle unless a trade-off is named; the opening does not quietly accept the deck.</li><li><strong>“Still MISSING” is not empty if scope or dates are blank</strong> — a complete-sounding script that filled the hours and the due date is a failed brief.</li></ol><p>If a line only works by inventing a policy or a quote, delete it.</p><h3>Step 4: How to use the brief in the actual conversation</h3><p>Use the opening line, then listen. If Jordan says something that is not one of the three hypotheses, do not force a memorized response; answer the actual sentence, still without inventing facts. If they name a due date or a trade-off in the room, write it down yourself — the brief must not pretend it was already known. If they ask “how long would a first look take?” and the paste has no hours, say you don’t have that number. After the meeting, one human note should carry any real decision, any real date, and nothing that is still MISSING.</p><p>A rehearsal that already “wins” the conversation on paper is not a rehearsal. It is a script that will break at the first unexpected sentence.</p><h3>How this differs from “just ask AI to write me a script”</h3><p>A soft “write me a script” returns fluent lines, a borrowed policy, a workload number you never recorded, and a prediction of the other person’s feelings. A rehearsal brief caps itself at the paste. It keeps the opening inside the goal, labels pushbacks as hypotheses, leaves MISSING visible, and refuses tactics that pressure the other person with facts you do not have.</p><p>A useful rehearsal is not a speech with no blanks. It is a short opening plus honest gaps, so you can stay in the conversation when it leaves the script.</p><h3>Common failure modes the prompt should catch</h3><p>“I already work 55 hours”; a handbook clause; “you said this was optional” when the paste only has “first look”; hypothesized pushbacks written as quotations from Jordan; “leadership is already expecting this” stated as fact; a colleague named to take the work; an HR escalation you did not plan; guilt or flattery used as leverage.</p><p>If any of those appear, delete the line. Re-run with the gap table and Jordan’s only quoted sentence pinned at the top of the paste.</p><h3>Five mistakes that fake a “ready” script</h3><p><strong>1. Inventing a company policy or a salary band so the ask sounds official.</strong> If the paste has no handbook line, the brief has no handbook line.</p><p><strong>2. Inventing quotes the other person never said.</strong> Hypotheses are rehearsal tools. They are not evidence.</p><p><strong>3. Using manipulative tactics to “win” the room.</strong> Guilt, false urgency, and flattery-as-a-trap are not a rehearsal; they are pressure. This workflow does not use them.</p><p><strong>4. Filling MISSING dates, hours, or scope so the script sounds complete.</strong> A made-up “two-hour first look” becomes a commitment the moment you say it.</p><p><strong>5. Treating the three pushbacks as a complete map of the conversation.</strong> If they say something else, listen. Do not drag them back onto a card they did not play.</p><h3>Make twelve minutes sacred before you go in</h3><p>Keep the rehearsal prompt in snippets. When a hard conversation is coming, run gap table → prompt → human pass before you sit down. The model is a structure engine, not a witness to policies, hours, or sentences that were never said.</p><p>When someone says “just give me a script,” you already have an opening line, three labeled hypotheses, and a visible gap list — with MISSING still showing until a human fills it.</p><p>— -</p><p>*Next difficult conversation: will you walk in with a fluent script that invents a policy, a number, and a quote — or a rehearsal whose blanks are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8fa26a79ff49" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn a Raw Spreadsheet Export into 3 Takeaways Before Your Next Meeting]]></title>
            <link>https://medium.com/@workflow-notes/turn-a-raw-spreadsheet-export-into-3-takeaways-before-your-next-meeting-e1910e596960?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/e1910e596960</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[meetings]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 01:20:30 GMT</pubDate>
            <atom:updated>2026-11-13T01:20:30.610Z</atom:updated>
            <content:encoded><![CDATA[<p>Tuesday 8:50 a.m. The weekly review starts at 9:00. A teammate drops a file on you:</p><blockquote><em>“This is the raw export. Give me three takeaways, one number that looks off, and one question to ask in the room. Do not invent a story about why the numbers moved.”</em></blockquote><p>That is not a request for a dashboard narrative. It is a request for a <strong>meeting brief</strong>. Ask a chatbot to “analyze this spreadsheet” and you often get a confident memo: a cause nobody measured, a blank cell filled with last week’s average, and a company total that quietly ignores a missing row.</p><p>What you need is a <strong>three-takeaway brief</strong>: three points the table actually supports, one anomaly you can compute, and one follow-up question — with blanks left blank and causes left unclaimed. Below is a twelve-minute workflow. Paste the export (sales, survey scores, or a timesheet — the same prompt works). Leave blank cells as MISSING. Free-tier chatbot is enough. The model computes; you refuse invented causes, filled-in blanks, and totals the missing cells make impossible.