{"course":{"slug":"ai-careers","title":"Working in AI","url":"https://addaly.com/learn/ai-careers","level":"beginner","track":"work"},"lesson":{"slug":"writing-for-people-who-decide","title":"The one-pager that gets a decision","url":"https://addaly.com/learn/ai-careers/writing-for-people-who-decide","position":84,"of":85,"minutes":9,"answer":"Lead with the recommendation, its cost and the decision needed by when, and attach to every number its definition, date, sample and uncertainty, because stating uncertainty is what makes a reader trust the rest.","keyIdea":"Lead with the recommendation, its cost and the decision needed by when, and attach to every number its definition, date, sample and uncertainty, because stating uncertainty is what makes a reader trust the rest."},"body":{"format":"markdown","text":"## Analysis is not a decision\n\nA great deal of good work dies in a document that describes what was found and never says what to do. The reader (busy, three levels up, holding a budget) reaches the end without knowing what is being asked of them, and files it.\n\nThe skill that converts technical work into influence is the ability to write a page that produces a decision. It is learnable, it is rarer than it should be, and it is the most reliable route from doing good work to being trusted with more of it.\n\n## Put the ask first\n\nTechnical writing is trained backwards: method, results, then conclusion. Decision writing inverts it, because the reader may stop at any point and you want them to have the answer before they do.\n\nCompare the same work in two openings.\n\n> \"We analysed churn across the last four quarterly cohorts using a survival model, controlling for plan type and acquisition channel. The results indicate…\"\n\nNobody above your manager reads past that.\n\n> \"**Recommendation:** move the annual renewal reminder from 7 days before expiry to 30 days. Expected effect: 1.5 to 3 percentage points of annual churn, worth ₹40 to ₹80 lakh a year. Cost: about two engineering weeks. **Decision needed by 20 October** to make the January cohort.\"\n\nEverything after that is evidence for a reader who wants it. Both documents contain the same work; only one of them will be acted on.\n\n## The structure\n\n1. **Recommendation.** One or two sentences.\n2. **What it costs and what it is worth**, with numbers.\n3. **The decision needed, from whom, by when.**\n4. **The evidence**, briefly, with the strongest thing first.\n5. **What would change my mind**, or what we are uncertain about.\n6. **Alternatives considered and why they were rejected**: this is what separates a recommendation from an opinion.\n7. **Appendix** for the method, the queries, the caveats.\n\nOne page above the appendix. If it will not fit, the recommendation is not clear enough yet.\n\n> Figure: One page, in this order. Layers, top to bottom: 1. Recommendation (one or two sentences); 2. What it costs and what it is worth (with numbers: 1.5 to 3 points of churn, 40 to 80 lakh, two engineering weeks); 3. The decision needed, from whom, by when (decision needed by 20 October); 4. The evidence, strongest thing first (each number with its definition, date, sample and uncertainty); 5. What would change my mind (or what we are uncertain about); 6. Alternatives considered and why rejected (a recommendation, not an opinion); 7. Appendix (method, queries, caveats). The reader may stop at any point, so the ask comes first. Section six is what separates a recommendation from an opinion, and if the page will not fit above the appendix the recommendation is not clear enough yet.\n\n## Number discipline\n\nEvery number in a decision document carries four things: what it measures, when, on what sample, and how uncertain it is.\n\n> Churn is 12%.\n\nversus\n\n> Churn was **12.1%** for the January cohort (n=400, ±3 points at 95%), where churn means no renewal within 30 days of contract end.\n\nThe second is longer and it is the one people trust, because a reader can tell what would break it. Stating uncertainty raises your credibility rather than lowering it: you have shown you know the difference between what you measured and what you know, which is exactly what a decision-maker is trying to establish about you.\n\n## Anticipate three questions\n\nEvery senior reader asks the same three, so answer them before they do:\n\n- **How do we know?** One line on the data and its limits.\n- **What does it cost?** Money, time and opportunity.\n- **What happens if we do nothing?** The most commonly omitted section and often the most persuasive one.\n\n## The meeting\n\nA pre-read circulated 24 hours in advance beats a presentation, because people can think. Some organisations go further and read the document silently at the start of the meeting (a practice Amazon is known for) precisely because slides let a weak argument sound strong.\n\nIf you must present: the first slide is the recommendation, the last slide is the recommendation, and the middle is evidence you may skip if the room has already agreed. Never build to a conclusion. There is no suspense in a budget meeting.\n\n## Email, briefly\n\nSame rules, smaller. The **subject line contains the ask**: \"Decision needed by Friday: renewal reminder timing\". The first line repeats it. The detail follows. If there is a deadline, it appears twice.\n\n## The limitation\n\nGood writing does not defeat organisational politics. A perfectly argued memo proposing something a director does not want will sit unanswered, and you will not be told why.