</p><h3>The scenario we’ll use</h3><p>An illustrative weekly sales export, incomplete on purpose — like most files that land on your desk at 8:50.</p><blockquote><strong><em>Weekly units by region (illustrative):</em></strong></blockquote><ul><li><strong>Region</strong>: North — <strong>Week of 10/6</strong>: 42 — <strong>Week of 10/13</strong>: 45 — <strong>Week of 10/20</strong>: 44 — <strong>Week of 10/27</strong>: 48</li><li><strong>Region</strong>: South — <strong>Week of 10/6</strong>: 31 — <strong>Week of 10/13</strong>: 29 — <strong>Week of 10/20</strong>: 18 — <strong>Week of 10/27</strong>: 28</li><li><strong>Region</strong>: East — <strong>Week of 10/6</strong>: 22 — <strong>Week of 10/13</strong>: 24 — <strong>Week of 10/20</strong>: 25 — <strong>Week of 10/27</strong>: 23</li><li><strong>Region</strong>: West — <strong>Week of 10/6</strong>: 15 — <strong>Week of 10/13</strong>: (blank) — <strong>Week of 10/20</strong>: 16 — <strong>Week of 10/27</strong>: 14</li></ul><p><strong><em>What the export does NOT contain</em></strong><em> is listed in the gap table below. There are no targets, quotas, year-ago figures, or notes that explain movement. A blank cell is not a zero. “Week of” labels are labels, not a claim that every week is complete.</em></p><h3>Step 1: Mark blanks, and what you can actually add, before you paste</h3><p>Do not ask the model to “find insights.” Without a gap pass, West’s blank becomes 15, and South’s dip becomes “seasonality” or “a staffing problem.”</p><ul><li><strong>Export fact</strong>: North: 42, 45, 44, 48 — <strong>Still open</strong>: Four-week sum can be computed: 42+45+44+48 = 179. Cause of the last-week rise: not in the export</li><li><strong>Export fact</strong>: South: 31, 29, 18, 28 — <strong>Still open</strong>: Four-week sum can be computed: 31+29+18+28 = 106. Cause of the week-of-10/20 dip: not in the export</li><li><strong>Export fact</strong>: East: 22, 24, 25, 23 — <strong>Still open</strong>: Four-week sum can be computed: 22+24+25+23 = 94</li><li><strong>Export fact</strong>: West: 15, blank, 16, 14 — <strong>Still open</strong>: Week of 10/13: MISSING. Four-week West total: cannot be computed. Three known weeks sum to 15+16+14 = 45</li><li><strong>Export fact</strong>: Company total, week of 10/6 / 10/20 / 10/27 — <strong>Still open</strong>: Can be computed: 110, 103, 113</li><li><strong>Export fact</strong>: Company total, week of 10/13 — <strong>Still open</strong>: MISSING (West is blank). Do not report a four-region total for that week</li><li><strong>Export fact</strong>: Targets, quotas, prior-year figures, reasons for movement — <strong>Still open</strong>: MISSING</li></ul><p>If a cell, total, target, or cause is not in the export, it is not a fact. <strong>MISSING stays visible on the brief.</strong></p><h3>Step 2: Run the export → three-takeaway brief prompt</h3><p><strong>Prompt — Raw spreadsheet export → 3 takeaways, 1 anomaly, 1 follow-up question</strong></p><pre>You are turning a raw spreadsheet export into a short brief for someone<br>about to walk into a meeting.</pre><pre>Hard rules:<br>- Use only figures and labels in the pasted table. Do not invent<br>  cells, totals, targets, quotas, year-ago figures, or causes.<br>- If a cell is blank or marked blank, write MISSING. A blank is not<br>  zero. Do not fill it with an average, a prior week, or a guess.<br>- If a total or comparison cannot be computed because of a MISSING<br>  cell, say so. Do not report that total as a fact.<br>- Takeaways must be computable from the table. Show the supporting<br>  arithmetic briefly next to each point.<br>- Do not invent a cause for any movement. If you speculate, label it<br>  explicitly as a hypothesis, not a finding. Prefer to leave cause<br>  as MISSING.<br>- Do not forecast, and do not recommend a restructure, hire, or<br>  budget change the table does not support.<br>- Return exactly three takeaways, one anomaly, and one follow-up<br>  question — not a hidden fourth takeaway dressed up as the question.</pre><pre>Return markdown with exactly these sections:</pre><pre>## 3 takeaways<br>1. A point computable from the table, with brief arithmetic.<br>2. A second computable point.<br>3. A third computable point, which may be that a total cannot be<br>   computed because of MISSING cells.</pre><pre>## 1 anomaly<br>- The value that stands out by arithmetic (for example the largest<br>  week-to-week drop among known adjacent values). Show the<br>  comparison. Cause: MISSING unless the table itself states one.<br>  A blank cell is a data gap, not automatically the anomaly.</pre><pre>## 1 suggested follow-up question<br>- One question a human should ask in the meeting. It may request a<br>  missing cell or source data that could explain an anomaly.<br>  Do not disguise a recommendation as a question.</pre><pre>## Gaps still open<br>- Bullet each MISSING cell, blocked total, and absent cause or<br>  target.