\n\nTwo things partly compensate. **Know whose decision it actually is** before writing: a document addressed to nobody in particular is decided by nobody in particular. And **find out what that person is measured on**, because a recommendation framed against their target is a different document from one framed against yours, even when the analysis is identical."},"check":{"question":"An analyst's memo opens with three paragraphs of methodology before reaching a recommendation on the final page. Why does this reliably fail with senior readers?","options":[{"text":"Senior readers do not understand statistical methods and are alienated by them.","correct":false,"tempting":"Assumes incomprehension when the issue is attention and ordering; the same readers will read a method section placed in an appendix."},{"text":"Methodology sections invite challenges to the analysis that distract from the conclusion.","correct":false,"tempting":"Suggests concealing the method, which is what destroys trust when the challenge eventually comes."},{"text":"A reader may stop at any point, so a document that withholds the ask until the end is one that most readers finish without knowing what is being requested of them.","correct":true},{"text":"The document is too long, and cutting it to one page would solve the problem.","correct":false,"tempting":"Fixes the symptom; a one-page document that still buries the recommendation on the last line fails the same way."}]},"prev":{"slug":"the-long-game","title":"The second job, the plateau, and the ten-year view","url":"https://addaly.com/learn/ai-careers/the-long-game"},"next":{"slug":"promotion-on-evidence","title":"Getting promoted, and the case somebody has to make","url":"https://addaly.com/learn/ai-careers/promotion-on-evidence"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'The one-pager that gets a decision', in Working in AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-careers/writing-for-people-who-decide (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"The one-pager that gets a decision","courseTitle":"Working in AI","module":{"index":9,"slug":"staying","title":"Staying: the first ninety days and the ten years after"},"answer":null,"keyIdea":"Lead with the recommendation, its cost and the decision needed by when, and attach to every number its definition, date, sample and uncertainty, because stating uncertainty is what makes a reader trust the rest.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"stack","title":"One page, in this order","layers":[{"label":"1. Recommendation","detail":"one or two sentences"},{"label":"2. What it costs and what it is worth","detail":"with numbers: 1.5 to 3 points of churn, 40 to 80 lakh, two engineering weeks"},{"label":"3. The decision needed, from whom, by when","detail":"decision needed by 20 October"},{"label":"4. The evidence, strongest thing first","detail":"each number with its definition, date, sample and uncertainty"},{"label":"5. What would change my mind","detail":"or what we are uncertain about"},{"label":"6. Alternatives considered and why rejected","detail":"a recommendation, not an opinion"},{"label":"7. Appendix","detail":"method, queries, caveats"}],"caption":"The reader may stop at any point, so the ask comes first. Section six is what separates a recommendation from an opinion, and if the page will not fit above the appendix the recommendation is not clear enough yet."},"alt":"Figure: One page, in this order. Layers, top to bottom: 1. Recommendation (one or two sentences); 2. What it costs and what it is worth (with numbers: 1.5 to 3 points of churn, 40 to 80 lakh, two engineering weeks); 3. The decision needed, from whom, by when (decision needed by 20 October); 4. The evidence, strongest thing first (each number with its definition, date, sample and uncertainty); 5. What would change my mind (or what we are uncertain about); 6. Alternatives considered and why rejected (a recommendation, not an opinion); 7. Appendix (method, queries, caveats). The reader may stop at any point, so the ask comes first. Section six is what separates a recommendation from an opinion, and if the page will not fit above the appendix the recommendation is not clear enough yet."}},"blocks":[{"t":"h2","id":"analysis-is-not-a-decision","runs":[{"text":"Analysis is not a decision"}]},{"t":"p","runs":[{"text":"A great deal of good work dies in a document that describes what was found and never says what to do. The reader (busy, three levels up, holding a budget) reaches the end without knowing what is being asked of them, and files it."}]},{"t":"p","runs":[{"text":"The skill that converts technical work into influence is the ability to write a page that produces a decision. It is learnable, it is rarer than it should be, and it is the most reliable route from doing good work to being trusted with more of it."}]},{"t":"h2","id":"put-the-ask-first","runs":[{"text":"Put the ask first"}]},{"t":"p","runs":[{"text":"Technical writing is trained backwards: method, results, then conclusion. Decision writing inverts it, because the reader may stop at any point and you want them to have the answer before they do."}]},{"t":"p","runs":[{"text":"Compare the same work in two openings."}]},{"t":"quote","runs":[{"text":"“We analysed churn across the last four quarterly cohorts using a survival model, controlling for plan type and acquisition channel. The results indicate…”"}]},{"t":"p","runs":[{"text":"Nobody above your manager reads past that."}]},{"t":"quote","runs":[{"text":"“"},{"text":"Recommendation:","b":true},{"text":" move the annual renewal reminder from 7 days before expiry to 30 days. Expected effect: 1.5 to 3 percentage points of annual churn, worth ₹40 to ₹80 lakh a year. Cost: about two engineering weeks. "},{"text":"Decision needed by 20 October","b":true},{"text":" to make the January cohort.”"