</pre><pre>Pasted export:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a raw export arrives too close to a meeting.</p><h3>What a healthy output looks like (illustrative)</h3><p>Not a live workbook — an <strong>illustrative shape</strong> for the sample export above.</p><p><strong>3 takeaways (illustrative):</strong></p><ol><li>North has the highest complete four-week total in the export, 179 (42+45+44+48), and its week of 10/27 (48) is its highest week in the table.</li><li>For the three weeks with all four regions filled (10/6, 10/20, 10/27), company totals are 110, 103, and 113 (North+South+East+West). The week of 10/20 is the lowest complete company week, 103, which is 7 below the week of 10/6 (110).</li><li>A four-week West total, and a four-region total for the week of 10/13, cannot be computed: West’s week of 10/13 is MISSING. West’s three known weeks sum to 45 (15+16+14).</li></ol><p><strong>1 anomaly (illustrative):</strong> South in the week of 10/20 is 18. That is 11 below the prior week (29) and 11.3 below the average of South’s other three weeks, (31+29+28)/3 = 29.3. It is the largest week-to-week drop among adjacent known values in the table. Cause: MISSING — the export does not contain one.</p><p><strong>1 suggested follow-up question (illustrative):</strong> “Can someone supply the missing West figure for the week of 10/13, and is there source data — not in this export — that explains South’s drop from 29 to 18 in the week of 10/20?”</p><p><strong>Gaps still open (illustrative):</strong> West, week of 10/13: MISSING. Company total for that week: cannot be computed. West four-week total: cannot be computed. Targets, quotas, year-ago figures, and causes of movement: MISSING.</p><p>Notice what stayed out: no “South is underperforming this quarter” (the table has no target); no “North’s promotion worked”; no West total of 60 built from a filled blank; no company total of 98 for the week of 10/13 treated as complete; no forecast; no “we should restructure South.”</p><h3>Step 3: Human pass — five checks before you walk into the room</h3><p>Do not ask the model to “make this more insightful” until the gaps are locked.</p><ol><li><strong>Fact parity</strong> — every figure is in the table or explicitly MISSING. Supporting arithmetic matches the cells.</li><li><strong>No invented cause</strong> — “seasonality,” “staffing,” or “a key account left” is not a takeaway unless the export says so. A guess, if you keep one, stays labeled as a hypothesis.</li><li><strong>Three means three</strong> — the follow-up is a question, not a hidden fourth finding or a recommendation.</li><li><strong>Incomplete math stays incomplete</strong> — West has no four-week total; the week of 10/13 has no four-region total.</li><li><strong>A blank is a gap, not automatically the anomaly</strong> — West’s missing cell is flagged as MISSING; the computed anomaly is South&#39;s drop to 18.</li></ol><p>If a takeaway only works by filling a blank or inventing a why, rewrite it or drop it.</p><h3>Step 4: How to use the brief in the actual meeting</h3><p>Lead with the three takeaways, then the anomaly and the question. Do not read a cause you do not have. If someone asks “what’s the West number for the 13th?” and the cell is still blank, say MISSING. If someone asks &quot;why did South drop?,&quot; say the export does not contain a cause, and ask whether anyone has source data that does. After the meeting, a human should fill the missing cell or record the cause — the brief should not pretend either was already known.</p><p>A brief that already “explains” the dip and presents a complete company trend is not a brief. It is an unauthorized story.</p><h3>How this differs from “just ask AI to analyze this spreadsheet”</h3><p>A soft “analyze this” returns a fluent narrative, a guessed cause, a filled-in blank, and a recommendation that feels like strategy. A three-takeaway brief caps itself at the table: the arithmetic is visible, MISSING stays visible, causes stay unclaimed, and the question you take into the room is a question.</p><p>A useful brief is not a shorter dashboard story. It is three supported points, one computed anomaly, and an honest gap list.</p><h3>Common failure modes the prompt should catch</h3><p>West’s blank filled with 15 or with an average; a company total for the week of 10/13 reported as 98 and treated as complete; “South is underperforming the quarter” with no target in the table; “North’s promotion worked”; a forecast for November; “we should restructure South” dressed up as the follow-up question.</p><p>If any of those appear, delete the cell. Re-run with the gap table pinned at the top of the paste.</p><h3>Five mistakes that fake an “insightful” pre-meeting brief</h3><p><strong>1. Inventing a cause so the takeaway sounds like analysis.</strong> A dip is a number. A why is a fact only if the export (or a human) supplies it.</p><p><strong>2. Filling blank cells with averages or last week.</strong> A filled blank makes every later total look more complete than the file is.</p><p><strong>3. Computing totals that the blanks make impossible.</strong> If West is MISSING for a week, that week&#39;s company total is not a fact.</p><p><strong>4. Treating a missing cell as the anomaly.</strong> The blank is a data gap. The anomaly is the value that stands out by arithmetic — here, South’s drop to 18.</p><p><strong>5. Turning the follow-up question into a recommendation.</strong> “Can we get the missing West figure?” is a question. “We should restructure South” is not, unless the table already supports that action, which this one does not.</p><h3>Make twelve minutes sacred before the review</h3><p>Keep the three-takeaway prompt in snippets. When a raw export arrives too close to a meeting, run gap table → prompt → human pass before you walk in. The model is a calculation engine, not a witness to why the numbers moved.</p><p>When someone says “just give me the highlights,” you already have three points, one computed anomaly, and a visible gap list — with MISSING still showing until a human fills it.</p><p>— -</p><p>*Next raw export: will you walk in with a fluent story that invents a cause and a filled-in blank — or three takeaways that stay honest about what the table actually shows?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e1910e596960" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Draft a Leave Handover Brief from Your Calendar and Open Tasks]]></title>
            <link>https://medium.com/@workflow-notes/draft-a-leave-handover-brief-from-your-calendar-and-open-tasks-c4456c321c31?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4456c321c31</guid>
            <category><![CDATA[time-management]]></category>
            <category><![CDATA[work]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 01:30:09 GMT</pubDate>
            <atom:updated>2026-11-12T01:30:09.130Z</atom:updated>
            <content:encoded><![CDATA[<p>Wednesday 4:50 p.m. You are out tomorrow and Friday. A teammate pings you:</p><blockquote><em>“Before you go, send a handover: who covers which meeting, what’s still open, and who I should ping if something breaks. Don’t invent a cover if nobody agreed to take it.”</em></blockquote><p>That is not a request for a cheerful out-of-office note. It is a request for a <strong>coverage map</strong>. Ask a chatbot to “write my handover” and you often get a tidy document in which every meeting has a named cover, every task is due before Monday, and “ask around” has turned into a person who never agreed.</p><p>What you need is a <strong>leave handover brief</strong>: who must cover what, which decisions are still open, and who the cover person should escalate to — with blanks left blank. Below is a fifteen-minute workflow. Paste your leave window, meetings, and open tasks; leave blank covers as UNASSIGNED, blank dates as NO DATE, and blank facts as MISSING. Free-tier chatbot is enough. The model structures; you refuse invented covers, dates, and escalation names the paste never contained.</p><h3>The scenario we’ll use</h3><p>An illustrative paste, incomplete on purpose — like most handovers written the afternoon before leave.</p><blockquote><strong><em>Leave window (illustrative):</em></strong><em> “I am out Thursday and Friday. Back Monday morning.”</em></blockquote><blockquote><strong><em>Calendar (illustrative):</em></strong><em><br>- Thu 10:00–10:30 Weekly standup. Note already on the event: “Jordan to cover.”<br>- Thu 15:00–15:45 Client check-in. No cover noted.<br>- Fri 11:00–12:00 Vendor review. No cover noted.<br>- Fri 14:00–15:00 Internal project review. Attendees are listed; no cover is named.</em></blockquote><blockquote><strong><em>Open tasks (illustrative):</em></strong><em> Finish the onboarding deck (no due date); reply to the invoice question (who is waiting is not written); “Help with the legal thing?” (owner and next action unclear); scope add-on still waiting on absorb vs re-quote (no decision); weekly status (no due date).</em></blockquote><blockquote><strong><em>Note from the paste:</em></strong><em> “If something urgent happens, ask around on the team channel.”</em></blockquote><blockquote><strong><em>What the paste does NOT contain</em></strong><em> is listed in the gap table below. The paste names only one cover (Jordan, standup). “Ask around” is not an escalation person. Weekday words are not numbered calendar dates.</em></blockquote><h3>Step 1: Separate named covers from blanks before you paste</h3><p>Do not ask the model to “write a professional handover.” Without a split, Jordan — the only named cover — will be assigned every meeting, and “ask around” will become a director.</p><ul><li><strong>Item in the paste:</strong> Leave: out Thursday and Friday; back Monday morning — <strong>Still open:</strong> Weekday words only; no numbered calendar date in the paste</li><li><strong>Item in the paste:</strong> Thu 10:00 standup — <strong>Still open:</strong> Cover: Jordan (named in the event). Backup if Jordan is out: MISSING</li><li><strong>Item in the paste:</strong> Thu 15:00 client check-in / Fri 11:00 vendor review / Fri 14:00 project review — <strong>Still open:</strong> Cover: UNASSIGNED; attendees listed are not a cover</li><li><strong>Item in the paste:</strong> Onboarding deck / weekly status — <strong>Still open:</strong> Due: NO DATE; cover: UNASSIGNED</li><li><strong>Item in the paste:</strong> Invoice question — <strong>Still open:</strong> Who is waiting: MISSING; due: NO DATE; cover: UNASSIGNED</li><li><strong>Item in the paste:</strong> Legal thing (hedged) — <strong>Still open:</strong> Owner: UNASSIGNED; next action: MISSING</li><li><strong>Item in the paste:</strong> Scope add-on, absorb vs re-quote — <strong>Still open:</strong> Decision: not in the paste; do not invent one</li><li><strong>Item in the paste:</strong> “Ask around on the team channel” — <strong>Still open:</strong> Escalation person: UNASSIGNED; channel name: MISSING</li></ul><p>If a cover, date, or escalation name is not in the paste, it is not a fact. <strong>MISSING / UNASSIGNED / NO DATE stay visible on the brief.