}]},{"t":"p","runs":[{"text":"Everything after that is evidence for a reader who wants it. Both documents contain the same work; only one of them will be acted on."}]},{"t":"h2","id":"the-structure","runs":[{"text":"The structure"}]},{"t":"ol","start":1,"items":[[{"text":"Recommendation.","b":true},{"text":" One or two sentences."}],[{"text":"What it costs and what it is worth","b":true},{"text":", with numbers."}],[{"text":"The decision needed, from whom, by when.","b":true}],[{"text":"The evidence","b":true},{"text":", briefly, with the strongest thing first."}],[{"text":"What would change my mind","b":true},{"text":", or what we are uncertain about."}],[{"text":"Alternatives considered and why they were rejected","b":true},{"text":": this is what separates a recommendation from an opinion."}],[{"text":"Appendix","b":true},{"text":" for the method, the queries, the caveats."}]]},{"t":"p","runs":[{"text":"One page above the appendix. If it will not fit, the recommendation is not clear enough yet."}]},{"t":"h2","id":"number-discipline","runs":[{"text":"Number discipline"}]},{"t":"p","runs":[{"text":"Every number in a decision document carries four things: what it measures, when, on what sample, and how uncertain it is."}]},{"t":"quote","runs":[{"text":"Churn is 12%."}]},{"t":"p","runs":[{"text":"versus"}]},{"t":"quote","runs":[{"text":"Churn was "},{"text":"12.1%","b":true},{"text":" for the January cohort (n=400, ±3 points at 95%), where churn means no renewal within 30 days of contract end."}]},{"t":"p","runs":[{"text":"The second is longer and it is the one people trust, because a reader can tell what would break it. Stating uncertainty raises your credibility rather than lowering it: you have shown you know the difference between what you measured and what you know, which is exactly what a decision-maker is trying to establish about you."}]},{"t":"h2","id":"anticipate-three-questions","runs":[{"text":"Anticipate three questions"}]},{"t":"p","runs":[{"text":"Every senior reader asks the same three, so answer them before they do:"}]},{"t":"ul","items":[[{"text":"How do we know?","b":true},{"text":" One line on the data and its limits."}],[{"text":"What does it cost?","b":true},{"text":" Money, time and opportunity."}],[{"text":"What happens if we do nothing?","b":true},{"text":" The most commonly omitted section and often the most persuasive one."}]]},{"t":"h2","id":"the-meeting","runs":[{"text":"The meeting"}]},{"t":"p","runs":[{"text":"A pre-read circulated 24 hours in advance beats a presentation, because people can think. Some organisations go further and read the document silently at the start of the meeting (a practice Amazon is known for) precisely because slides let a weak argument sound strong."}]},{"t":"p","runs":[{"text":"If you must present: the first slide is the recommendation, the last slide is the recommendation, and the middle is evidence you may skip if the room has already agreed. Never build to a conclusion. There is no suspense in a budget meeting."}]},{"t":"h2","id":"email-briefly","runs":[{"text":"Email, briefly"}]},{"t":"p","runs":[{"text":"Same rules, smaller. The "},{"text":"subject line contains the ask","b":true},{"text":": “Decision needed by Friday: renewal reminder timing”. The first line repeats it. The detail follows. If there is a deadline, it appears twice."}]},{"t":"h2","id":"the-limitation","runs":[{"text":"The limitation"}],"role":"limit"},{"t":"p","runs":[{"text":"Good writing does not defeat organisational politics. A perfectly argued memo proposing something a director does not want will sit unanswered, and you will not be told why."}]},{"t":"p","runs":[{"text":"Two things partly compensate. "},{"text":"Know whose decision it actually is","b":true},{"text":" before writing: a document addressed to nobody in particular is decided by nobody in particular. And "},{"text":"find out what that person is measured on","b":true},{"text":", because a recommendation framed against their target is a different document from one framed against yours, even when the analysis is identical."}]}],"check":{"question":[{"text":"An analyst’s memo opens with three paragraphs of methodology before reaching a recommendation on the final page. Why does this reliably fail with senior readers?"}],"options":[{"text":[{"text":"Senior readers do not understand statistical methods and are alienated by them."}],"correct":false,"tempting":[{"text":"Assumes incomprehension when the issue is attention and ordering; the same readers will read a method section placed in an appendix."}]},{"text":[{"text":"Methodology sections invite challenges to the analysis that distract from the conclusion."}],"correct":false,"tempting":[{"text":"Suggests concealing the method, which is what destroys trust when the challenge eventually comes."}]},{"text":[{"text":"A reader may stop at any point, so a document that withholds the ask until the end is one that most readers finish without knowing what is being requested of them."}],"correct":true,"tempting":null},{"text":[{"text":"The document is too long, and cutting it to one page would solve the problem."}],"correct":false,"tempting":[{"text":"Fixes the symptom; a one-page document that still buries the recommendation on the last line fails the same way."}]}],"order":[3,0,2,1],"why":null},"prev":{"slug":"the-long-game","title":"The second job, the plateau, and the ten-year view","minutes":8},"next":{"slug":"promotion-on-evidence","title":"Getting promoted, and the case somebody has to make","minutes":9}}}