</strong></p><h3>Step 2: Run the calendar + tasks → handover brief prompt</h3><p><strong>Prompt — Leave dates, calendar, and open tasks → handover brief</strong></p><pre>You are turning a leave window, calendar meetings, and open tasks<br>into a handover brief for the people covering the time off.</pre><pre>Hard rules:<br>- Use only facts in the paste. Do not invent coverage owners,<br>  due dates, decisions, escalation names, or channel names.<br>- If a meeting or task has no named cover, write UNASSIGNED.<br>  Do not extend a cover who was named for one meeting onto<br>  other meetings.<br>- If a due date or numbered calendar date is absent, write NO DATE.<br>  Weekday words may be repeated as written; do not convert them<br>  into a dated calendar.<br>- If an owner, waiting party, channel, backup, or next action is<br>  absent or hedged (&quot;ask around&quot;), write UNASSIGNED or MISSING.<br>- Open decisions stay open. Do not close a choice the paste did<br>  not make.<br>- &quot;Who to escalate to&quot; may only name a person or role in the<br>  paste. Otherwise write UNASSIGNED.</pre><pre>Return markdown with exactly these sections:</pre><pre>## Leave window (from paste)<br>- Out / back only as written. Flag NO DATE if no numbered date.</pre><pre>## Coverage (meetings)<br>| Meeting | When (from paste) | Cover | Notes / gaps |<br>| … | … | UNASSIGNED unless named | Backup MISSING unless named |</pre><pre>## Open tasks<br>| Task | Cover | Due | Notes / gaps |<br>| … | UNASSIGNED unless named | NO DATE unless named | Flag MISSING details |</pre><pre>## Open decisions<br>- Only decisions the paste already marks as open. Do not resolve them.</pre><pre>## Escalation for the cover person<br>- Who to ping, and how, only if named. Otherwise UNASSIGNED / MISSING.</pre><pre>## Still cannot hand off (do not fake-close)<br>- Meetings or tasks that need a human to name a cover, a date,<br>  or a missing detail before you leave.</pre><pre>Leave window:<br>[paste]</pre><pre>Calendar:<br>[paste]</pre><pre>Open tasks:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time you go out with meetings still on the calendar.</p><h3>What a healthy output looks like (illustrative)</h3><p>Not a live personal handover — an <strong>illustrative shape</strong> for the sample paste above.</p><p><strong>Leave window (from paste):</strong> Out Thursday and Friday; back Monday morning. No numbered calendar date (NO DATE for a dated form).</p><p><strong>Coverage (illustrative):</strong> Thu 10:00 standup — Jordan (named on the event); backup MISSING. Thu 15:00 client check-in, Fri 11:00 vendor review, and Fri 14:00 project review — cover UNASSIGNED. Attendees listed on the project review are not a cover.</p><p><strong>Open tasks (illustrative):</strong> Onboarding deck, invoice question, and weekly status — cover UNASSIGNED; due NO DATE. Invoice waiting party: MISSING. Legal thing — owner UNASSIGNED; next action MISSING; do not promote it into a covered task.</p><p><strong>Open decisions (illustrative):</strong> Scope add-on, absorb vs re-quote — still open. Do not pick a side or assign an unnamed decider.</p><p><strong>Escalation (illustrative):</strong> The paste says “ask around on the team channel.” Escalation person: UNASSIGNED. Channel name: MISSING.</p><p><strong>Still cannot hand off:</strong> the three UNASSIGNED meetings need a named cover before Thursday; invoice needs a waiting party or a cover; legal help needs an owner and a next action; escalation needs a person if “ask around” is not enough.</p><p>Notice what stayed out: no “Jordan will cover all four meetings,” no “deck due Friday,” no invented legal owner, no “escalate to the director,” no fake close on absorb vs re-quote, no invented channel name.</p><h3>Step 3: Human pass — five checks before you send the handover</h3><p>Do not ask the model to “fill the gaps so this looks complete” until the blanks are locked.</p><ol><li><strong>Fact parity</strong> — every cover, date, decision, and escalation name is in the paste or explicitly UNASSIGNED / NO DATE / MISSING.</li><li><strong>One cover does not become every cover</strong> — Jordan stays on standup only.</li><li><strong>Open decisions stay open</strong> — absorb vs re-quote is not closed in the brief.</li><li><strong>Hedged escalation stays hedged</strong> — “ask around on the team channel” does not become a named person or a named channel.</li><li><strong>“Still cannot hand off” is not empty if covers are blank</strong> — UNASSIGNED meetings must remain visible, not dropped without a flag.</li></ol><p>If a meeting would be unattended on Thursday, name a cover yourself, or flag it — do not let the model invent one.</p><h3>Step 4: How to use the brief before you actually leave</h3><p>Send the coverage table and the “still cannot hand off” list first. This breaks if it waits until Thursday morning. Ask a human to take each UNASSIGNED meeting, in writing. If nobody takes it, say so in the calendar note; do not publish a fake cover. For tasks with due NO DATE, tell the cover what “done enough” means — the model must not mint a Friday deadline. When you return Monday, compare the brief to what actually moved.</p><p>A handover that assigns every row and names an escalator you never asked is not a handover — it is a surprise job description.</p><h3>How this differs from “just ask AI to write my out-of-office handover”</h3><p>A soft “write my handover” returns a complete-looking memo: a cover on every meeting, a due date on every task, and “ping the director if needed.” A leave handover brief caps itself at the paste. It keeps a named cover specific, leaves UNASSIGNED / NO DATE / MISSING visible, keeps open decisions open, and treats “still cannot hand off” as part of the output.</p><p>A useful handover is not a document with no blanks. It is a document whose blanks are still readable.</p><h3>Common failure modes the prompt should catch</h3><p>Jordan auto-assigned to other meetings; “ask around” turned into a director; “Monday morning” turned into a dated return; “deck due Friday”; an invented legal owner; absorb vs re-quote quietly closed; an invented channel name; attendees treated as the cover.</p><p>If any of those appear, delete the cell. Re-run with the gap table and the one named cover (Jordan, standup only) pinned at the top of the paste.</p><h3>Five mistakes that fake a “complete” handover</h3><p><strong>1. Inventing covers so every meeting looks staffed.</strong> An UNASSIGNED client check-in is uncomfortable. A fake name is worse: the client still meets nobody.</p><p><strong>2. Extending one named cover to the whole calendar.</strong> Jordan agreed to standup, per the event note — not to Friday’s vendor review.</p><p><strong>3. Minting due dates so work looks safe until you return.</strong> Anxiety is not a due date. If the paste has no date, the brief stays NO DATE.</p><p><strong>4. Turning “ask around” into an escalation person.</strong> A cover who follows that line will ping someone who was never asked.</p><p><strong>5. Closing open decisions in the handover.</strong> The brief is not the place to pick absorb vs re-quote, or invent who will pick it, unless the paste already says so.</p><h3>Make fifteen minutes sacred the afternoon before you go</h3><p>Keep the handover prompt in snippets. When leave is coming and meetings remain, run gap table → prompt → human pass, then fill UNASSIGNED covers with real names — or leave them visibly open. The model is a structure engine, not a staffing desk.</p><p>When someone says “just send a handover,” you already have a coverage table and an escalation line that still shows UNASSIGNED and MISSING until a human fills them.</p><p>— -</p><p>*Next time you go out: will you send a tidy memo that invents a cover, a due date, and an escalator — or a handover brief whose blanks are still honest?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4456c321c31" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn a Messy Email Thread into a 3-Line Brief for Your Manager]]></title>
            <link>https://medium.com/@workflow-notes/turn-a-messy-email-thread-into-a-3-line-brief-for-your-manager-ad4d306f9efa?source=rss-ae857f7c0838------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad4d306f9efa</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[email]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Workflow Notes]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 01:22:20 GMT</pubDate>
            <atom:updated>2026-11-12T01:22:20.131Z</atom:updated>
            <content:encoded><![CDATA[<h4>Paste the thread → background, where it’s stuck, and the next step — or why your manager must decide. No invented owners or dates.</h4><p>Thursday 9:10 a.m. Your manager pings you ten minutes before standup:</p><blockquote><em>“I cannot read a forty-message thread before the meeting. Give me three lines: what started this, where it is stuck, and what you need from me. Do not invent a decision we have not made.”</em></blockquote><p>That is not a request for a fluent summary. It is a request for a <strong>decision brief</strong>. Ask a chatbot to “catch my manager up” and you often get a confident memo: a named owner who was never in the thread, a Friday built from “soon,” and a recommendation to absorb work nobody agreed to absorb.</p><p>What you need is a <strong>3-line manager brief</strong>: background, where it is stuck now, and a recommended next step — or why only your manager can decide. Below is a twelve-minute workflow. Paste the thread; leave blank owners as UNASSIGNED, blank dates as NO DATE, and blank facts as MISSING. Free-tier chatbot is enough. The model compresses; you refuse invented decisions, owners, and dates the thread never contained.</p><h3>The scenario we’ll use</h3><p>An illustrative thread, incomplete on purpose — like most threads that land on your desk at 9:00.</p><blockquote><strong><em>Email thread (illustrative):</em></strong></blockquote><blockquote><em>From: Morgan (client) — “Can we add the extra usage report to this month’s delivery? We need it before the review.”</em></blockquote><blockquote><em>From: Sam (internal) — “We didn’t quote this. I can draft a scope note, but I don’t know if we absorb it or re-quote. Also the vendor said the data feed might slip.”</em></blockquote><blockquote><em>From: vendor contact — “The data feed is delayed. We’ll update you when we have a new date.”</em></blockquote><blockquote><em>From: Riley (internal) — “Legal still hasn’t confirmed the wording on the add-on. Not sure who owns chasing that.”</em></blockquote><blockquote><em>From: Morgan (client) — “Please confirm soon? Our review is coming up. Who is the point person?”</em></blockquote><blockquote><strong><em>What the thread does NOT contain</em></strong><em> is listed in the gap table below. Do not treat “soon,” “this month,” “coming up,” or “when we have a new date” as calendar dates.</em></blockquote><h3>Step 1: Mark decisions, owners, and dates before you paste</h3><p>Do not ask the model to “summarize this for my boss.” Without a gap pass, “confirm soon” becomes Friday, and “not sure who owns chasing that” becomes a named legal owner.</p><ul><li><strong>Thread fact:</strong> Extra usage report for “this month’s delivery,” needed “before the review” — <strong>Still open:</strong> Delivery date: NO DATE; review date: NO DATE</li><li><strong>Thread fact:</strong> Work was not quoted; Sam can draft a scope note; absorb vs re-quote — <strong>Still open:</strong> Decision: not in the thread; do not invent one</li><li><strong>Thread fact:</strong> Vendor data feed delayed; update “when we have a new date” — <strong>Still open:</strong> New feed date: NO DATE</li><li><strong>Thread fact:</strong> Legal wording on the add-on unconfirmed — <strong>Still open:</strong> Chaser: UNASSIGNED</li><li><strong>Thread fact:</strong> Client asks for a point person and a confirmation “soon” — <strong>Still open:</strong> Point person: UNASSIGNED; “soon” is not a date (NO DATE)</li><li><strong>Thread fact:</strong> Cost, hours, or fee for the extra report — <strong>Still open:</strong> MISSING</li></ul><p>If a decision, owner, or date is not in the thread, it is not a fact. <strong>MISSING / UNASSIGNED / NO DATE stay visible on the brief.</strong></p><h3>Step 2: Run the messy-thread → 3-line brief prompt</h3><p><strong>Prompt — Email thread → 3-line manager brief</strong></p><pre>You are turning a messy email thread into a 3-line brief for a manager<br>who has not read the thread.</pre><pre>Hard rules:<br>- Use only facts in the thread. Do not invent decisions, owners,<br>  dates, costs, hours, or commitments.<br>- If a decision was not made in the thread (for example absorb vs<br>  re-quote), do not pick a side. Say the decision is open.<br>- If an owner or point person is unnamed or hedged (&quot;not sure who<br>  owns&quot;), write UNASSIGNED.<br>- If a date is absent, vague, or only a word like &quot;soon,&quot; &quot;this<br>  month,&quot; &quot;coming up,&quot; or &quot;when we have a new date,&quot; write NO DATE.<br>  Do not turn those words into a calendar date.<br>- If a cost, hour estimate, or other figure is absent, write MISSING.<br>- Line 3 may be a recommended next step ONLY if the thread already<br>  supports it (a person already offered to do it, or a question only<br>  the manager can close). Do not disguise a fake management decision<br>  as a &quot;recommendation.&quot;<br>- If the next move requires a choice the thread never made, write<br>  why the manager must decide, and list only options the thread named.</pre><pre>Return markdown with exactly these sections:</pre><pre>## 3-line brief<br>1. Background: what started this, in one or two sentences, facts only.<br>2. Where it is stuck now: the open block, with UNASSIGNED / NO DATE /<br>   MISSING visible.<br>3. Recommended next step, OR why the manager must decide: one of<br>   these, not both dressed up as a decision already made.</pre><pre>## Gaps still open<br>- Bullet each UNASSIGNED owner, NO DATE item, and MISSING fact</pre><pre>## What I can do without a new decision<br>- Only actions someone in the thread already offered, or clarifying<br>  questions. If none, write &quot;None in the thread.&quot;</pre><pre>## Do not tell the client until<br>- Facts a reply would need that are still UNASSIGNED, NO DATE, or<br>  MISSING</pre><pre>Email thread:<br>[paste]</pre><p>Save it as a snippet. Same prompt every time a thread is too long to forward.</p><h3>What a healthy output looks like (illustrative)</h3><p>Not a live client file — an <strong>illustrative shape</strong> for the sample thread above.</p><p><strong>3-line brief (illustrative):</strong></p><ol><li><strong>Background.</strong> Client Morgan asked to add an extra usage report to this month’s delivery, needed before their review. Sam noted it was not quoted, offered to draft a scope note, and named two options: absorb or re-quote. The vendor said the data feed is delayed. Riley reported legal wording on the add-on is unconfirmed.</li><li><strong>Where it is stuck now.</strong> Absorb vs re-quote is still open. Vendor feed date, review date, and “this month’s delivery” date: NO DATE. Legal chaser and point person: UNASSIGNED. Cost or hours: MISSING.</li><li><strong>Why the manager must decide.</strong> The thread names two options and contains no choice. Confirming “yes” or a date to the client would also need a point person, which the thread does not name. A next step that does not pretend the choice is made: Sam can draft the already-offered scope note; the manager needs to pick absorb or re-quote before anyone confirms to Morgan.</li></ol><p><strong>Gaps still open (illustrative):</strong> point person and legal chaser UNASSIGNED; vendor, review, and delivery dates NO DATE; cost/hours MISSING; absorb vs re-quote undecided.</p><p><strong>What I can do without a new decision:</strong> Sam’s already-offered scope note — not a confirmation to the client, a point person, or a vendor date.</p><p><strong>Do not tell the client until:</strong> absorb vs re-quote is decided; a point person is named; and any date in a reply is no longer NO DATE.</p><p>Notice what stayed out: no “I recommend we absorb it,” no invented legal owner, no Friday slip, no review on the 20th, no fee, no “I’ll confirm to Morgan today.”</p><h3>Step 3: Human pass — five checks before you send the brief</h3><p>Do not ask the model to “make this more decisive” until the gaps are locked.</p><ol><li><strong>Fact parity</strong> — every owner, date, cost, and decision is in the thread or explicitly UNASSIGNED / NO DATE / MISSING.</li><li><strong>No comfort decision</strong> — absorb vs re-quote did not quietly become a recommendation.</li><li><strong>Three means three</strong> — line 3 is a supported next step or a manager decision, not a hidden fourth commitment.</li><li><strong>Vague time stays vague</strong> — “soon,” “this month,” and “coming up” are still NO DATE.</li><li><strong>Client-facing line stays honest</strong> — the brief does not invent a confirmation you could paste back to Morgan.</li></ol><p>If line 3 only works by inventing a choice, rewrite it as “why the manager must decide.”</p><h3>Step 4: How to use the brief in the actual conversation</h3><p>Send the three lines first, then the gaps list if your manager wants the detail. Do not paste the forty messages underneath unless they ask. If they pick absorb or re-quote, write that decision down yourself. If they ask “what’s the date?” and the thread still says NO DATE, say NO DATE. After the meeting, one human reply should carry the real decision, the real point person if one was named, and nothing that is still blank.</p><p>A brief that already “confirms” a date to the client is not a brief. It is an unauthorized reply.</p><h3>How this differs from “just ask AI to summarize the thread”</h3><p>A soft “summarize this email” returns five polite paragraphs, a guessed owner, a made-up Friday, and a recommendation that feels like leadership. A 3-line manager brief is short on purpose: UNASSIGNED / NO DATE / MISSING stay visible, Sam’s offered scope note stays separate from what only the manager can close, and the client does not get an answer the thread does not contain.</p><p>A useful brief is not a shorter memo. It is three lines plus an honest gap list.</p><h3>Common failure modes the prompt should catch</h3><p>A fake choice (“we should absorb this”); a named legal owner the thread never used; “soon” turned into Friday; a fee or hour count; “Riley will be the point person” when Riley only flagged a gap; a client-ready confirmation in the brief.</p><p>If any of those appear, delete the cell. Re-run with the gap table pinned at the top of the paste.</p><h3>Five mistakes that fake a “crisp” update</h3><p><strong>1. Inventing a decision so the brief looks decisive.</strong> Picking absorb vs re-quote for your manager is a decision you did not have authority to make.</p><p><strong>2. Naming an owner who is not in the thread.</strong> “Not sure who owns chasing that” stays UNASSIGNED.</p><p><strong>3. Turning “soon” or “coming up” into a calendar date.</strong> Anxiety and a month word are not a date. If you need a date, ask a human.</p><p><strong>4. Dressing a management choice up as a “recommended next step.”</strong> Sam drafting a scope note is supported. “We should absorb the work and confirm to Morgan” is not, unless the thread already says so.</p><p><strong>5. Skipping the “do not tell the client until” list.</strong> Without it, someone will send a confident reply and discover the date and the point person were never real.</p><h3>Make twelve minutes sacred before standup</h3><p>Keep the 3-line-brief prompt in snippets. When a thread is too long to forward, run gap table → prompt → human pass before you message your manager. The model is a compression engine, not a decision-maker.</p><p>When someone says “just catch me up,” you already have three lines and a visible gap list — with MISSING, UNASSIGNED, and NO DATE still showing until a human fills them.</p><p>— -</p><p>*Next messy thread: will you send a fluent memo that invents a decision, an owner, and a Friday — or a three-line brief that stays honest about where the work is stuck?*</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad4d306f9efa" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>